| // Copyright (c) Facebook, Inc. and its affiliates. |
| // All rights reserved. |
| // |
| // Copyright 2019 Google LLC |
| // |
| // This source code is licensed under the BSD-style license found in the |
| // LICENSE file in the root directory of this source tree. |
| // |
| // Auto-generated file. Do not edit! |
| // Specification: test/f32-gemm.yaml |
| // Generator: tools/generate-gemm-test.py |
| |
| |
| #include <gtest/gtest.h> |
| |
| #include <xnnpack/common.h> |
| #include <xnnpack/isa-checks.h> |
| |
| #include <xnnpack/gemm.h> |
| #include <xnnpack/igemm.h> |
| #include <xnnpack/ppmm.h> |
| #include "gemm-microkernel-tester.h" |
| |
| |
| #if XNN_ARCH_ARM64 |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(8) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(8) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(8) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_8_subtile_m) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_8_subtile_n) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(16) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_16_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(16) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_16_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(16) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, k_lt_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, k_lt_16_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, k_lt_16_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, k_gt_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, k_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, k_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, k_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, k_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(83) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, k_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, n_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, qmin) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(8) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, qmax) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(8) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A57, strided_cm) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(8) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a57); |
| } |
| #endif // XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM64 |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(8) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(8) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(8) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_8_subtile_m) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_8_subtile_n) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(16) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_16_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(16) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_16_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(16) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, k_lt_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, k_lt_16_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, k_lt_16_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, k_gt_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, k_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, k_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, k_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, k_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(83) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, k_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, n_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, qmin) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(8) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, qmax) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(8) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_1X8__AARCH64_NEONFMA_CORTEX_A75, strided_cm) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(8) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__aarch64_neonfma_cortex_a75); |
| } |
| #endif // XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM64 |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(8) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(8) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(8) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_8_subtile_m) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_8_subtile_n) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(16) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_16_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(16) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_16_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(16) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, k_lt_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, k_lt_16_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, k_lt_16_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, k_gt_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, k_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, k_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, k_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, k_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(83) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, k_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, n_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, qmin) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(8) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, qmax) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(8) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A57, strided_cm) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(8) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a57); |
| } |
| #endif // XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM64 |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(8) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(8) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(8) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_8_subtile_m) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_8_subtile_n) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(16) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_16_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(16) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_16_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(16) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, k_lt_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, k_lt_16_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, k_lt_16_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, k_gt_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, k_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, k_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, k_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, k_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(83) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, k_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, n_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, qmin) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(8) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, qmax) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(8) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_CORTEX_A75, strided_cm) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(8) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_cortex_a75); |
| } |
| #endif // XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM64 |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(8) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(8) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(8) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 5; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_8_subtile_m) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 5; m++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_8_subtile_n) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(n) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(16) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_16_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(16) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_16_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 5; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(16) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, k_lt_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, k_lt_16_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, k_lt_16_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| for (uint32_t m = 1; m <= 5; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, k_gt_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, k_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, k_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| for (uint32_t m = 1; m <= 5; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, k_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, k_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .a_stride(83) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, k_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| for (uint32_t m = 1; m <= 5; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(n) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 5; m++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, n_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(n) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 5; m++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 5; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, qmin) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(8) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, qmax) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(8) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_5X8__AARCH64_NEONFMA_CORTEX_A75, strided_cm) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(8) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_5x8__aarch64_neonfma_cortex_a75); |
| } |
| #endif // XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM64 |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(8) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(8) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(8) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_8_subtile_m) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_8_subtile_n) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(16) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_16_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(16) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, k_eq_16_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(16) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, k_lt_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, k_lt_16_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, k_lt_16_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, k_gt_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, k_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, k_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, k_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, k_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(83) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, k_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, n_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, qmin) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(8) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, qmax) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(8) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A57, strided_cm) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(8) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a57); |
| } |
| #endif // XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM64 |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, k_eq_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(8) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(8) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, k_eq_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(8) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, k_eq_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, k_eq_8_subtile_m) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, k_eq_8_subtile_n) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, k_eq_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(16) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, k_eq_16_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(16) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, k_eq_16_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(16) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, k_lt_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, k_lt_16_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, k_lt_16_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, k_gt_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, k_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, k_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, k_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, k_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(83) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, k_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, n_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, qmin) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(8) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, qmax) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(8) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A73, strided_cm) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(8) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a73); |
| } |
| #endif // XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM64 |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(8) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(8) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(8) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_8_subtile_m) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_8_subtile_n) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(16) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_16_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(16) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, k_eq_16_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(16) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, k_lt_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, k_lt_16_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, k_lt_16_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 16; k++) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, k_gt_16) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, k_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(19) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, k_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 17; k < 16; k++) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, k_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, k_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(83) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, k_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 24; k <= 80; k += 8) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, n_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k <= 40; k += 9) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, qmin) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(8) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, qmax) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(8) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_CORTEX_A75, strided_cm) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(8) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_cortex_a75); |
| } |
| #endif // XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM64 |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, k_eq_4) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(12) |
| .k(4) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(12) |
| .k(4) |
| .cn_stride(17) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, k_eq_4_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(12) |
| .k(4) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, k_eq_4_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, k_eq_4_subtile_m) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(12) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, k_eq_4_subtile_n) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, k_eq_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(12) |
| .k(8) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, k_eq_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(12) |
| .k(8) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, k_eq_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, k_lt_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(12) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, k_lt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(12) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, k_lt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 8; k++) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, k_gt_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 9; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(12) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, k_gt_4_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 9; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(12) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, k_gt_4_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 9; k < 8; k++) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, k_div_4) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 12; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(12) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, k_div_4_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 12; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(12) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, k_div_4_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 12; k <= 40; k += 4) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, n_gt_12) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 13; n < 24; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(12) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, n_gt_12_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 13; n < 24; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(12) |
| .k(k) |
| .cn_stride(17) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, n_gt_12_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 13; n < 24; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, n_gt_12_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 13; n < 24; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, n_div_12) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 24; n <= 36; n += 12) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(12) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, n_div_12_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 24; n <= 36; n += 12) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .cn_stride(17) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, n_div_12_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 24; n <= 36; n += 12) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, n_div_12_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 24; n <= 36; n += 12) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(17) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, qmin) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(12) |
| .k(4) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, qmax) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(12) |
| .k(4) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| |
| TEST(F32_GEMM_1X12__AARCH64_NEONFMA_CORTEX_A53, strided_cm) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(12) |
| .k(4) |
| .cm_stride(17) |
| .Test(xnn_f32_gemm_ukernel_1x12__aarch64_neonfma_cortex_a53); |
| } |
| #endif // XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM64 |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, k_eq_4) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(4) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(4) |
| .cn_stride(17) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, k_eq_4_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(4) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, k_eq_4_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, k_eq_4_subtile_m) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(12) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, k_eq_4_subtile_n) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, k_eq_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(8) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, k_eq_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(8) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, k_eq_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(8) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, k_lt_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, k_lt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, k_lt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 8; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, k_gt_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 9; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, k_gt_4_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 9; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, k_gt_4_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 9; k < 8; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, k_div_4) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 12; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, k_div_4_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 12; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, k_div_4_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 12; k <= 40; k += 4) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, n_gt_12) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 13; n < 24; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, n_gt_12_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 13; n < 24; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .cn_stride(17) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, n_gt_12_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 13; n < 24; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, n_gt_12_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 13; n < 24; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, n_div_12) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 24; n <= 36; n += 12) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, n_div_12_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 24; n <= 36; n += 12) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(17) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, n_div_12_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 24; n <= 36; n += 12) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, n_div_12_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 24; n <= 36; n += 12) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(17) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, qmin) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(4) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, qmax) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(4) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| |
| TEST(F32_GEMM_4X12__AARCH64_NEONFMA_CORTEX_A53, strided_cm) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(4) |
| .cm_stride(17) |
| .Test(xnn_f32_gemm_ukernel_4x12__aarch64_neonfma_cortex_a53); |
| } |
| #endif // XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM64 |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, k_eq_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(2) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(2) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, k_eq_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(2) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, k_eq_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, k_eq_2_subtile_m) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, k_eq_2_subtile_n) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, k_lt_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, k_lt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, k_lt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 2; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, k_gt_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, k_gt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, k_gt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 3; k < 4; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, k_div_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, k_div_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, k_div_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 4; k <= 20; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, n_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, qmin) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(2) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, qmax) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(2) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD64, strided_cm) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(2) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld64); |
| } |
| #endif // XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM64 |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, k_eq_4) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, k_eq_4_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, k_eq_4_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, k_eq_4_subtile_m) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, k_eq_4_subtile_n) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, k_lt_4) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, k_lt_4_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, k_lt_4_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 4; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, k_gt_4) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, k_gt_4_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, k_gt_4_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 5; k < 8; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, k_div_4) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, k_div_4_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, k_div_4_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 8; k <= 40; k += 4) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, n_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, qmin) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, qmax) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| |
| TEST(F32_GEMM_4X8__AARCH64_NEONFMA_LD128, strided_cm) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__aarch64_neonfma_ld128); |
| } |
| #endif // XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM64 |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, k_eq_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(2) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(2) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, k_eq_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(2) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, k_eq_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, k_eq_2_subtile_m) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, k_eq_2_subtile_n) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, k_lt_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, k_lt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, k_lt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 2; k++) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, k_gt_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, k_gt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, k_gt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 3; k < 4; k++) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, k_div_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, k_div_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, k_div_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 4; k <= 20; k += 2) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, n_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, qmin) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(2) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, qmax) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(2) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD64, strided_cm) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(2) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64); |
| } |
| #endif // XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM64 |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, k_eq_4) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(4) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(4) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, k_eq_4_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(4) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, k_eq_4_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, k_eq_4_subtile_m) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, k_eq_4_subtile_n) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, k_lt_4) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, k_lt_4_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, k_lt_4_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 4; k++) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, k_gt_4) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, k_gt_4_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, k_gt_4_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 5; k < 8; k++) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, k_div_4) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, k_div_4_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, k_div_4_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 8; k <= 40; k += 4) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, n_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, qmin) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(4) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, qmax) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(4) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| |
| TEST(F32_GEMM_6X8__AARCH64_NEONFMA_LD128, strided_cm) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(4) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld128); |
| } |
| #endif // XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(F32_GEMM_4X12__NEON_LD64, k_eq_2) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(2) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(2) |
| .cn_stride(17) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, k_eq_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(2) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, k_eq_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, k_eq_2_subtile_m) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(12) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, k_eq_2_subtile_n) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, k_lt_2) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, k_lt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, k_lt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k < 2; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, k_gt_2) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, k_gt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, k_gt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 3; k < 4; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, k_div_2) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, k_div_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, k_div_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 4; k <= 20; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, n_gt_12) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 13; n < 24; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, n_gt_12_strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 13; n < 24; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .cn_stride(17) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, n_gt_12_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 13; n < 24; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, n_gt_12_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 13; n < 24; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, n_div_12) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 24; n <= 36; n += 12) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, n_div_12_strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 24; n <= 36; n += 12) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(17) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, n_div_12_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 24; n <= 36; n += 12) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, n_div_12_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 24; n <= 36; n += 12) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(17) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, qmin) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(2) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, qmax) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(2) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_4X12__NEON_LD64, strided_cm) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(2) |
| .cm_stride(17) |
| .Test(xnn_f32_gemm_ukernel_4x12__neon_ld64); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(F32_GEMM_1X8__NEON_LD64, k_eq_2) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(2) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(2) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, k_eq_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(2) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, k_eq_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, k_eq_2_subtile_m) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, k_eq_2_subtile_n) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, k_lt_2) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, k_lt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, k_lt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k < 2; k++) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, k_gt_2) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, k_gt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, k_gt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 3; k < 4; k++) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, k_div_2) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, k_div_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, k_div_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 4; k <= 20; k += 2) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, n_div_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, qmin) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(2) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, qmax) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(2) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_1X8__NEON_LD64, strided_cm) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(2) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__neon_ld64); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(F32_GEMM_4X2__NEON_LD64, k_eq_2) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(2) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(2) |
| .cn_stride(5) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, k_eq_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(2) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, k_eq_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 2; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, k_eq_2_subtile_m) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(2) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, k_eq_2_subtile_n) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 1; n <= 2; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, k_lt_2) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, k_lt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(k) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, k_lt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k < 2; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 2; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, k_gt_2) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, k_gt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, k_gt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 3; k < 4; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 2; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, k_div_2) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, k_div_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, k_div_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 4; k <= 20; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 2; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, n_gt_2) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 3; n < 4; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, n_gt_2_strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 3; n < 4; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(k) |
| .cn_stride(5) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, n_gt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 3; n < 4; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, n_gt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 3; n < 4; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, n_div_2) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 4; n <= 6; n += 2) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, n_div_2_strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 4; n <= 6; n += 2) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(5) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, n_div_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 4; n <= 6; n += 2) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, n_div_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 4; n <= 6; n += 2) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 2; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(5) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, qmin) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(2) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, qmax) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(2) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_4X2__NEON_LD64, strided_cm) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(2) |
| .cm_stride(5) |
| .Test(xnn_f32_gemm_ukernel_4x2__neon_ld64); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(F32_GEMM_4X8__NEON_LD64, k_eq_2) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(2) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(2) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, k_eq_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(2) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, k_eq_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, k_eq_2_subtile_m) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, k_eq_2_subtile_n) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, k_lt_2) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, k_lt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, k_lt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k < 2; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, k_gt_2) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, k_gt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, k_gt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 3; k < 4; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, k_div_2) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, k_div_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, k_div_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 4; k <= 20; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, n_div_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, qmin) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(2) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, qmax) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(2) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD64, strided_cm) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(2) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld64); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(F32_GEMM_4X8__NEON_LD128, k_eq_4) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, k_eq_4_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, k_eq_4_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, k_eq_4_subtile_m) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, k_eq_4_subtile_n) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, k_lt_4) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, k_lt_4_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, k_lt_4_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k < 4; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, k_gt_4) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, k_gt_4_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, k_gt_4_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 5; k < 8; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, k_div_4) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, k_div_4_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, k_div_4_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 8; k <= 40; k += 4) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, n_div_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, qmin) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, qmax) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| |
| TEST(F32_GEMM_4X8__NEON_LD128, strided_cm) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__neon_ld128); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(F32_GEMM_5X8__NEON_LD64, k_eq_2) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(2) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(2) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, k_eq_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(2) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, k_eq_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t m = 1; m <= 5; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, k_eq_2_subtile_m) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t m = 1; m <= 5; m++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, k_eq_2_subtile_n) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, k_lt_2) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, k_lt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, k_lt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k < 2; k++) { |
| for (uint32_t m = 1; m <= 5; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, k_gt_2) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, k_gt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, k_gt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 3; k < 4; k++) { |
| for (uint32_t m = 1; m <= 5; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, k_div_2) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, k_div_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, k_div_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 4; k <= 20; k += 2) { |
| for (uint32_t m = 1; m <= 5; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 5; m++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, n_div_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 5; m++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 5; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, qmin) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(2) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, qmax) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(2) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_5X8__NEON_LD64, strided_cm) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(2) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_5x8__neon_ld64); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(F32_GEMM_6X8__NEON_LD64, k_eq_2) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(2) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(2) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, k_eq_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(2) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, k_eq_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, k_eq_2_subtile_m) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, k_eq_2_subtile_n) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, k_lt_2) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, k_lt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, k_lt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k < 2; k++) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, k_gt_2) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, k_gt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, k_gt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 3; k < 4; k++) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, k_div_2) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, k_div_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, k_div_2_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 4; k <= 20; k += 2) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, n_div_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, qmin) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(2) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, qmax) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(2) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| |
| TEST(F32_GEMM_6X8__NEON_LD64, strided_cm) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(2) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__neon_ld64); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, k_eq_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(2) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(2) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, k_eq_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(2) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, k_eq_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, k_eq_2_subtile_m) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, k_eq_2_subtile_n) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, k_lt_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, k_lt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, k_lt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 2; k++) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, k_gt_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, k_gt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, k_gt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 3; k < 4; k++) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, k_div_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, k_div_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, k_div_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 4; k <= 20; k += 2) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, n_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, qmin) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(2) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, qmax) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(2) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_1X8__NEONFMA_LD64, strided_cm) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(2) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__neonfma_ld64); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, k_eq_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(2) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(2) |
| .cn_stride(17) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, k_eq_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(2) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, k_eq_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, k_eq_2_subtile_m) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(12) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, k_eq_2_subtile_n) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, k_lt_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, k_lt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, k_lt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 2; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, k_gt_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, k_gt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, k_gt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 3; k < 4; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, k_div_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, k_div_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, k_div_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 4; k <= 20; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, n_gt_12) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 13; n < 24; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, n_gt_12_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 13; n < 24; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .cn_stride(17) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, n_gt_12_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 13; n < 24; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, n_gt_12_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 13; n < 24; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, n_div_12) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 24; n <= 36; n += 12) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, n_div_12_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 24; n <= 36; n += 12) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(17) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, n_div_12_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 24; n <= 36; n += 12) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, n_div_12_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 24; n <= 36; n += 12) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 12; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(17) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, qmin) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(2) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, qmax) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(2) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_4X12__NEONFMA_LD64, strided_cm) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(12) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(12) |
| .k(2) |
| .cm_stride(17) |
| .Test(xnn_f32_gemm_ukernel_4x12__neonfma_ld64); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, k_eq_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(2) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(2) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, k_eq_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(2) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, k_eq_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, k_eq_2_subtile_m) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, k_eq_2_subtile_n) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, k_lt_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, k_lt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, k_lt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 2; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, k_gt_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, k_gt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, k_gt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 3; k < 4; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, k_div_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, k_div_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, k_div_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 4; k <= 20; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, n_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, qmin) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(2) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, qmax) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(2) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD64, strided_cm) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(2) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld64); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, k_eq_4) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, k_eq_4_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, k_eq_4_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, k_eq_4_subtile_m) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, k_eq_4_subtile_n) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, k_lt_4) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, k_lt_4_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, k_lt_4_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 4; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, k_gt_4) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, k_gt_4_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, k_gt_4_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 5; k < 8; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, k_div_4) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, k_div_4_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, k_div_4_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 8; k <= 40; k += 4) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, n_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, qmin) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, qmax) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| |
| TEST(F32_GEMM_4X8__NEONFMA_LD128, strided_cm) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__neonfma_ld128); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, k_eq_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(2) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(2) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, k_eq_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(2) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, k_eq_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 5; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, k_eq_2_subtile_m) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 5; m++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, k_eq_2_subtile_n) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, k_lt_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, k_lt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, k_lt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 2; k++) { |
| for (uint32_t m = 1; m <= 5; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, k_gt_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, k_gt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, k_gt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 3; k < 4; k++) { |
| for (uint32_t m = 1; m <= 5; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, k_div_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, k_div_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, k_div_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 4; k <= 20; k += 2) { |
| for (uint32_t m = 1; m <= 5; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 5; m++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, n_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 5; m++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 5; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, qmin) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(2) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, qmax) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(2) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_5X8__NEONFMA_LD64, strided_cm) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(5) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(5) |
| .n(8) |
| .k(2) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_5x8__neonfma_ld64); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, k_eq_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(2) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(2) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, k_eq_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(2) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, k_eq_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, k_eq_2_subtile_m) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, k_eq_2_subtile_n) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(2) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, k_lt_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, k_lt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 2; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(5) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, k_lt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k < 2; k++) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, k_gt_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, k_gt_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 3; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, k_gt_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 3; k < 4; k++) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, k_div_2) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, k_div_2_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 4; k <= 20; k += 2) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, k_div_2_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 4; k <= 20; k += 2) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, n_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .a_stride(13) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k <= 10; k += 3) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, qmin) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(2) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, qmax) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(2) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| |
| TEST(F32_GEMM_6X8__NEONFMA_LD64, strided_cm) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(2) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__neonfma_ld64); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| TEST(F32_GEMM_1X8__SSE_LOAD1, k_eq_1) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_load1); |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_LOAD1, strided_cn) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(1) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_load1); |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_LOAD1, k_eq_1_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(1) |
| .a_stride(3) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_load1); |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_LOAD1, k_eq_1_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_load1); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_LOAD1, k_eq_1_subtile_m) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_load1); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_LOAD1, k_eq_1_subtile_n) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_load1); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_LOAD1, k_gt_1) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_load1); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_LOAD1, k_gt_1_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_load1); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_LOAD1, k_gt_1_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 2; k < 10; k++) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_load1); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_LOAD1, n_gt_8) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_load1); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_LOAD1, n_gt_8_strided_cn) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_load1); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_LOAD1, n_gt_8_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_load1); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_LOAD1, n_gt_8_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_load1); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_LOAD1, n_div_8) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_load1); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_LOAD1, n_div_8_strided_cn) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_load1); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_LOAD1, n_div_8_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_load1); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_LOAD1, n_div_8_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_load1); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_LOAD1, strided_cm_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_load1); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_LOAD1, qmin) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(1) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_load1); |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_LOAD1, qmax) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(1) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_load1); |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_LOAD1, strided_cm) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(1) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_load1); |
| } |
| #endif // XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| |
| |
| #if XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| TEST(F32_GEMM_4X8__SSE_LOAD1, k_eq_1) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_load1); |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_LOAD1, strided_cn) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_load1); |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_LOAD1, k_eq_1_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .a_stride(3) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_load1); |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_LOAD1, k_eq_1_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_load1); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_LOAD1, k_eq_1_subtile_m) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_load1); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_LOAD1, k_eq_1_subtile_n) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_load1); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_LOAD1, k_gt_1) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_load1); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_LOAD1, k_gt_1_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_load1); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_LOAD1, k_gt_1_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 2; k < 10; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_load1); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_LOAD1, n_gt_8) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_load1); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_LOAD1, n_gt_8_strided_cn) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_load1); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_LOAD1, n_gt_8_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_load1); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_LOAD1, n_gt_8_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_load1); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_LOAD1, n_div_8) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_load1); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_LOAD1, n_div_8_strided_cn) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_load1); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_LOAD1, n_div_8_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_load1); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_LOAD1, n_div_8_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_load1); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_LOAD1, strided_cm_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_load1); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_LOAD1, qmin) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_load1); |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_LOAD1, qmax) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_load1); |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_LOAD1, strided_cm) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_load1); |
| } |
| #endif // XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| |
| |
| #if XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| TEST(F32_GEMM_1X8__SSE_DUP, k_eq_4) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(4) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, strided_cn) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(4) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, k_eq_4_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(4) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, k_eq_4_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, k_eq_4_subtile_m) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, k_eq_4_subtile_n) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, k_lt_4) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, k_lt_4_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, k_lt_4_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 1; k < 4; k++) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, k_gt_4) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, k_gt_4_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, k_gt_4_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 5; k < 8; k++) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, k_div_4) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, k_div_4_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, k_div_4_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 8; k <= 40; k += 4) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, n_gt_8) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, n_gt_8_strided_cn) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, n_gt_8_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, n_gt_8_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, n_div_8) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, n_div_8_strided_cn) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, n_div_8_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, n_div_8_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, strided_cm_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, qmin) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(4) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, qmax) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(4) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| |
| TEST(F32_GEMM_1X8__SSE_DUP, strided_cm) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(4) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__sse_dup); |
| } |
| #endif // XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| |
| |
| #if XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| TEST(F32_GEMM_4X8__SSE_DUP, k_eq_4) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, strided_cn) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, k_eq_4_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, k_eq_4_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, k_eq_4_subtile_m) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, k_eq_4_subtile_n) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, k_lt_4) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, k_lt_4_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, k_lt_4_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 1; k < 4; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, k_gt_4) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, k_gt_4_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, k_gt_4_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 5; k < 8; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, k_div_4) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, k_div_4_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, k_div_4_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 8; k <= 40; k += 4) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, n_gt_8) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, n_gt_8_strided_cn) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, n_gt_8_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, n_gt_8_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, n_div_8) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, n_div_8_strided_cn) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, n_div_8_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, n_div_8_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, strided_cm_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, qmin) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, qmax) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| |
| TEST(F32_GEMM_4X8__SSE_DUP, strided_cm) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__sse_dup); |
| } |
| #endif // XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| |
| |
| #if XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| TEST(F32_GEMM_1X8S4__SSE, k_eq_4) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(4) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, strided_cn) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(4) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, k_eq_4_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(4) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, k_eq_4_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, k_eq_4_subtile_m) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(8) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, k_eq_4_subtile_n) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, k_lt_4) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, k_lt_4_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, k_lt_4_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 1; k < 4; k++) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, k_gt_4) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, k_gt_4_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, k_gt_4_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 5; k < 8; k++) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, k_div_4) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, k_div_4_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, k_div_4_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 8; k <= 40; k += 4) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, n_gt_8) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, n_gt_8_strided_cn) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, n_gt_8_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, n_gt_8_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, n_div_8) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, n_div_8_strided_cn) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, n_div_8_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, n_div_8_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, strided_cm_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, qmin) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(4) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, qmax) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(4) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| |
| TEST(F32_GEMM_1X8S4__SSE, strided_cm) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(4) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__sse); |
| } |
| #endif // XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| |
| |
| #if XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| TEST(F32_GEMM_4X8S4__SSE, k_eq_4) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(4) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, strided_cn) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(4) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, k_eq_4_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(4) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, k_eq_4_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, k_eq_4_subtile_m) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(8) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, k_eq_4_subtile_n) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, k_lt_4) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, k_lt_4_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, k_lt_4_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 1; k < 4; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, k_gt_4) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, k_gt_4_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, k_gt_4_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 5; k < 8; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, k_div_4) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, k_div_4_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, k_div_4_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 8; k <= 40; k += 4) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, n_gt_8) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, n_gt_8_strided_cn) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, n_gt_8_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, n_gt_8_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, n_div_8) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, n_div_8_strided_cn) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, n_div_8_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, n_div_8_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, strided_cm_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, qmin) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(4) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, qmax) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(4) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| |
| TEST(F32_GEMM_4X8S4__SSE, strided_cm) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(4) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__sse); |
| } |
| #endif // XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| |
| |
| #if !XNN_ARCH_ASMJS && !XNN_ARCH_WASM |
| TEST(F32_GEMM_1X8__PSIMD_LOADSPLAT, k_eq_1) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_LOADSPLAT, strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(1) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_LOADSPLAT, k_eq_1_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(1) |
| .a_stride(3) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_LOADSPLAT, k_eq_1_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_LOADSPLAT, k_eq_1_subtile_m) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_LOADSPLAT, k_eq_1_subtile_n) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_LOADSPLAT, k_gt_1) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_LOADSPLAT, k_gt_1_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_LOADSPLAT, k_gt_1_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 2; k < 10; k++) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_LOADSPLAT, n_gt_8) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_LOADSPLAT, n_gt_8_strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_LOADSPLAT, n_gt_8_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_LOADSPLAT, n_gt_8_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_LOADSPLAT, n_div_8) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_LOADSPLAT, n_div_8_strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_LOADSPLAT, n_div_8_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_LOADSPLAT, n_div_8_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_LOADSPLAT, strided_cm_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_LOADSPLAT, qmin) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(1) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_LOADSPLAT, qmax) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(1) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_LOADSPLAT, strided_cm) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(1) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| #endif // !XNN_ARCH_ASMJS && !XNN_ARCH_WASM |
| |
| |
| #if !XNN_ARCH_ASMJS && !XNN_ARCH_WASM |
| TEST(F32_GEMM_4X8__PSIMD_LOADSPLAT, k_eq_1) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_LOADSPLAT, strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_LOADSPLAT, k_eq_1_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .a_stride(3) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_LOADSPLAT, k_eq_1_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_LOADSPLAT, k_eq_1_subtile_m) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_LOADSPLAT, k_eq_1_subtile_n) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_LOADSPLAT, k_gt_1) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_LOADSPLAT, k_gt_1_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_LOADSPLAT, k_gt_1_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 2; k < 10; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_LOADSPLAT, n_gt_8) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_LOADSPLAT, n_gt_8_strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_LOADSPLAT, n_gt_8_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_LOADSPLAT, n_gt_8_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_LOADSPLAT, n_div_8) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_LOADSPLAT, n_div_8_strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_LOADSPLAT, n_div_8_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_LOADSPLAT, n_div_8_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_LOADSPLAT, strided_cm_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_LOADSPLAT, qmin) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_LOADSPLAT, qmax) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_LOADSPLAT, strided_cm) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| #endif // !XNN_ARCH_ASMJS && !XNN_ARCH_WASM |
| |
| |
| #if !XNN_ARCH_ASMJS && !XNN_ARCH_WASM |
| TEST(F32_GEMM_6X8__PSIMD_LOADSPLAT, k_eq_1) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_LOADSPLAT, strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(1) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_LOADSPLAT, k_eq_1_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(1) |
| .a_stride(3) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_LOADSPLAT, k_eq_1_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_LOADSPLAT, k_eq_1_subtile_m) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_LOADSPLAT, k_eq_1_subtile_n) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_LOADSPLAT, k_gt_1) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_LOADSPLAT, k_gt_1_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_LOADSPLAT, k_gt_1_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 2; k < 10; k++) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_LOADSPLAT, n_gt_8) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_LOADSPLAT, n_gt_8_strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_LOADSPLAT, n_gt_8_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_LOADSPLAT, n_gt_8_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_LOADSPLAT, n_div_8) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_LOADSPLAT, n_div_8_strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_LOADSPLAT, n_div_8_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_LOADSPLAT, n_div_8_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_LOADSPLAT, strided_cm_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_LOADSPLAT, qmin) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(1) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_LOADSPLAT, qmax) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(1) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_LOADSPLAT, strided_cm) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(1) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_loadsplat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| #endif // !XNN_ARCH_ASMJS && !XNN_ARCH_WASM |
| |
| |
| #if !XNN_ARCH_ASMJS && !XNN_ARCH_WASM |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, k_eq_4) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(4) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(4) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, k_eq_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(4) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, k_eq_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, k_eq_4_subtile_m) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, k_eq_4_subtile_n) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, k_lt_4) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, k_lt_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, k_lt_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k < 4; k++) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, k_gt_4) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, k_gt_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, k_gt_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 5; k < 8; k++) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, k_div_4) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, k_div_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, k_div_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 8; k <= 40; k += 4) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, n_gt_8) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, n_gt_8_strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, n_gt_8_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, n_gt_8_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, n_div_8) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, n_div_8_strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, n_div_8_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, n_div_8_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, strided_cm_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, qmin) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(4) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, qmax) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(4) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_1X8__PSIMD_SPLAT, strided_cm) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(8) |
| .k(4) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| #endif // !XNN_ARCH_ASMJS && !XNN_ARCH_WASM |
| |
| |
| #if !XNN_ARCH_ASMJS && !XNN_ARCH_WASM |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, k_eq_4) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, k_eq_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, k_eq_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, k_eq_4_subtile_m) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, k_eq_4_subtile_n) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, k_lt_4) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, k_lt_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, k_lt_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k < 4; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, k_gt_4) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, k_gt_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, k_gt_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 5; k < 8; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, k_div_4) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, k_div_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, k_div_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 8; k <= 40; k += 4) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, n_gt_8) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, n_gt_8_strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, n_gt_8_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, n_gt_8_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, n_div_8) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, n_div_8_strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, n_div_8_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, n_div_8_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, strided_cm_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, qmin) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, qmax) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X8__PSIMD_SPLAT, strided_cm) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(4) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| #endif // !XNN_ARCH_ASMJS && !XNN_ARCH_WASM |
| |
| |
| #if !XNN_ARCH_ASMJS && !XNN_ARCH_WASM |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, k_eq_4) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(4) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(4) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, k_eq_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(4) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, k_eq_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, k_eq_4_subtile_m) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, k_eq_4_subtile_n) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, k_lt_4) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, k_lt_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, k_lt_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k < 4; k++) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, k_gt_4) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, k_gt_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, k_gt_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 5; k < 8; k++) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, k_div_4) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, k_div_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, k_div_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 8; k <= 40; k += 4) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, n_gt_8) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, n_gt_8_strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, n_gt_8_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, n_gt_8_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, n_div_8) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, n_div_8_strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, n_div_8_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, n_div_8_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, strided_cm_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, qmin) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(4) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, qmax) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(4) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_6X8__PSIMD_SPLAT, strided_cm) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(6) |
| .n(8) |
| .k(4) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8__psimd_splat, GemmMicrokernelTester::Variant::Scalar); |
| } |
| #endif // !XNN_ARCH_ASMJS && !XNN_ARCH_WASM |
| |
| |
| #if !XNN_ARCH_ASMJS && !XNN_ARCH_WASM |
| TEST(F32_GEMM_1X8S4__PSIMD, k_eq_4) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(4) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(4) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, k_eq_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(4) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, k_eq_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, k_eq_4_subtile_m) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(8) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, k_eq_4_subtile_n) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, k_lt_4) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, k_lt_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, k_lt_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k < 4; k++) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, k_gt_4) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, k_gt_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, k_gt_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 5; k < 8; k++) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, k_div_4) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, k_div_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, k_div_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 8; k <= 40; k += 4) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, n_gt_8) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, n_gt_8_strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, n_gt_8_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, n_gt_8_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, n_div_8) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, n_div_8_strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, n_div_8_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, n_div_8_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, strided_cm_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, qmin) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(4) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, qmax) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(4) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_1X8S4__PSIMD, strided_cm) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(1) |
| .n(8) |
| .k(4) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| #endif // !XNN_ARCH_ASMJS && !XNN_ARCH_WASM |
| |
| |
| #if !XNN_ARCH_ASMJS && !XNN_ARCH_WASM |
| TEST(F32_GEMM_4X8S4__PSIMD, k_eq_4) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(4) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(4) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, k_eq_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(4) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, k_eq_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, k_eq_4_subtile_m) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(8) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, k_eq_4_subtile_n) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, k_lt_4) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, k_lt_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, k_lt_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k < 4; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, k_gt_4) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, k_gt_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, k_gt_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 5; k < 8; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, k_div_4) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, k_div_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, k_div_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 8; k <= 40; k += 4) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, n_gt_8) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, n_gt_8_strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, n_gt_8_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, n_gt_8_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, n_div_8) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, n_div_8_strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, n_div_8_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, n_div_8_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, strided_cm_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, qmin) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(4) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, qmax) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(4) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X8S4__PSIMD, strided_cm) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(4) |
| .n(8) |
| .k(4) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| #endif // !XNN_ARCH_ASMJS && !XNN_ARCH_WASM |
| |
| |
| #if !XNN_ARCH_ASMJS && !XNN_ARCH_WASM |
| TEST(F32_GEMM_6X8S4__PSIMD, k_eq_4) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(6) |
| .n(8) |
| .k(4) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(6) |
| .n(8) |
| .k(4) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, k_eq_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(6) |
| .n(8) |
| .k(4) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, k_eq_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, k_eq_4_subtile_m) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(8) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, k_eq_4_subtile_n) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(6) |
| .n(n) |
| .k(4) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, k_lt_4) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, k_lt_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k < 4; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, k_lt_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k < 4; k++) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, k_gt_4) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, k_gt_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 5; k < 8; k++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, k_gt_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 5; k < 8; k++) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, k_div_4) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, k_div_4_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 8; k <= 40; k += 4) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(6) |
| .n(8) |
| .k(k) |
| .a_stride(43) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, k_div_4_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 8; k <= 40; k += 4) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, n_gt_8) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, n_gt_8_strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(6) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, n_gt_8_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(6) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, n_gt_8_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, n_div_8) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(6) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, n_div_8_strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(6) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, n_div_8_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(6) |
| .n(n) |
| .k(k) |
| .a_stride(23) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, n_div_8_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, strided_cm_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k <= 20; k += 5) { |
| for (uint32_t m = 1; m <= 6; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, qmin) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(6) |
| .n(8) |
| .k(4) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, qmax) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(6) |
| .n(8) |
| .k(4) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_6X8S4__PSIMD, strided_cm) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(6) |
| .nr(8) |
| .kr(1) |
| .sr(4) |
| .m(6) |
| .n(8) |
| .k(4) |
| .cm_stride(11) |
| .Test(xnn_f32_gemm_ukernel_6x8s4__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| #endif // !XNN_ARCH_ASMJS && !XNN_ARCH_WASM |
| |
| |
| TEST(F32_GEMM_1X4__SCALAR, k_eq_1) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(4) |
| .k(1) |
| .Test(xnn_f32_gemm_ukernel_1x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_1X4__SCALAR, strided_cn) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(4) |
| .k(1) |
| .cn_stride(7) |
| .Test(xnn_f32_gemm_ukernel_1x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_1X4__SCALAR, k_eq_1_strided_a) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(4) |
| .k(1) |
| .a_stride(3) |
| .Test(xnn_f32_gemm_ukernel_1x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_1X4__SCALAR, k_eq_1_subtile) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 4; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X4__SCALAR, k_eq_1_subtile_m) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(4) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X4__SCALAR, k_eq_1_subtile_n) { |
| for (uint32_t n = 1; n <= 4; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X4__SCALAR, k_gt_1) { |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(4) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X4__SCALAR, k_gt_1_strided_a) { |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(4) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_1x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_1X4__SCALAR, k_gt_1_subtile) { |
| for (size_t k = 2; k < 10; k++) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 4; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X4__SCALAR, n_gt_4) { |
| for (uint32_t n = 5; n < 8; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(4) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X4__SCALAR, n_gt_4_strided_cn) { |
| for (uint32_t n = 5; n < 8; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(4) |
| .k(k) |
| .cn_stride(7) |
| .Test(xnn_f32_gemm_ukernel_1x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X4__SCALAR, n_gt_4_strided_a) { |
| for (uint32_t n = 5; n < 8; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_1x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X4__SCALAR, n_gt_4_subtile) { |
| for (uint32_t n = 5; n < 8; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X4__SCALAR, n_div_4) { |
| for (uint32_t n = 8; n <= 12; n += 4) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(4) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_1x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X4__SCALAR, n_div_4_strided_cn) { |
| for (uint32_t n = 8; n <= 12; n += 4) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .cn_stride(7) |
| .Test(xnn_f32_gemm_ukernel_1x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X4__SCALAR, n_div_4_strided_a) { |
| for (uint32_t n = 8; n <= 12; n += 4) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_1x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X4__SCALAR, n_div_4_subtile) { |
| for (uint32_t n = 8; n <= 12; n += 4) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X4__SCALAR, strided_cm_subtile) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 1; m++) { |
| for (uint32_t n = 1; n <= 4; n++) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(7) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_1x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_1X4__SCALAR, qmin) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(4) |
| .k(1) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_1x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_1X4__SCALAR, qmax) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(4) |
| .k(1) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_1x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_1X4__SCALAR, strided_cm) { |
| GemmMicrokernelTester() |
| .mr(1) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(1) |
| .n(4) |
| .k(1) |
| .cm_stride(7) |
| .Test(xnn_f32_gemm_ukernel_1x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| |
| TEST(F32_GEMM_2X4__SCALAR, k_eq_1) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(4) |
| .k(1) |
| .Test(xnn_f32_gemm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_2X4__SCALAR, strided_cn) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(4) |
| .k(1) |
| .cn_stride(7) |
| .Test(xnn_f32_gemm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_2X4__SCALAR, k_eq_1_strided_a) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(4) |
| .k(1) |
| .a_stride(3) |
| .Test(xnn_f32_gemm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_2X4__SCALAR, k_eq_1_subtile) { |
| for (uint32_t m = 1; m <= 2; m++) { |
| for (uint32_t n = 1; n <= 4; n++) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_2X4__SCALAR, k_eq_1_subtile_m) { |
| for (uint32_t m = 1; m <= 2; m++) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(4) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_2X4__SCALAR, k_eq_1_subtile_n) { |
| for (uint32_t n = 1; n <= 4; n++) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_2X4__SCALAR, k_gt_1) { |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(4) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_2X4__SCALAR, k_gt_1_strided_a) { |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(4) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_2X4__SCALAR, k_gt_1_subtile) { |
| for (size_t k = 2; k < 10; k++) { |
| for (uint32_t m = 1; m <= 2; m++) { |
| for (uint32_t n = 1; n <= 4; n++) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_2X4__SCALAR, n_gt_4) { |
| for (uint32_t n = 5; n < 8; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(4) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_2X4__SCALAR, n_gt_4_strided_cn) { |
| for (uint32_t n = 5; n < 8; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(4) |
| .k(k) |
| .cn_stride(7) |
| .Test(xnn_f32_gemm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_2X4__SCALAR, n_gt_4_strided_a) { |
| for (uint32_t n = 5; n < 8; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_2X4__SCALAR, n_gt_4_subtile) { |
| for (uint32_t n = 5; n < 8; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 2; m++) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_2X4__SCALAR, n_div_4) { |
| for (uint32_t n = 8; n <= 12; n += 4) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(4) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_2X4__SCALAR, n_div_4_strided_cn) { |
| for (uint32_t n = 8; n <= 12; n += 4) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(n) |
| .k(k) |
| .cn_stride(7) |
| .Test(xnn_f32_gemm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_2X4__SCALAR, n_div_4_strided_a) { |
| for (uint32_t n = 8; n <= 12; n += 4) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_2X4__SCALAR, n_div_4_subtile) { |
| for (uint32_t n = 8; n <= 12; n += 4) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 2; m++) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_2X4__SCALAR, strided_cm_subtile) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 2; m++) { |
| for (uint32_t n = 1; n <= 4; n++) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(7) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_2X4__SCALAR, qmin) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(4) |
| .k(1) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_2X4__SCALAR, qmax) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(4) |
| .k(1) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_2X4__SCALAR, strided_cm) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(4) |
| .k(1) |
| .cm_stride(7) |
| .Test(xnn_f32_gemm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| |
| TEST(F32_GEMM_4X4__SCALAR, k_eq_1) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(4) |
| .k(1) |
| .Test(xnn_f32_gemm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X4__SCALAR, strided_cn) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(4) |
| .k(1) |
| .cn_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X4__SCALAR, k_eq_1_strided_a) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(4) |
| .k(1) |
| .a_stride(3) |
| .Test(xnn_f32_gemm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X4__SCALAR, k_eq_1_subtile) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 4; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X4__SCALAR, k_eq_1_subtile_m) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(4) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X4__SCALAR, k_eq_1_subtile_n) { |
| for (uint32_t n = 1; n <= 4; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X4__SCALAR, k_gt_1) { |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(4) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X4__SCALAR, k_gt_1_strided_a) { |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(4) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X4__SCALAR, k_gt_1_subtile) { |
| for (size_t k = 2; k < 10; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 4; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X4__SCALAR, n_gt_4) { |
| for (uint32_t n = 5; n < 8; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(4) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X4__SCALAR, n_gt_4_strided_cn) { |
| for (uint32_t n = 5; n < 8; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(4) |
| .k(k) |
| .cn_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X4__SCALAR, n_gt_4_strided_a) { |
| for (uint32_t n = 5; n < 8; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X4__SCALAR, n_gt_4_subtile) { |
| for (uint32_t n = 5; n < 8; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X4__SCALAR, n_div_4) { |
| for (uint32_t n = 8; n <= 12; n += 4) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(4) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X4__SCALAR, n_div_4_strided_cn) { |
| for (uint32_t n = 8; n <= 12; n += 4) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X4__SCALAR, n_div_4_strided_a) { |
| for (uint32_t n = 8; n <= 12; n += 4) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X4__SCALAR, n_div_4_subtile) { |
| for (uint32_t n = 8; n <= 12; n += 4) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X4__SCALAR, strided_cm_subtile) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 4; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(7) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X4__SCALAR, qmin) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(4) |
| .k(1) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X4__SCALAR, qmax) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(4) |
| .k(1) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X4__SCALAR, strided_cm) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(4) |
| .k(1) |
| .cm_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| |
| TEST(F32_GEMM_4X2__SCALAR, k_eq_1) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(1) |
| .Test(xnn_f32_gemm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X2__SCALAR, strided_cn) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(1) |
| .cn_stride(5) |
| .Test(xnn_f32_gemm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X2__SCALAR, k_eq_1_strided_a) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(1) |
| .a_stride(3) |
| .Test(xnn_f32_gemm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X2__SCALAR, k_eq_1_subtile) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 2; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__SCALAR, k_eq_1_subtile_m) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(2) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__SCALAR, k_eq_1_subtile_n) { |
| for (uint32_t n = 1; n <= 2; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__SCALAR, k_gt_1) { |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__SCALAR, k_gt_1_strided_a) { |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(k) |
| .a_stride(11) |
| .Test(xnn_f32_gemm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__SCALAR, k_gt_1_subtile) { |
| for (size_t k = 2; k < 10; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 2; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__SCALAR, n_gt_2) { |
| for (uint32_t n = 3; n < 4; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__SCALAR, n_gt_2_strided_cn) { |
| for (uint32_t n = 3; n < 4; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(k) |
| .cn_stride(5) |
| .Test(xnn_f32_gemm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__SCALAR, n_gt_2_strided_a) { |
| for (uint32_t n = 3; n < 4; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__SCALAR, n_gt_2_subtile) { |
| for (uint32_t n = 3; n < 4; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__SCALAR, n_div_2) { |
| for (uint32_t n = 4; n <= 6; n += 2) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(k) |
| .Test(xnn_f32_gemm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__SCALAR, n_div_2_strided_cn) { |
| for (uint32_t n = 4; n <= 6; n += 2) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(5) |
| .Test(xnn_f32_gemm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__SCALAR, n_div_2_strided_a) { |
| for (uint32_t n = 4; n <= 6; n += 2) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_gemm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__SCALAR, n_div_2_subtile) { |
| for (uint32_t n = 4; n <= 6; n += 2) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__SCALAR, strided_cm_subtile) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 2; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(5) |
| .iterations(1) |
| .Test(xnn_f32_gemm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_GEMM_4X2__SCALAR, qmin) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(1) |
| .qmin(128) |
| .Test(xnn_f32_gemm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X2__SCALAR, qmax) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(1) |
| .qmax(128) |
| .Test(xnn_f32_gemm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_GEMM_4X2__SCALAR, strided_cm) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(1) |
| .cm_stride(5) |
| .Test(xnn_f32_gemm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |