| // 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/qu8-vmulc-minmax-fp32.yaml |
| // Generator: tools/generate-vbinary-test.py |
| |
| |
| #include <gtest/gtest.h> |
| |
| #include <xnnpack/common.h> |
| #include <xnnpack/isa-checks.h> |
| |
| #include <xnnpack/params-init.h> |
| #include <xnnpack/vmul.h> |
| #include "vmulc-microkernel-tester.h" |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X8, batch_eq_8) { |
| TEST_REQUIRES_ARM_NEON; |
| VMulCMicrokernelTester() |
| .batch_size(8) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X8, batch_div_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 16; batch_size < 80; batch_size += 8) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X8, batch_lt_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size < 8; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X8, batch_gt_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 9; batch_size < 16; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X8, inplace) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .inplace(true) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X8, a_zero_point) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (int32_t a_zero_point = -128; a_zero_point <= 127; a_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_zero_point(a_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X8, b_zero_point) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (int32_t b_zero_point = -128; b_zero_point <= 127; b_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_zero_point(b_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X8, y_zero_point) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (int32_t y_zero_point = -128; y_zero_point <= 127; y_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_zero_point(y_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X8, a_scale) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (float a_scale = 0.1f; a_scale <= 10.0f; a_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_scale(a_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X8, b_scale) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (float b_scale = 0.1f; b_scale <= 10.0f; b_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_scale(b_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X8, y_scale) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (float y_scale = 0.1f; y_scale <= 10.0f; y_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_scale(y_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X8, qmin) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmin(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X8, qmax) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmax(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X16, batch_eq_16) { |
| TEST_REQUIRES_ARM_NEON; |
| VMulCMicrokernelTester() |
| .batch_size(16) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X16, batch_div_16) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 32; batch_size < 160; batch_size += 16) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X16, batch_lt_16) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size < 16; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X16, batch_gt_16) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 17; batch_size < 32; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X16, inplace) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .inplace(true) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X16, a_zero_point) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t a_zero_point = -128; a_zero_point <= 127; a_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_zero_point(a_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X16, b_zero_point) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t b_zero_point = -128; b_zero_point <= 127; b_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_zero_point(b_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X16, y_zero_point) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t y_zero_point = -128; y_zero_point <= 127; y_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_zero_point(y_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X16, a_scale) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float a_scale = 0.1f; a_scale <= 10.0f; a_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_scale(a_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X16, b_scale) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float b_scale = 0.1f; b_scale <= 10.0f; b_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_scale(b_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X16, y_scale) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float y_scale = 0.1f; y_scale <= 10.0f; y_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_scale(y_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X16, qmin) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmin(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD64_X16, qmax) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmax(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD128_X16, batch_eq_16) { |
| TEST_REQUIRES_ARM_NEON; |
| VMulCMicrokernelTester() |
| .batch_size(16) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD128_X16, batch_div_16) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 32; batch_size < 160; batch_size += 16) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD128_X16, batch_lt_16) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size < 16; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD128_X16, batch_gt_16) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 17; batch_size < 32; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD128_X16, inplace) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .inplace(true) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD128_X16, a_zero_point) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t a_zero_point = -128; a_zero_point <= 127; a_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_zero_point(a_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD128_X16, b_zero_point) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t b_zero_point = -128; b_zero_point <= 127; b_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_zero_point(b_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD128_X16, y_zero_point) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t y_zero_point = -128; y_zero_point <= 127; y_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_zero_point(y_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD128_X16, a_scale) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float a_scale = 0.1f; a_scale <= 10.0f; a_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_scale(a_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD128_X16, b_scale) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float b_scale = 0.1f; b_scale <= 10.0f; b_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_scale(b_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD128_X16, y_scale) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float y_scale = 0.1f; y_scale <= 10.0f; y_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_scale(y_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD128_X16, qmin) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmin(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEON_LD128_X16, qmax) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmax(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neon_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neon_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X8, batch_eq_8) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| VMulCMicrokernelTester() |
| .batch_size(8) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X8, batch_div_8) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 16; batch_size < 80; batch_size += 8) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X8, batch_lt_8) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size < 8; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X8, batch_gt_8) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 9; batch_size < 16; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X8, inplace) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .inplace(true) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X8, a_zero_point) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (int32_t a_zero_point = -128; a_zero_point <= 127; a_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_zero_point(a_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X8, b_zero_point) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (int32_t b_zero_point = -128; b_zero_point <= 127; b_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_zero_point(b_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X8, y_zero_point) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (int32_t y_zero_point = -128; y_zero_point <= 127; y_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_zero_point(y_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X8, a_scale) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (float a_scale = 0.1f; a_scale <= 10.0f; a_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_scale(a_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X8, b_scale) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (float b_scale = 0.1f; b_scale <= 10.0f; b_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_scale(b_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X8, y_scale) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (float y_scale = 0.1f; y_scale <= 10.0f; y_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_scale(y_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X8, qmin) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmin(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X8, qmax) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmax(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x8, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X16, batch_eq_16) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| VMulCMicrokernelTester() |
| .batch_size(16) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X16, batch_div_16) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 32; batch_size < 160; batch_size += 16) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X16, batch_lt_16) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size < 16; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X16, batch_gt_16) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 17; batch_size < 32; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X16, inplace) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .inplace(true) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X16, a_zero_point) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t a_zero_point = -128; a_zero_point <= 127; a_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_zero_point(a_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X16, b_zero_point) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t b_zero_point = -128; b_zero_point <= 127; b_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_zero_point(b_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X16, y_zero_point) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t y_zero_point = -128; y_zero_point <= 127; y_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_zero_point(y_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X16, a_scale) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float a_scale = 0.1f; a_scale <= 10.0f; a_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_scale(a_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X16, b_scale) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float b_scale = 0.1f; b_scale <= 10.0f; b_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_scale(b_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X16, y_scale) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float y_scale = 0.1f; y_scale <= 10.0f; y_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_scale(y_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X16, qmin) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmin(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD64_X16, qmax) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmax(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld64_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD128_X16, batch_eq_16) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| VMulCMicrokernelTester() |
| .batch_size(16) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD128_X16, batch_div_16) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 32; batch_size < 160; batch_size += 16) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD128_X16, batch_lt_16) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size < 16; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD128_X16, batch_gt_16) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 17; batch_size < 32; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD128_X16, inplace) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .inplace(true) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD128_X16, a_zero_point) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t a_zero_point = -128; a_zero_point <= 127; a_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_zero_point(a_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD128_X16, b_zero_point) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t b_zero_point = -128; b_zero_point <= 127; b_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_zero_point(b_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD128_X16, y_zero_point) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t y_zero_point = -128; y_zero_point <= 127; y_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_zero_point(y_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD128_X16, a_scale) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float a_scale = 0.1f; a_scale <= 10.0f; a_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_scale(a_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD128_X16, b_scale) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float b_scale = 0.1f; b_scale <= 10.0f; b_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_scale(b_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD128_X16, y_scale) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float y_scale = 0.1f; y_scale <= 10.0f; y_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_scale(y_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD128_X16, qmin) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmin(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__NEONV8_LD128_X16, qmax) { |
| TEST_REQUIRES_ARM_NEON_V8; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmax(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__neonv8_ld128_x16, xnn_init_qu8_mul_minmax_fp32_neonv8_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X8, batch_eq_8) { |
| TEST_REQUIRES_X86_SSE2; |
| VMulCMicrokernelTester() |
| .batch_size(8) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X8, batch_div_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 16; batch_size < 80; batch_size += 8) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X8, batch_lt_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 1; batch_size < 8; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X8, batch_gt_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 9; batch_size < 16; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X8, inplace) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .inplace(true) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X8, a_zero_point) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (int32_t a_zero_point = -128; a_zero_point <= 127; a_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_zero_point(a_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X8, b_zero_point) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (int32_t b_zero_point = -128; b_zero_point <= 127; b_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_zero_point(b_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X8, y_zero_point) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (int32_t y_zero_point = -128; y_zero_point <= 127; y_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_zero_point(y_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X8, a_scale) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (float a_scale = 0.1f; a_scale <= 10.0f; a_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_scale(a_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X8, b_scale) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (float b_scale = 0.1f; b_scale <= 10.0f; b_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_scale(b_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X8, y_scale) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (float y_scale = 0.1f; y_scale <= 10.0f; y_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_scale(y_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X8, qmin) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmin(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X8, qmax) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmax(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| #endif // XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| |
| |
| #if XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X16, batch_eq_16) { |
| TEST_REQUIRES_X86_SSE2; |
| VMulCMicrokernelTester() |
| .batch_size(16) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X16, batch_div_16) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 32; batch_size < 160; batch_size += 16) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X16, batch_lt_16) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 1; batch_size < 16; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X16, batch_gt_16) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 17; batch_size < 32; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X16, inplace) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .inplace(true) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X16, a_zero_point) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t a_zero_point = -128; a_zero_point <= 127; a_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_zero_point(a_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X16, b_zero_point) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t b_zero_point = -128; b_zero_point <= 127; b_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_zero_point(b_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X16, y_zero_point) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t y_zero_point = -128; y_zero_point <= 127; y_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_zero_point(y_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X16, a_scale) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float a_scale = 0.1f; a_scale <= 10.0f; a_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_scale(a_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X16, b_scale) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float b_scale = 0.1f; b_scale <= 10.0f; b_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_scale(b_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X16, y_scale) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float y_scale = 0.1f; y_scale <= 10.0f; y_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_scale(y_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X16, qmin) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmin(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE2_MUL16_LD64_X16, qmax) { |
| TEST_REQUIRES_X86_SSE2; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmax(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse2_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| #endif // XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| |
| |
| #if XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X8, batch_eq_8) { |
| TEST_REQUIRES_X86_SSE41; |
| VMulCMicrokernelTester() |
| .batch_size(8) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X8, batch_div_8) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 16; batch_size < 80; batch_size += 8) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X8, batch_lt_8) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 1; batch_size < 8; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X8, batch_gt_8) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 9; batch_size < 16; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X8, inplace) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .inplace(true) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X8, a_zero_point) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (int32_t a_zero_point = -128; a_zero_point <= 127; a_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_zero_point(a_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X8, b_zero_point) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (int32_t b_zero_point = -128; b_zero_point <= 127; b_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_zero_point(b_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X8, y_zero_point) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (int32_t y_zero_point = -128; y_zero_point <= 127; y_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_zero_point(y_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X8, a_scale) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (float a_scale = 0.1f; a_scale <= 10.0f; a_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_scale(a_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X8, b_scale) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (float b_scale = 0.1f; b_scale <= 10.0f; b_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_scale(b_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X8, y_scale) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (float y_scale = 0.1f; y_scale <= 10.0f; y_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_scale(y_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X8, qmin) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmin(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X8, qmax) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmax(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| #endif // XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| |
| |
| #if XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X16, batch_eq_16) { |
| TEST_REQUIRES_X86_SSE41; |
| VMulCMicrokernelTester() |
| .batch_size(16) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X16, batch_div_16) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 32; batch_size < 160; batch_size += 16) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X16, batch_lt_16) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 1; batch_size < 16; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X16, batch_gt_16) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 17; batch_size < 32; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X16, inplace) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .inplace(true) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X16, a_zero_point) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t a_zero_point = -128; a_zero_point <= 127; a_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_zero_point(a_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X16, b_zero_point) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t b_zero_point = -128; b_zero_point <= 127; b_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_zero_point(b_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X16, y_zero_point) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t y_zero_point = -128; y_zero_point <= 127; y_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_zero_point(y_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X16, a_scale) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float a_scale = 0.1f; a_scale <= 10.0f; a_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_scale(a_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X16, b_scale) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float b_scale = 0.1f; b_scale <= 10.0f; b_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_scale(b_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X16, y_scale) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float y_scale = 0.1f; y_scale <= 10.0f; y_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_scale(y_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X16, qmin) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmin(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SSE41_MUL16_LD64_X16, qmax) { |
| TEST_REQUIRES_X86_SSE41; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmax(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__sse41_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| #endif // XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| |
| |
| #if XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X8, batch_eq_8) { |
| TEST_REQUIRES_X86_AVX; |
| VMulCMicrokernelTester() |
| .batch_size(8) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X8, batch_div_8) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 16; batch_size < 80; batch_size += 8) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X8, batch_lt_8) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 1; batch_size < 8; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X8, batch_gt_8) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 9; batch_size < 16; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X8, inplace) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .inplace(true) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X8, a_zero_point) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (int32_t a_zero_point = -128; a_zero_point <= 127; a_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_zero_point(a_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X8, b_zero_point) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (int32_t b_zero_point = -128; b_zero_point <= 127; b_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_zero_point(b_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X8, y_zero_point) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (int32_t y_zero_point = -128; y_zero_point <= 127; y_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_zero_point(y_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X8, a_scale) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (float a_scale = 0.1f; a_scale <= 10.0f; a_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_scale(a_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X8, b_scale) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (float b_scale = 0.1f; b_scale <= 10.0f; b_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_scale(b_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X8, y_scale) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (float y_scale = 0.1f; y_scale <= 10.0f; y_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_scale(y_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X8, qmin) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmin(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X8, qmax) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmax(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x8, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| #endif // XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| |
| |
| #if XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X16, batch_eq_16) { |
| TEST_REQUIRES_X86_AVX; |
| VMulCMicrokernelTester() |
| .batch_size(16) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X16, batch_div_16) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 32; batch_size < 160; batch_size += 16) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X16, batch_lt_16) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 1; batch_size < 16; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X16, batch_gt_16) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 17; batch_size < 32; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X16, inplace) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .inplace(true) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X16, a_zero_point) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t a_zero_point = -128; a_zero_point <= 127; a_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_zero_point(a_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X16, b_zero_point) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t b_zero_point = -128; b_zero_point <= 127; b_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_zero_point(b_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X16, y_zero_point) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t y_zero_point = -128; y_zero_point <= 127; y_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_zero_point(y_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X16, a_scale) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float a_scale = 0.1f; a_scale <= 10.0f; a_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_scale(a_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X16, b_scale) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float b_scale = 0.1f; b_scale <= 10.0f; b_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_scale(b_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X16, y_scale) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float y_scale = 0.1f; y_scale <= 10.0f; y_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_scale(y_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X16, qmin) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmin(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__AVX_MUL16_LD64_X16, qmax) { |
| TEST_REQUIRES_X86_AVX; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmax(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__avx_mul16_ld64_x16, xnn_init_qu8_mul_minmax_fp32_sse2_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| #endif // XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| |
| |
| #if XNN_ARCH_WASMSIMD |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X8, batch_eq_8) { |
| VMulCMicrokernelTester() |
| .batch_size(8) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x8, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X8, batch_div_8) { |
| for (size_t batch_size = 16; batch_size < 80; batch_size += 8) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x8, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X8, batch_lt_8) { |
| for (size_t batch_size = 1; batch_size < 8; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x8, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X8, batch_gt_8) { |
| for (size_t batch_size = 9; batch_size < 16; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x8, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X8, inplace) { |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .inplace(true) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x8, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X8, a_zero_point) { |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (int32_t a_zero_point = -128; a_zero_point <= 127; a_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_zero_point(a_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x8, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X8, b_zero_point) { |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (int32_t b_zero_point = -128; b_zero_point <= 127; b_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_zero_point(b_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x8, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X8, y_zero_point) { |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (int32_t y_zero_point = -128; y_zero_point <= 127; y_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_zero_point(y_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x8, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X8, a_scale) { |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (float a_scale = 0.1f; a_scale <= 10.0f; a_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_scale(a_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x8, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X8, b_scale) { |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (float b_scale = 0.1f; b_scale <= 10.0f; b_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_scale(b_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x8, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X8, y_scale) { |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| for (float y_scale = 0.1f; y_scale <= 10.0f; y_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_scale(y_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x8, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X8, qmin) { |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmin(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x8, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X8, qmax) { |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmax(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x8, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| #endif // XNN_ARCH_WASMSIMD |
| |
| |
| #if XNN_ARCH_WASMSIMD |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X16, batch_eq_16) { |
| VMulCMicrokernelTester() |
| .batch_size(16) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x16, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X16, batch_div_16) { |
| for (size_t batch_size = 32; batch_size < 160; batch_size += 16) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x16, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X16, batch_lt_16) { |
| for (size_t batch_size = 1; batch_size < 16; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x16, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X16, batch_gt_16) { |
| for (size_t batch_size = 17; batch_size < 32; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x16, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X16, inplace) { |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .inplace(true) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x16, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X16, a_zero_point) { |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t a_zero_point = -128; a_zero_point <= 127; a_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_zero_point(a_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x16, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X16, b_zero_point) { |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t b_zero_point = -128; b_zero_point <= 127; b_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_zero_point(b_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x16, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X16, y_zero_point) { |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (int32_t y_zero_point = -128; y_zero_point <= 127; y_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_zero_point(y_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x16, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X16, a_scale) { |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float a_scale = 0.1f; a_scale <= 10.0f; a_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_scale(a_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x16, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X16, b_scale) { |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float b_scale = 0.1f; b_scale <= 10.0f; b_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_scale(b_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x16, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X16, y_scale) { |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| for (float y_scale = 0.1f; y_scale <= 10.0f; y_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_scale(y_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x16, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X16, qmin) { |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmin(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x16, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__WASMSIMD_MUL32_LD64_X16, qmax) { |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmax(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__wasmsimd_mul32_ld64_x16, xnn_init_qu8_mul_minmax_fp32_wasmsimd_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| #endif // XNN_ARCH_WASMSIMD |
| |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X1, batch_eq_1) { |
| VMulCMicrokernelTester() |
| .batch_size(1) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x1, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X1, batch_gt_1) { |
| for (size_t batch_size = 2; batch_size < 10; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x1, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X1, inplace) { |
| for (size_t batch_size = 1; batch_size <= 5; batch_size += 1) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .inplace(true) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x1, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X1, a_zero_point) { |
| for (size_t batch_size = 1; batch_size <= 5; batch_size += 1) { |
| for (int32_t a_zero_point = -128; a_zero_point <= 127; a_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_zero_point(a_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x1, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X1, b_zero_point) { |
| for (size_t batch_size = 1; batch_size <= 5; batch_size += 1) { |
| for (int32_t b_zero_point = -128; b_zero_point <= 127; b_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_zero_point(b_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x1, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X1, y_zero_point) { |
| for (size_t batch_size = 1; batch_size <= 5; batch_size += 1) { |
| for (int32_t y_zero_point = -128; y_zero_point <= 127; y_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_zero_point(y_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x1, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X1, a_scale) { |
| for (size_t batch_size = 1; batch_size <= 5; batch_size += 1) { |
| for (float a_scale = 0.1f; a_scale <= 10.0f; a_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_scale(a_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x1, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X1, b_scale) { |
| for (size_t batch_size = 1; batch_size <= 5; batch_size += 1) { |
| for (float b_scale = 0.1f; b_scale <= 10.0f; b_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_scale(b_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x1, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X1, y_scale) { |
| for (size_t batch_size = 1; batch_size <= 5; batch_size += 1) { |
| for (float y_scale = 0.1f; y_scale <= 10.0f; y_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_scale(y_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x1, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X1, qmin) { |
| for (size_t batch_size = 1; batch_size <= 5; batch_size += 1) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmin(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x1, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X1, qmax) { |
| for (size_t batch_size = 1; batch_size <= 5; batch_size += 1) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmax(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x1, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X2, batch_eq_2) { |
| VMulCMicrokernelTester() |
| .batch_size(2) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x2, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X2, batch_div_2) { |
| for (size_t batch_size = 4; batch_size < 20; batch_size += 2) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x2, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X2, batch_lt_2) { |
| for (size_t batch_size = 1; batch_size < 2; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x2, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X2, batch_gt_2) { |
| for (size_t batch_size = 3; batch_size < 4; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x2, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X2, inplace) { |
| for (size_t batch_size = 1; batch_size <= 10; batch_size += 1) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .inplace(true) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x2, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X2, a_zero_point) { |
| for (size_t batch_size = 1; batch_size <= 10; batch_size += 1) { |
| for (int32_t a_zero_point = -128; a_zero_point <= 127; a_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_zero_point(a_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x2, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X2, b_zero_point) { |
| for (size_t batch_size = 1; batch_size <= 10; batch_size += 1) { |
| for (int32_t b_zero_point = -128; b_zero_point <= 127; b_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_zero_point(b_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x2, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X2, y_zero_point) { |
| for (size_t batch_size = 1; batch_size <= 10; batch_size += 1) { |
| for (int32_t y_zero_point = -128; y_zero_point <= 127; y_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_zero_point(y_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x2, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X2, a_scale) { |
| for (size_t batch_size = 1; batch_size <= 10; batch_size += 1) { |
| for (float a_scale = 0.1f; a_scale <= 10.0f; a_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_scale(a_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x2, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X2, b_scale) { |
| for (size_t batch_size = 1; batch_size <= 10; batch_size += 1) { |
| for (float b_scale = 0.1f; b_scale <= 10.0f; b_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_scale(b_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x2, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X2, y_scale) { |
| for (size_t batch_size = 1; batch_size <= 10; batch_size += 1) { |
| for (float y_scale = 0.1f; y_scale <= 10.0f; y_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_scale(y_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x2, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X2, qmin) { |
| for (size_t batch_size = 1; batch_size <= 10; batch_size += 1) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmin(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x2, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X2, qmax) { |
| for (size_t batch_size = 1; batch_size <= 10; batch_size += 1) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmax(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x2, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X4, batch_eq_4) { |
| VMulCMicrokernelTester() |
| .batch_size(4) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x4, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X4, batch_div_4) { |
| for (size_t batch_size = 8; batch_size < 40; batch_size += 4) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x4, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X4, batch_lt_4) { |
| for (size_t batch_size = 1; batch_size < 4; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x4, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X4, batch_gt_4) { |
| for (size_t batch_size = 5; batch_size < 8; batch_size++) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x4, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X4, inplace) { |
| for (size_t batch_size = 1; batch_size <= 20; batch_size += 3) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .inplace(true) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x4, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X4, a_zero_point) { |
| for (size_t batch_size = 1; batch_size <= 20; batch_size += 3) { |
| for (int32_t a_zero_point = -128; a_zero_point <= 127; a_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_zero_point(a_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x4, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X4, b_zero_point) { |
| for (size_t batch_size = 1; batch_size <= 20; batch_size += 3) { |
| for (int32_t b_zero_point = -128; b_zero_point <= 127; b_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_zero_point(b_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x4, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X4, y_zero_point) { |
| for (size_t batch_size = 1; batch_size <= 20; batch_size += 3) { |
| for (int32_t y_zero_point = -128; y_zero_point <= 127; y_zero_point += 51) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_zero_point(y_zero_point) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x4, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X4, a_scale) { |
| for (size_t batch_size = 1; batch_size <= 20; batch_size += 3) { |
| for (float a_scale = 0.1f; a_scale <= 10.0f; a_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .a_scale(a_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x4, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X4, b_scale) { |
| for (size_t batch_size = 1; batch_size <= 20; batch_size += 3) { |
| for (float b_scale = 0.1f; b_scale <= 10.0f; b_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .b_scale(b_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x4, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X4, y_scale) { |
| for (size_t batch_size = 1; batch_size <= 20; batch_size += 3) { |
| for (float y_scale = 0.1f; y_scale <= 10.0f; y_scale *= 3.14f) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .y_scale(y_scale) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x4, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X4, qmin) { |
| for (size_t batch_size = 1; batch_size <= 20; batch_size += 3) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmin(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x4, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |
| |
| TEST(QU8_VMULC_MINMAX_FP32__SCALAR_X4, qmax) { |
| for (size_t batch_size = 1; batch_size <= 20; batch_size += 3) { |
| VMulCMicrokernelTester() |
| .batch_size(batch_size) |
| .qmax(128) |
| .Test(xnn_qu8_vmulc_minmax_fp32_ukernel__scalar_x4, xnn_init_qu8_mul_minmax_fp32_scalar_params, xnn_init_qu8_requantization_fp32_params, xnn_qu8_requantize_fp32); |
| } |
| } |