| // Copyright 2021 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/x32-transpose.yaml |
| // Generator: tools/generate-transpose-test.py |
| |
| |
| #include <gtest/gtest.h> |
| |
| #include <xnnpack/common.h> |
| #include <xnnpack/isa-checks.h> |
| |
| #include <xnnpack/transpose.h> |
| #include "transpose-microkernel-tester.h" |
| |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_INT, bh_1_bw_2) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(1) |
| .block_width(2) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_INT, bh_1_2_bw_1_4) { |
| for(size_t i = 1; i <= 2; ++i){ |
| for(size_t j = 1; j <= 4; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_int); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_INT, bh_1_bw_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(1) |
| .block_width(4) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_INT, bh_1_bw_3_4) { |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(1) |
| .block_width(i) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_INT, bh_2_bw_3_4) { |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(2) |
| .block_width(i) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_INT, bh_2_bw_2) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(2) |
| .block_width(2) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_INT, bh_2_2_bw_2){ |
| for(size_t i = 2; i < 2; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(i) |
| .block_width(2) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_INT, bh_2_2_bw_4){ |
| for(size_t i = 2; i < 2; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_INT, bh_2_2_bw_3_4) { |
| for(size_t i = 2; i < 2; ++i){ |
| for(size_t j = 3; j < 4; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_int); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_INT, bh_1_bw_2_is_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(1) |
| .block_width(2) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_INT, bh_1_bw_2_os_2) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(2) |
| .block_width(2) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_INT, bh_1_bw_2_is_4_os_2) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(2) |
| .block_width(2) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_INT, bh_1_bw_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(1) |
| .block_width(4) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_INT, bh_1_2_bw_1_8) { |
| for(size_t i = 1; i <= 2; ++i){ |
| for(size_t j = 1; j <= 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_int); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_INT, bh_1_bw_8) { |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(1) |
| .block_width(8) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_INT, bh_1_bw_5_8) { |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(1) |
| .block_width(i) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_INT, bh_2_bw_5_8) { |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(2) |
| .block_width(i) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_INT, bh_2_bw_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(2) |
| .block_width(4) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_INT, bh_2_2_bw_4){ |
| for(size_t i = 2; i < 2; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_INT, bh_2_2_bw_8){ |
| for(size_t i = 2; i < 2; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(i) |
| .block_width(8) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_INT, bh_2_2_bw_5_8) { |
| for(size_t i = 2; i < 2; ++i){ |
| for(size_t j = 5; j < 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_int); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_INT, bh_1_bw_4_is_8) { |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(1) |
| .block_width(4) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_INT, bh_1_bw_4_os_2) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(2) |
| .block_width(4) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_INT, bh_1_bw_4_is_8_os_2) { |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(2) |
| .block_width(4) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_INT, bh_2_bw_1) { |
| TransposeMicrokernelTester() |
| .input_stride(1) |
| .output_stride(2) |
| .block_width(1) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_INT, bh_1_4_bw_1_2) { |
| for(size_t i = 1; i <= 4; ++i){ |
| for(size_t j = 1; j <= 2; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_int); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_INT, bh_2_bw_2) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(2) |
| .block_width(2) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_INT, bh_2_bw_2_2) { |
| for(size_t i = 2; i < 2; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(2) |
| .block_width(i) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_INT, bh_4_bw_2_2) { |
| for(size_t i = 2; i < 2; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_INT, bh_4_bw_1) { |
| TransposeMicrokernelTester() |
| .input_stride(1) |
| .output_stride(4) |
| .block_width(1) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_INT, bh_3_4_bw_1){ |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(1) |
| .output_stride(i) |
| .block_width(1) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_INT, bh_3_4_bw_2){ |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(i) |
| .block_width(2) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_INT, bh_3_4_bw_2_2) { |
| for(size_t i = 3; i < 4; ++i){ |
| for(size_t j = 2; j < 2; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_int); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_INT, bh_2_bw_1_is_2) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(2) |
| .block_width(1) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_INT, bh_2_bw_1_os_4) { |
| TransposeMicrokernelTester() |
| .input_stride(1) |
| .output_stride(4) |
| .block_width(1) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_INT, bh_2_bw_1_is_2_os_4) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(4) |
| .block_width(1) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_INT, bh_2_bw_2) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(2) |
| .block_width(2) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_INT, bh_1_4_bw_1_4) { |
| for(size_t i = 1; i <= 4; ++i){ |
| for(size_t j = 1; j <= 4; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_int); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_INT, bh_2_bw_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(2) |
| .block_width(4) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_INT, bh_2_bw_3_4) { |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(2) |
| .block_width(i) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_INT, bh_4_bw_3_4) { |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_INT, bh_4_bw_2) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(4) |
| .block_width(2) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_INT, bh_3_4_bw_2){ |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(i) |
| .block_width(2) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_INT, bh_3_4_bw_4){ |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_INT, bh_3_4_bw_3_4) { |
| for(size_t i = 3; i < 4; ++i){ |
| for(size_t j = 3; j < 4; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_int); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_INT, bh_2_bw_2_is_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(2) |
| .block_width(2) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_INT, bh_2_bw_2_os_4) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(4) |
| .block_width(2) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_INT, bh_2_bw_2_is_4_os_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(2) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_INT, bh_2_bw_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(2) |
| .block_width(4) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_INT, bh_1_4_bw_1_8) { |
| for(size_t i = 1; i <= 4; ++i){ |
| for(size_t j = 1; j <= 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_int); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_INT, bh_2_bw_8) { |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(2) |
| .block_width(8) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_INT, bh_2_bw_5_8) { |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(2) |
| .block_width(i) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_INT, bh_4_bw_5_8) { |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_INT, bh_4_bw_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_INT, bh_3_4_bw_4){ |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_INT, bh_3_4_bw_8){ |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(i) |
| .block_width(8) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_INT, bh_3_4_bw_5_8) { |
| for(size_t i = 3; i < 4; ++i){ |
| for(size_t j = 5; j < 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_int); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_INT, bh_2_bw_4_is_8) { |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(2) |
| .block_width(4) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_INT, bh_2_bw_4_os_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_INT, bh_2_bw_4_is_8_os_4) { |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_INT, bh_4_bw_1) { |
| TransposeMicrokernelTester() |
| .input_stride(1) |
| .output_stride(4) |
| .block_width(1) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_INT, bh_1_8_bw_1_2) { |
| for(size_t i = 1; i <= 8; ++i){ |
| for(size_t j = 1; j <= 2; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_int); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_INT, bh_4_bw_2) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(4) |
| .block_width(2) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_INT, bh_4_bw_2_2) { |
| for(size_t i = 2; i < 2; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_INT, bh_8_bw_2_2) { |
| for(size_t i = 2; i < 2; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(8) |
| .block_width(i) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_INT, bh_8_bw_1) { |
| TransposeMicrokernelTester() |
| .input_stride(1) |
| .output_stride(8) |
| .block_width(1) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_INT, bh_5_8_bw_1){ |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(1) |
| .output_stride(i) |
| .block_width(1) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_INT, bh_5_8_bw_2){ |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(i) |
| .block_width(2) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_INT, bh_5_8_bw_2_2) { |
| for(size_t i = 5; i < 8; ++i){ |
| for(size_t j = 2; j < 2; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_int); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_INT, bh_4_bw_1_is_2) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(4) |
| .block_width(1) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_INT, bh_4_bw_1_os_8) { |
| TransposeMicrokernelTester() |
| .input_stride(1) |
| .output_stride(8) |
| .block_width(1) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_INT, bh_4_bw_1_is_2_os_8) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(8) |
| .block_width(1) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_INT, bh_4_bw_2) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(4) |
| .block_width(2) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_INT, bh_1_8_bw_1_4) { |
| for(size_t i = 1; i <= 8; ++i){ |
| for(size_t j = 1; j <= 4; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_int); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_INT, bh_4_bw_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_INT, bh_4_bw_3_4) { |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_INT, bh_8_bw_3_4) { |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(8) |
| .block_width(i) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_INT, bh_8_bw_2) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(8) |
| .block_width(2) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_INT, bh_5_8_bw_2){ |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(i) |
| .block_width(2) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_INT, bh_5_8_bw_4){ |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_INT, bh_5_8_bw_3_4) { |
| for(size_t i = 5; i < 8; ++i){ |
| for(size_t j = 3; j < 4; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_int); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_INT, bh_4_bw_2_is_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(2) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_INT, bh_4_bw_2_os_8) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(8) |
| .block_width(2) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_INT, bh_4_bw_2_is_4_os_8) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(2) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_INT, bh_4_bw_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_INT, bh_1_8_bw_1_8) { |
| for(size_t i = 1; i <= 8; ++i){ |
| for(size_t j = 1; j <= 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_int); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_INT, bh_4_bw_8) { |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(8) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_INT, bh_4_bw_5_8) { |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_INT, bh_8_bw_5_8) { |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(8) |
| .block_width(i) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_INT, bh_8_bw_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_INT, bh_5_8_bw_4){ |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_INT, bh_5_8_bw_8){ |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(i) |
| .block_width(8) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_int); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_INT, bh_5_8_bw_5_8) { |
| for(size_t i = 5; i < 8; ++i){ |
| for(size_t j = 5; j < 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_int); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_INT, bh_4_bw_4_is_8) { |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_INT, bh_4_bw_4_os_8) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_INT, bh_4_bw_4_is_8_os_8) { |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_int); |
| } |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_FLOAT, bh_1_bw_2) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(1) |
| .block_width(2) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_FLOAT, bh_1_2_bw_1_4) { |
| for(size_t i = 1; i <= 2; ++i){ |
| for(size_t j = 1; j <= 4; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_float); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_FLOAT, bh_1_bw_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(1) |
| .block_width(4) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_FLOAT, bh_1_bw_3_4) { |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(1) |
| .block_width(i) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_FLOAT, bh_2_bw_3_4) { |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(2) |
| .block_width(i) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_FLOAT, bh_2_bw_2) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(2) |
| .block_width(2) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_FLOAT, bh_2_2_bw_2){ |
| for(size_t i = 2; i < 2; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(i) |
| .block_width(2) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_FLOAT, bh_2_2_bw_4){ |
| for(size_t i = 2; i < 2; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_FLOAT, bh_2_2_bw_3_4) { |
| for(size_t i = 2; i < 2; ++i){ |
| for(size_t j = 3; j < 4; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_float); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_FLOAT, bh_1_bw_2_is_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(1) |
| .block_width(2) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_FLOAT, bh_1_bw_2_os_2) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(2) |
| .block_width(2) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__1X2_SCALAR_FLOAT, bh_1_bw_2_is_4_os_2) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(2) |
| .block_width(2) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x2_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_FLOAT, bh_1_bw_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(1) |
| .block_width(4) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_FLOAT, bh_1_2_bw_1_8) { |
| for(size_t i = 1; i <= 2; ++i){ |
| for(size_t j = 1; j <= 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_float); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_FLOAT, bh_1_bw_8) { |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(1) |
| .block_width(8) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_FLOAT, bh_1_bw_5_8) { |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(1) |
| .block_width(i) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_FLOAT, bh_2_bw_5_8) { |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(2) |
| .block_width(i) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_FLOAT, bh_2_bw_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(2) |
| .block_width(4) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_FLOAT, bh_2_2_bw_4){ |
| for(size_t i = 2; i < 2; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_FLOAT, bh_2_2_bw_8){ |
| for(size_t i = 2; i < 2; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(i) |
| .block_width(8) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_FLOAT, bh_2_2_bw_5_8) { |
| for(size_t i = 2; i < 2; ++i){ |
| for(size_t j = 5; j < 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_float); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_FLOAT, bh_1_bw_4_is_8) { |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(1) |
| .block_width(4) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_FLOAT, bh_1_bw_4_os_2) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(2) |
| .block_width(4) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__1X4_SCALAR_FLOAT, bh_1_bw_4_is_8_os_2) { |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(2) |
| .block_width(4) |
| .block_height(1) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__1x4_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_FLOAT, bh_2_bw_1) { |
| TransposeMicrokernelTester() |
| .input_stride(1) |
| .output_stride(2) |
| .block_width(1) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_FLOAT, bh_1_4_bw_1_2) { |
| for(size_t i = 1; i <= 4; ++i){ |
| for(size_t j = 1; j <= 2; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_float); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_FLOAT, bh_2_bw_2) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(2) |
| .block_width(2) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_FLOAT, bh_2_bw_2_2) { |
| for(size_t i = 2; i < 2; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(2) |
| .block_width(i) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_FLOAT, bh_4_bw_2_2) { |
| for(size_t i = 2; i < 2; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_FLOAT, bh_4_bw_1) { |
| TransposeMicrokernelTester() |
| .input_stride(1) |
| .output_stride(4) |
| .block_width(1) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_FLOAT, bh_3_4_bw_1){ |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(1) |
| .output_stride(i) |
| .block_width(1) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_FLOAT, bh_3_4_bw_2){ |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(i) |
| .block_width(2) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_FLOAT, bh_3_4_bw_2_2) { |
| for(size_t i = 3; i < 4; ++i){ |
| for(size_t j = 2; j < 2; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_float); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_FLOAT, bh_2_bw_1_is_2) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(2) |
| .block_width(1) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_FLOAT, bh_2_bw_1_os_4) { |
| TransposeMicrokernelTester() |
| .input_stride(1) |
| .output_stride(4) |
| .block_width(1) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__2X1_SCALAR_FLOAT, bh_2_bw_1_is_2_os_4) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(4) |
| .block_width(1) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x1_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_FLOAT, bh_2_bw_2) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(2) |
| .block_width(2) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_FLOAT, bh_1_4_bw_1_4) { |
| for(size_t i = 1; i <= 4; ++i){ |
| for(size_t j = 1; j <= 4; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_float); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_FLOAT, bh_2_bw_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(2) |
| .block_width(4) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_FLOAT, bh_2_bw_3_4) { |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(2) |
| .block_width(i) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_FLOAT, bh_4_bw_3_4) { |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_FLOAT, bh_4_bw_2) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(4) |
| .block_width(2) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_FLOAT, bh_3_4_bw_2){ |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(i) |
| .block_width(2) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_FLOAT, bh_3_4_bw_4){ |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_FLOAT, bh_3_4_bw_3_4) { |
| for(size_t i = 3; i < 4; ++i){ |
| for(size_t j = 3; j < 4; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_float); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_FLOAT, bh_2_bw_2_is_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(2) |
| .block_width(2) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_FLOAT, bh_2_bw_2_os_4) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(4) |
| .block_width(2) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__2X2_SCALAR_FLOAT, bh_2_bw_2_is_4_os_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(2) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x2_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_FLOAT, bh_2_bw_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(2) |
| .block_width(4) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_FLOAT, bh_1_4_bw_1_8) { |
| for(size_t i = 1; i <= 4; ++i){ |
| for(size_t j = 1; j <= 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_float); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_FLOAT, bh_2_bw_8) { |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(2) |
| .block_width(8) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_FLOAT, bh_2_bw_5_8) { |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(2) |
| .block_width(i) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_FLOAT, bh_4_bw_5_8) { |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_FLOAT, bh_4_bw_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_FLOAT, bh_3_4_bw_4){ |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_FLOAT, bh_3_4_bw_8){ |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(i) |
| .block_width(8) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_FLOAT, bh_3_4_bw_5_8) { |
| for(size_t i = 3; i < 4; ++i){ |
| for(size_t j = 5; j < 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_float); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_FLOAT, bh_2_bw_4_is_8) { |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(2) |
| .block_width(4) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_FLOAT, bh_2_bw_4_os_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__2X4_SCALAR_FLOAT, bh_2_bw_4_is_8_os_4) { |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(2) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__2x4_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_FLOAT, bh_4_bw_1) { |
| TransposeMicrokernelTester() |
| .input_stride(1) |
| .output_stride(4) |
| .block_width(1) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_FLOAT, bh_1_8_bw_1_2) { |
| for(size_t i = 1; i <= 8; ++i){ |
| for(size_t j = 1; j <= 2; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_float); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_FLOAT, bh_4_bw_2) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(4) |
| .block_width(2) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_FLOAT, bh_4_bw_2_2) { |
| for(size_t i = 2; i < 2; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_FLOAT, bh_8_bw_2_2) { |
| for(size_t i = 2; i < 2; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(8) |
| .block_width(i) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_FLOAT, bh_8_bw_1) { |
| TransposeMicrokernelTester() |
| .input_stride(1) |
| .output_stride(8) |
| .block_width(1) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_FLOAT, bh_5_8_bw_1){ |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(1) |
| .output_stride(i) |
| .block_width(1) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_FLOAT, bh_5_8_bw_2){ |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(i) |
| .block_width(2) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_FLOAT, bh_5_8_bw_2_2) { |
| for(size_t i = 5; i < 8; ++i){ |
| for(size_t j = 2; j < 2; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_float); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_FLOAT, bh_4_bw_1_is_2) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(4) |
| .block_width(1) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_FLOAT, bh_4_bw_1_os_8) { |
| TransposeMicrokernelTester() |
| .input_stride(1) |
| .output_stride(8) |
| .block_width(1) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__4X1_SCALAR_FLOAT, bh_4_bw_1_is_2_os_8) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(8) |
| .block_width(1) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x1_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_FLOAT, bh_4_bw_2) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(4) |
| .block_width(2) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_FLOAT, bh_1_8_bw_1_4) { |
| for(size_t i = 1; i <= 8; ++i){ |
| for(size_t j = 1; j <= 4; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_float); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_FLOAT, bh_4_bw_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_FLOAT, bh_4_bw_3_4) { |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_FLOAT, bh_8_bw_3_4) { |
| for(size_t i = 3; i < 4; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(8) |
| .block_width(i) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_FLOAT, bh_8_bw_2) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(8) |
| .block_width(2) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_FLOAT, bh_5_8_bw_2){ |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(i) |
| .block_width(2) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_FLOAT, bh_5_8_bw_4){ |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_FLOAT, bh_5_8_bw_3_4) { |
| for(size_t i = 5; i < 8; ++i){ |
| for(size_t j = 3; j < 4; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_float); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_FLOAT, bh_4_bw_2_is_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(2) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_FLOAT, bh_4_bw_2_os_8) { |
| TransposeMicrokernelTester() |
| .input_stride(2) |
| .output_stride(8) |
| .block_width(2) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__4X2_SCALAR_FLOAT, bh_4_bw_2_is_4_os_8) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(2) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x2_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_FLOAT, bh_4_bw_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_FLOAT, bh_1_8_bw_1_8) { |
| for(size_t i = 1; i <= 8; ++i){ |
| for(size_t j = 1; j <= 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_float); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_FLOAT, bh_4_bw_8) { |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(8) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_FLOAT, bh_4_bw_5_8) { |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_FLOAT, bh_8_bw_5_8) { |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(8) |
| .block_width(i) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_FLOAT, bh_8_bw_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_FLOAT, bh_5_8_bw_4){ |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_FLOAT, bh_5_8_bw_8){ |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(i) |
| .block_width(8) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_float); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_FLOAT, bh_5_8_bw_5_8) { |
| for(size_t i = 5; i < 8; ++i){ |
| for(size_t j = 5; j < 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_float); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_FLOAT, bh_4_bw_4_is_8) { |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_FLOAT, bh_4_bw_4_os_8) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_float); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SCALAR_FLOAT, bh_4_bw_4_is_8_os_8) { |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_scalar_float); |
| } |
| |
| #if XNN_ARCH_ARM64 |
| TEST(X32_TRANSPOSE__4X4_AARCH64_NEON_TBL, bh_4_bw_4) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_aarch64_neon_tbl); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_AARCH64_NEON_TBL, bh_1_8_bw_1_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 1; i <= 8; ++i){ |
| for(size_t j = 1; j <= 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_aarch64_neon_tbl); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_AARCH64_NEON_TBL, bh_4_bw_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(8) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_aarch64_neon_tbl); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_AARCH64_NEON_TBL, bh_4_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_aarch64_neon_tbl); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_AARCH64_NEON_TBL, bh_8_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(8) |
| .block_width(i) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_aarch64_neon_tbl); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_AARCH64_NEON_TBL, bh_8_bw_4) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_aarch64_neon_tbl); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_AARCH64_NEON_TBL, bh_5_8_bw_4){ |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_aarch64_neon_tbl); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_AARCH64_NEON_TBL, bh_5_8_bw_8){ |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(i) |
| .block_width(8) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_aarch64_neon_tbl); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_AARCH64_NEON_TBL, bh_5_8_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| for(size_t j = 5; j < 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_aarch64_neon_tbl); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_AARCH64_NEON_TBL, bh_4_bw_4_is_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_aarch64_neon_tbl); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_AARCH64_NEON_TBL, bh_4_bw_4_os_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_aarch64_neon_tbl); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_AARCH64_NEON_TBL, bh_4_bw_4_is_8_os_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_aarch64_neon_tbl); |
| } |
| #endif // XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_SSE2, bh_4_bw_4) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_SSE2, bh_1_8_bw_1_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 1; i <= 8; ++i){ |
| for(size_t j = 1; j <= 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_sse2); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_SSE2, bh_4_bw_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(8) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_SSE2, bh_4_bw_5_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_SSE2, bh_8_bw_5_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(8) |
| .block_width(i) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_SSE2, bh_8_bw_4) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_SSE2, bh_5_8_bw_4){ |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_SSE2, bh_5_8_bw_8){ |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(i) |
| .block_width(8) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_SSE2, bh_5_8_bw_5_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| for(size_t j = 5; j < 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_sse2); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_SSE2, bh_4_bw_4_is_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_SSE2, bh_4_bw_4_os_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_SSE2, bh_4_bw_4_is_8_os_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_sse2); |
| } |
| #endif // XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| |
| |
| #if XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_SSE2, bh_4_bw_4) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_SSE2, bh_1_8_bw_1_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 1; i <= 8; ++i){ |
| for(size_t j = 1; j <= 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_sse2); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_SSE2, bh_4_bw_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(8) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_SSE2, bh_4_bw_5_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_SSE2, bh_8_bw_5_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(8) |
| .block_width(i) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_SSE2, bh_8_bw_4) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_SSE2, bh_5_8_bw_4){ |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_SSE2, bh_5_8_bw_8){ |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(i) |
| .block_width(8) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_SSE2, bh_5_8_bw_5_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| for(size_t j = 5; j < 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_sse2); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_SSE2, bh_4_bw_4_is_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_SSE2, bh_4_bw_4_os_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_SSE2, bh_4_bw_4_is_8_os_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_sse2); |
| } |
| #endif // XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| |
| |
| #if XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_SSE2, bh_4_bw_4) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_SSE2, bh_1_8_bw_1_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 1; i <= 8; ++i){ |
| for(size_t j = 1; j <= 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_sse2); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_SSE2, bh_4_bw_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(8) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_SSE2, bh_4_bw_5_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_SSE2, bh_8_bw_5_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(8) |
| .block_width(i) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_SSE2, bh_8_bw_4) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_SSE2, bh_5_8_bw_4){ |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_SSE2, bh_5_8_bw_8){ |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(i) |
| .block_width(8) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_SSE2, bh_5_8_bw_5_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| for(size_t j = 5; j < 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_sse2); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_SSE2, bh_4_bw_4_is_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_SSE2, bh_4_bw_4_os_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_SSE2, bh_4_bw_4_is_8_os_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_sse2); |
| } |
| #endif // XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| |
| |
| #if XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_SSE2, bh_4_bw_4) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_SSE2, bh_1_8_bw_1_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 1; i <= 8; ++i){ |
| for(size_t j = 1; j <= 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_sse2); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_SSE2, bh_4_bw_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(8) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_SSE2, bh_4_bw_5_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_SSE2, bh_8_bw_5_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(8) |
| .block_width(i) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_SSE2, bh_8_bw_4) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_SSE2, bh_5_8_bw_4){ |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_SSE2, bh_5_8_bw_8){ |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(i) |
| .block_width(8) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_SSE2, bh_5_8_bw_5_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| for(size_t j = 5; j < 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_sse2); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_SSE2, bh_4_bw_4_is_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_SSE2, bh_4_bw_4_os_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_SSE2, bh_4_bw_4_is_8_os_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_sse2); |
| } |
| #endif // XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| |
| |
| #if XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_SSE2, bh_4_bw_4) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_SSE2, bh_1_8_bw_1_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 1; i <= 8; ++i){ |
| for(size_t j = 1; j <= 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_sse2); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_SSE2, bh_4_bw_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(8) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_SSE2, bh_4_bw_5_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_SSE2, bh_8_bw_5_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(8) |
| .block_width(i) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_SSE2, bh_8_bw_4) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_SSE2, bh_5_8_bw_4){ |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_SSE2, bh_5_8_bw_8){ |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(i) |
| .block_width(8) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_SSE2, bh_5_8_bw_5_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| for(size_t j = 5; j < 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_sse2); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_SSE2, bh_4_bw_4_is_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_SSE2, bh_4_bw_4_os_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_SSE2, bh_4_bw_4_is_8_os_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_sse2); |
| } |
| #endif // XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| |
| |
| #if XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_SSE2, bh_4_bw_4) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_SSE2, bh_1_8_bw_1_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 1; i <= 8; ++i){ |
| for(size_t j = 1; j <= 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_sse2); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_SSE2, bh_4_bw_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(8) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_SSE2, bh_4_bw_5_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_SSE2, bh_8_bw_5_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(8) |
| .block_width(i) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_SSE2, bh_8_bw_4) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_SSE2, bh_5_8_bw_4){ |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_SSE2, bh_5_8_bw_8){ |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(i) |
| .block_width(8) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_sse2); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_SSE2, bh_5_8_bw_5_8) { |
| TEST_REQUIRES_X86_SSE2; |
| for(size_t i = 5; i < 8; ++i){ |
| for(size_t j = 5; j < 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_sse2); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_SSE2, bh_4_bw_4_is_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_SSE2, bh_4_bw_4_os_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_sse2); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_SSE2, bh_4_bw_4_is_8_os_8) { |
| TEST_REQUIRES_X86_SSE2; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_sse2); |
| } |
| #endif // XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| |
| |
| #if XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| TEST(X32_TRANSPOSE__4X4_SSE, bh_4_bw_4) { |
| TEST_REQUIRES_X86_SSE; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_sse); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SSE, bh_1_8_bw_1_8) { |
| TEST_REQUIRES_X86_SSE; |
| for(size_t i = 1; i <= 8; ++i){ |
| for(size_t j = 1; j <= 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_sse); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SSE, bh_4_bw_8) { |
| TEST_REQUIRES_X86_SSE; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(8) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_sse); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SSE, bh_4_bw_5_8) { |
| TEST_REQUIRES_X86_SSE; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_sse); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SSE, bh_8_bw_5_8) { |
| TEST_REQUIRES_X86_SSE; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(8) |
| .block_width(i) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_sse); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SSE, bh_8_bw_4) { |
| TEST_REQUIRES_X86_SSE; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_sse); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SSE, bh_5_8_bw_4){ |
| TEST_REQUIRES_X86_SSE; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_sse); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SSE, bh_5_8_bw_8){ |
| TEST_REQUIRES_X86_SSE; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(i) |
| .block_width(8) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_sse); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SSE, bh_5_8_bw_5_8) { |
| TEST_REQUIRES_X86_SSE; |
| for(size_t i = 5; i < 8; ++i){ |
| for(size_t j = 5; j < 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_sse); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SSE, bh_4_bw_4_is_8) { |
| TEST_REQUIRES_X86_SSE; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_sse); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SSE, bh_4_bw_4_os_8) { |
| TEST_REQUIRES_X86_SSE; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_sse); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_SSE, bh_4_bw_4_is_8_os_8) { |
| TEST_REQUIRES_X86_SSE; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_sse); |
| } |
| #endif // XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| |
| |
| #if XNN_ARCH_WASMSIMD || XNN_ARCH_WASMRELAXEDSIMD |
| TEST(X32_TRANSPOSE__4X4_WASMSIMD, bh_4_bw_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_wasmsimd); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_WASMSIMD, bh_1_8_bw_1_8) { |
| for(size_t i = 1; i <= 8; ++i){ |
| for(size_t j = 1; j <= 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_wasmsimd); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_WASMSIMD, bh_4_bw_8) { |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(8) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_wasmsimd); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_WASMSIMD, bh_4_bw_5_8) { |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_wasmsimd); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_WASMSIMD, bh_8_bw_5_8) { |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(8) |
| .block_width(i) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_wasmsimd); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_WASMSIMD, bh_8_bw_4) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_wasmsimd); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_WASMSIMD, bh_5_8_bw_4){ |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_wasmsimd); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_WASMSIMD, bh_5_8_bw_8){ |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(i) |
| .block_width(8) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_wasmsimd); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_WASMSIMD, bh_5_8_bw_5_8) { |
| for(size_t i = 5; i < 8; ++i){ |
| for(size_t j = 5; j < 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_wasmsimd); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_WASMSIMD, bh_4_bw_4_is_8) { |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_wasmsimd); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_WASMSIMD, bh_4_bw_4_os_8) { |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_wasmsimd); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_WASMSIMD, bh_4_bw_4_is_8_os_8) { |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_wasmsimd); |
| } |
| #endif // XNN_ARCH_WASMSIMD || XNN_ARCH_WASMRELAXEDSIMD |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(X32_TRANSPOSE__4X4_MULTI_DEC_ZIP_NEON, bh_4_bw_4) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_dec_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_DEC_ZIP_NEON, bh_1_8_bw_1_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 1; i <= 8; ++i){ |
| for(size_t j = 1; j <= 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_dec_zip_neon); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_DEC_ZIP_NEON, bh_4_bw_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(8) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_dec_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_DEC_ZIP_NEON, bh_4_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_dec_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_DEC_ZIP_NEON, bh_8_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(8) |
| .block_width(i) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_dec_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_DEC_ZIP_NEON, bh_8_bw_4) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_dec_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_DEC_ZIP_NEON, bh_5_8_bw_4){ |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_dec_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_DEC_ZIP_NEON, bh_5_8_bw_8){ |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(i) |
| .block_width(8) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_dec_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_DEC_ZIP_NEON, bh_5_8_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| for(size_t j = 5; j < 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_dec_zip_neon); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_DEC_ZIP_NEON, bh_4_bw_4_is_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_dec_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_DEC_ZIP_NEON, bh_4_bw_4_os_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_dec_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_DEC_ZIP_NEON, bh_4_bw_4_is_8_os_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_dec_zip_neon); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_ZIP_NEON, bh_4_bw_4) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_ZIP_NEON, bh_1_8_bw_1_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 1; i <= 8; ++i){ |
| for(size_t j = 1; j <= 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_zip_neon); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_ZIP_NEON, bh_4_bw_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(8) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_ZIP_NEON, bh_4_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_ZIP_NEON, bh_8_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(8) |
| .block_width(i) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_ZIP_NEON, bh_8_bw_4) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_ZIP_NEON, bh_5_8_bw_4){ |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_ZIP_NEON, bh_5_8_bw_8){ |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(i) |
| .block_width(8) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_ZIP_NEON, bh_5_8_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| for(size_t j = 5; j < 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_zip_neon); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_ZIP_NEON, bh_4_bw_4_is_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_ZIP_NEON, bh_4_bw_4_os_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MOV_ZIP_NEON, bh_4_bw_4_is_8_os_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_mov_zip_neon); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_ZIP_NEON, bh_4_bw_4) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_ZIP_NEON, bh_1_8_bw_1_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 1; i <= 8; ++i){ |
| for(size_t j = 1; j <= 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_zip_neon); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_ZIP_NEON, bh_4_bw_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(8) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_ZIP_NEON, bh_4_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_ZIP_NEON, bh_8_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(8) |
| .block_width(i) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_ZIP_NEON, bh_8_bw_4) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_ZIP_NEON, bh_5_8_bw_4){ |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_ZIP_NEON, bh_5_8_bw_8){ |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(i) |
| .block_width(8) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_ZIP_NEON, bh_5_8_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| for(size_t j = 5; j < 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_zip_neon); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_ZIP_NEON, bh_4_bw_4_is_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_ZIP_NEON, bh_4_bw_4_os_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_MULTI_ZIP_NEON, bh_4_bw_4_is_8_os_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_multi_zip_neon); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_ZIP_NEON, bh_4_bw_4) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_ZIP_NEON, bh_1_8_bw_1_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 1; i <= 8; ++i){ |
| for(size_t j = 1; j <= 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_zip_neon); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_ZIP_NEON, bh_4_bw_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(8) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_ZIP_NEON, bh_4_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_ZIP_NEON, bh_8_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(8) |
| .block_width(i) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_ZIP_NEON, bh_8_bw_4) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_ZIP_NEON, bh_5_8_bw_4){ |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_ZIP_NEON, bh_5_8_bw_8){ |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(i) |
| .block_width(8) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_ZIP_NEON, bh_5_8_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| for(size_t j = 5; j < 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_zip_neon); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_ZIP_NEON, bh_4_bw_4_is_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_ZIP_NEON, bh_4_bw_4_os_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_MULTI_SWITCH_ZIP_NEON, bh_4_bw_4_is_8_os_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_multi_switch_zip_neon); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(X32_TRANSPOSE__4X4_REUSE_DEC_ZIP_NEON, bh_4_bw_4) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_dec_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_DEC_ZIP_NEON, bh_1_8_bw_1_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 1; i <= 8; ++i){ |
| for(size_t j = 1; j <= 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_dec_zip_neon); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_DEC_ZIP_NEON, bh_4_bw_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(8) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_dec_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_DEC_ZIP_NEON, bh_4_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_dec_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_DEC_ZIP_NEON, bh_8_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(8) |
| .block_width(i) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_dec_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_DEC_ZIP_NEON, bh_8_bw_4) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_dec_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_DEC_ZIP_NEON, bh_5_8_bw_4){ |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_dec_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_DEC_ZIP_NEON, bh_5_8_bw_8){ |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(i) |
| .block_width(8) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_dec_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_DEC_ZIP_NEON, bh_5_8_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| for(size_t j = 5; j < 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_dec_zip_neon); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_DEC_ZIP_NEON, bh_4_bw_4_is_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_dec_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_DEC_ZIP_NEON, bh_4_bw_4_os_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_dec_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_DEC_ZIP_NEON, bh_4_bw_4_is_8_os_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_dec_zip_neon); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_ZIP_NEON, bh_4_bw_4) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_ZIP_NEON, bh_1_8_bw_1_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 1; i <= 8; ++i){ |
| for(size_t j = 1; j <= 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_zip_neon); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_ZIP_NEON, bh_4_bw_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(8) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_ZIP_NEON, bh_4_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_ZIP_NEON, bh_8_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(8) |
| .block_width(i) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_ZIP_NEON, bh_8_bw_4) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_ZIP_NEON, bh_5_8_bw_4){ |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_ZIP_NEON, bh_5_8_bw_8){ |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(i) |
| .block_width(8) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_ZIP_NEON, bh_5_8_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| for(size_t j = 5; j < 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_zip_neon); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_ZIP_NEON, bh_4_bw_4_is_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_ZIP_NEON, bh_4_bw_4_os_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MOV_ZIP_NEON, bh_4_bw_4_is_8_os_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_mov_zip_neon); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_ZIP_NEON, bh_4_bw_4) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_ZIP_NEON, bh_1_8_bw_1_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 1; i <= 8; ++i){ |
| for(size_t j = 1; j <= 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_zip_neon); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_ZIP_NEON, bh_4_bw_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(8) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_ZIP_NEON, bh_4_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_ZIP_NEON, bh_8_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(8) |
| .block_width(i) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_ZIP_NEON, bh_8_bw_4) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_ZIP_NEON, bh_5_8_bw_4){ |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_ZIP_NEON, bh_5_8_bw_8){ |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(i) |
| .block_width(8) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_ZIP_NEON, bh_5_8_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| for(size_t j = 5; j < 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_zip_neon); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_ZIP_NEON, bh_4_bw_4_is_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_ZIP_NEON, bh_4_bw_4_os_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_MULTI_ZIP_NEON, bh_4_bw_4_is_8_os_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_multi_zip_neon); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_ZIP_NEON, bh_4_bw_4) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_ZIP_NEON, bh_1_8_bw_1_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 1; i <= 8; ++i){ |
| for(size_t j = 1; j <= 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_zip_neon); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_ZIP_NEON, bh_4_bw_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(8) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_ZIP_NEON, bh_4_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(4) |
| .block_width(i) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_ZIP_NEON, bh_8_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(i) |
| .output_stride(8) |
| .block_width(i) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_ZIP_NEON, bh_8_bw_4) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(8) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_ZIP_NEON, bh_5_8_bw_4){ |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(i) |
| .block_width(4) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_ZIP_NEON, bh_5_8_bw_8){ |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(i) |
| .block_width(8) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_zip_neon); |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_ZIP_NEON, bh_5_8_bw_5_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for(size_t i = 5; i < 8; ++i){ |
| for(size_t j = 5; j < 8; ++j){ |
| TransposeMicrokernelTester() |
| .input_stride(j) |
| .output_stride(i) |
| .block_width(j) |
| .block_height(i) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_zip_neon); |
| } |
| } |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_ZIP_NEON, bh_4_bw_4_is_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(4) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_ZIP_NEON, bh_4_bw_4_os_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(4) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_zip_neon); |
| } |
| |
| TEST(X32_TRANSPOSE__4X4_REUSE_SWITCH_ZIP_NEON, bh_4_bw_4_is_8_os_8) { |
| TEST_REQUIRES_ARM_NEON; |
| TransposeMicrokernelTester() |
| .input_stride(8) |
| .output_stride(8) |
| .block_width(4) |
| .block_height(4) |
| .iterations(1) |
| .Test(xnn_x32_transpose_ukernel__4x4_reuse_switch_zip_neon); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |