| // Copyright (c) Facebook, Inc. and its affiliates. |
| // All rights reserved. |
| // |
| // Copyright 2019 Google LLC |
| // |
| // This source code is licensed under the BSD-style license found in the |
| // LICENSE file in the root directory of this source tree. |
| // |
| // Auto-generated file. Do not edit! |
| // Specification: test/f32-ppmm.yaml |
| // Generator: tools/generate-gemm-test.py |
| |
| |
| #include <gtest/gtest.h> |
| |
| #include <xnnpack/common.h> |
| #include <xnnpack/isa-checks.h> |
| |
| #include <xnnpack/gemm.h> |
| #include <xnnpack/igemm.h> |
| #include <xnnpack/ppmm.h> |
| #include "gemm-microkernel-tester.h" |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(F32_PPMM_4X8__NEON, k_eq_1) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neon); |
| } |
| |
| TEST(F32_PPMM_4X8__NEON, strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .cn_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neon); |
| } |
| |
| TEST(F32_PPMM_4X8__NEON, k_eq_1_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .a_stride(3) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neon); |
| } |
| |
| TEST(F32_PPMM_4X8__NEON, k_eq_1_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neon); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEON, k_eq_1_subtile_m) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neon); |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEON, k_eq_1_subtile_n) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neon); |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEON, k_gt_1) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neon); |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEON, k_gt_1_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 2; k < 10; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neon); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEON, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neon); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEON, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neon); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEON, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neon); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEON, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neon); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEON, n_div_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neon); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEON, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neon); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEON, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neon); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEON, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neon); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEON, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neon); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEON, qmin) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .qmin(128) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neon); |
| } |
| |
| TEST(F32_PPMM_4X8__NEON, qmax) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .qmax(128) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neon); |
| } |
| |
| TEST(F32_PPMM_4X8__NEON, strided_cm) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .cm_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neon); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(F32_PPMM_4X8__NEONFMA, k_eq_1) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neonfma); |
| } |
| |
| TEST(F32_PPMM_4X8__NEONFMA, strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .cn_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neonfma); |
| } |
| |
| TEST(F32_PPMM_4X8__NEONFMA, k_eq_1_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .a_stride(3) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neonfma); |
| } |
| |
| TEST(F32_PPMM_4X8__NEONFMA, k_eq_1_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neonfma); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEONFMA, k_eq_1_subtile_m) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neonfma); |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEONFMA, k_eq_1_subtile_n) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neonfma); |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEONFMA, k_gt_1) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neonfma); |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEONFMA, k_gt_1_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 2; k < 10; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neonfma); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEONFMA, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neonfma); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEONFMA, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neonfma); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEONFMA, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neonfma); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEONFMA, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neonfma); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEONFMA, n_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neonfma); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEONFMA, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neonfma); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEONFMA, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neonfma); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEONFMA, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neonfma); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEONFMA, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neonfma); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__NEONFMA, qmin) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .qmin(128) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neonfma); |
| } |
| |
| TEST(F32_PPMM_4X8__NEONFMA, qmax) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .qmax(128) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neonfma); |
| } |
| |
| TEST(F32_PPMM_4X8__NEONFMA, strided_cm) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .cm_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_4x8__neonfma); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(F32_PPMM_8X8__NEON, k_eq_1) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(8) |
| .k(1) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neon); |
| } |
| |
| TEST(F32_PPMM_8X8__NEON, strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(8) |
| .k(1) |
| .cn_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neon); |
| } |
| |
| TEST(F32_PPMM_8X8__NEON, k_eq_1_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(8) |
| .k(1) |
| .a_stride(3) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neon); |
| } |
| |
| TEST(F32_PPMM_8X8__NEON, k_eq_1_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t m = 1; m <= 8; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neon); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEON, k_eq_1_subtile_m) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t m = 1; m <= 8; m++) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neon); |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEON, k_eq_1_subtile_n) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neon); |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEON, k_gt_1) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neon); |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEON, k_gt_1_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 2; k < 10; k++) { |
| for (uint32_t m = 1; m <= 8; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neon); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEON, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neon); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEON, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neon); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEON, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neon); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEON, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 8; m++) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neon); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEON, n_div_8) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neon); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEON, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neon); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEON, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neon); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEON, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 8; m++) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neon); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEON, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON; |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 8; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neon); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEON, qmin) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(8) |
| .k(1) |
| .qmin(128) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neon); |
| } |
| |
| TEST(F32_PPMM_8X8__NEON, qmax) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(8) |
| .k(1) |
| .qmax(128) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neon); |
| } |
| |
| TEST(F32_PPMM_8X8__NEON, strided_cm) { |
| TEST_REQUIRES_ARM_NEON; |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(8) |
| .k(1) |
| .cm_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neon); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| TEST(F32_PPMM_8X8__NEONFMA, k_eq_1) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(8) |
| .k(1) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neonfma); |
| } |
| |
| TEST(F32_PPMM_8X8__NEONFMA, strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(8) |
| .k(1) |
| .cn_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neonfma); |
| } |
| |
| TEST(F32_PPMM_8X8__NEONFMA, k_eq_1_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(8) |
| .k(1) |
| .a_stride(3) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neonfma); |
| } |
| |
| TEST(F32_PPMM_8X8__NEONFMA, k_eq_1_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 8; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neonfma); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEONFMA, k_eq_1_subtile_m) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t m = 1; m <= 8; m++) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neonfma); |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEONFMA, k_eq_1_subtile_n) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neonfma); |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEONFMA, k_gt_1) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neonfma); |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEONFMA, k_gt_1_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 2; k < 10; k++) { |
| for (uint32_t m = 1; m <= 8; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neonfma); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEONFMA, n_gt_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neonfma); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEONFMA, n_gt_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neonfma); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEONFMA, n_gt_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neonfma); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEONFMA, n_gt_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 8; m++) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neonfma); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEONFMA, n_div_8) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neonfma); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEONFMA, n_div_8_strided_cn) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neonfma); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEONFMA, n_div_8_strided_a) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neonfma); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEONFMA, n_div_8_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 8; m++) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neonfma); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEONFMA, strided_cm_subtile) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 8; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neonfma); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_8X8__NEONFMA, qmin) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(8) |
| .k(1) |
| .qmin(128) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neonfma); |
| } |
| |
| TEST(F32_PPMM_8X8__NEONFMA, qmax) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(8) |
| .k(1) |
| .qmax(128) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neonfma); |
| } |
| |
| TEST(F32_PPMM_8X8__NEONFMA, strided_cm) { |
| TEST_REQUIRES_ARM_NEON_FMA; |
| GemmMicrokernelTester() |
| .mr(8) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(8) |
| .n(8) |
| .k(1) |
| .cm_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_8x8__neonfma); |
| } |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| TEST(F32_PPMM_4X8__SSE, k_eq_1) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__sse); |
| } |
| |
| TEST(F32_PPMM_4X8__SSE, strided_cn) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .cn_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_4x8__sse); |
| } |
| |
| TEST(F32_PPMM_4X8__SSE, k_eq_1_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .a_stride(3) |
| .Test(xnn_f32_ppmm_ukernel_4x8__sse); |
| } |
| |
| TEST(F32_PPMM_4X8__SSE, k_eq_1_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__sse); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__SSE, k_eq_1_subtile_m) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__sse); |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__SSE, k_eq_1_subtile_n) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__sse); |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__SSE, k_gt_1) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_4x8__sse); |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__SSE, k_gt_1_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 2; k < 10; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__sse); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__SSE, n_gt_8) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_4x8__sse); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__SSE, n_gt_8_strided_cn) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_4x8__sse); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__SSE, n_gt_8_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_4x8__sse); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__SSE, n_gt_8_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__sse); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__SSE, n_div_8) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_4x8__sse); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__SSE, n_div_8_strided_cn) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_4x8__sse); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__SSE, n_div_8_strided_a) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_4x8__sse); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__SSE, n_div_8_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__sse); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__SSE, strided_cm_subtile) { |
| TEST_REQUIRES_X86_SSE; |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__sse); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__SSE, qmin) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .qmin(128) |
| .Test(xnn_f32_ppmm_ukernel_4x8__sse); |
| } |
| |
| TEST(F32_PPMM_4X8__SSE, qmax) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .qmax(128) |
| .Test(xnn_f32_ppmm_ukernel_4x8__sse); |
| } |
| |
| TEST(F32_PPMM_4X8__SSE, strided_cm) { |
| TEST_REQUIRES_X86_SSE; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .cm_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_4x8__sse); |
| } |
| #endif // XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| |
| |
| #if !XNN_ARCH_ASMJS && !XNN_ARCH_WASM |
| TEST(F32_PPMM_4X8__PSIMD, k_eq_1) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_4X8__PSIMD, strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .cn_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_4x8__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_4X8__PSIMD, k_eq_1_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .a_stride(3) |
| .Test(xnn_f32_ppmm_ukernel_4x8__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_4X8__PSIMD, k_eq_1_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__PSIMD, k_eq_1_subtile_m) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(8) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__PSIMD, k_eq_1_subtile_n) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__PSIMD, k_gt_1) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_4x8__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__PSIMD, k_gt_1_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 2; k < 10; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__PSIMD, n_gt_8) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_4x8__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__PSIMD, n_gt_8_strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_4x8__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__PSIMD, n_gt_8_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_4x8__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__PSIMD, n_gt_8_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 9; n < 16; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__PSIMD, n_div_8) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_4x8__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__PSIMD, n_div_8_strided_cn) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_4x8__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__PSIMD, n_div_8_strided_a) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_4x8__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__PSIMD, n_div_8_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (uint32_t n = 16; n <= 24; n += 8) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__PSIMD, strided_cm_subtile) { |
| TEST_REQUIRES_PSIMD; |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 8; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(11) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x8__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X8__PSIMD, qmin) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .qmin(128) |
| .Test(xnn_f32_ppmm_ukernel_4x8__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_4X8__PSIMD, qmax) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .qmax(128) |
| .Test(xnn_f32_ppmm_ukernel_4x8__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_4X8__PSIMD, strided_cm) { |
| TEST_REQUIRES_PSIMD; |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(8) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(8) |
| .k(1) |
| .cm_stride(11) |
| .Test(xnn_f32_ppmm_ukernel_4x8__psimd, GemmMicrokernelTester::Variant::Scalar); |
| } |
| #endif // !XNN_ARCH_ASMJS && !XNN_ARCH_WASM |
| |
| |
| TEST(F32_PPMM_4X2__SCALAR, k_eq_1) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(1) |
| .Test(xnn_f32_ppmm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_4X2__SCALAR, strided_cn) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(1) |
| .cn_stride(5) |
| .Test(xnn_f32_ppmm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_4X2__SCALAR, k_eq_1_strided_a) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(1) |
| .a_stride(3) |
| .Test(xnn_f32_ppmm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_4X2__SCALAR, k_eq_1_subtile) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 2; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X2__SCALAR, k_eq_1_subtile_m) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(2) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_PPMM_4X2__SCALAR, k_eq_1_subtile_n) { |
| for (uint32_t n = 1; n <= 2; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_PPMM_4X2__SCALAR, k_gt_1) { |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_PPMM_4X2__SCALAR, k_gt_1_subtile) { |
| for (size_t k = 2; k < 10; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 2; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X2__SCALAR, n_gt_2) { |
| for (uint32_t n = 3; n < 4; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X2__SCALAR, n_gt_2_strided_cn) { |
| for (uint32_t n = 3; n < 4; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(k) |
| .cn_stride(5) |
| .Test(xnn_f32_ppmm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X2__SCALAR, n_gt_2_strided_a) { |
| for (uint32_t n = 3; n < 4; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X2__SCALAR, n_gt_2_subtile) { |
| for (uint32_t n = 3; n < 4; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X2__SCALAR, n_div_2) { |
| for (uint32_t n = 4; n <= 6; n += 2) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X2__SCALAR, n_div_2_strided_cn) { |
| for (uint32_t n = 4; n <= 6; n += 2) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(5) |
| .Test(xnn_f32_ppmm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X2__SCALAR, n_div_2_strided_a) { |
| for (uint32_t n = 4; n <= 6; n += 2) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X2__SCALAR, n_div_2_subtile) { |
| for (uint32_t n = 4; n <= 6; n += 2) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X2__SCALAR, strided_cm_subtile) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 2; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(5) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X2__SCALAR, qmin) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(1) |
| .qmin(128) |
| .Test(xnn_f32_ppmm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_4X2__SCALAR, qmax) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(1) |
| .qmax(128) |
| .Test(xnn_f32_ppmm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_4X2__SCALAR, strided_cm) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(2) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(2) |
| .k(1) |
| .cm_stride(5) |
| .Test(xnn_f32_ppmm_ukernel_4x2__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| |
| TEST(F32_PPMM_2X4__SCALAR, k_eq_1) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(4) |
| .k(1) |
| .Test(xnn_f32_ppmm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_2X4__SCALAR, strided_cn) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(4) |
| .k(1) |
| .cn_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_2X4__SCALAR, k_eq_1_strided_a) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(4) |
| .k(1) |
| .a_stride(3) |
| .Test(xnn_f32_ppmm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_2X4__SCALAR, k_eq_1_subtile) { |
| for (uint32_t m = 1; m <= 2; m++) { |
| for (uint32_t n = 1; n <= 4; n++) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_2X4__SCALAR, k_eq_1_subtile_m) { |
| for (uint32_t m = 1; m <= 2; m++) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(4) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_PPMM_2X4__SCALAR, k_eq_1_subtile_n) { |
| for (uint32_t n = 1; n <= 4; n++) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_PPMM_2X4__SCALAR, k_gt_1) { |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(4) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_PPMM_2X4__SCALAR, k_gt_1_subtile) { |
| for (size_t k = 2; k < 10; k++) { |
| for (uint32_t m = 1; m <= 2; m++) { |
| for (uint32_t n = 1; n <= 4; n++) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_2X4__SCALAR, n_gt_4) { |
| for (uint32_t n = 5; n < 8; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(4) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_2X4__SCALAR, n_gt_4_strided_cn) { |
| for (uint32_t n = 5; n < 8; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(4) |
| .k(k) |
| .cn_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_2X4__SCALAR, n_gt_4_strided_a) { |
| for (uint32_t n = 5; n < 8; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_2X4__SCALAR, n_gt_4_subtile) { |
| for (uint32_t n = 5; n < 8; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 2; m++) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_2X4__SCALAR, n_div_4) { |
| for (uint32_t n = 8; n <= 12; n += 4) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(4) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_2X4__SCALAR, n_div_4_strided_cn) { |
| for (uint32_t n = 8; n <= 12; n += 4) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(n) |
| .k(k) |
| .cn_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_2X4__SCALAR, n_div_4_strided_a) { |
| for (uint32_t n = 8; n <= 12; n += 4) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_2X4__SCALAR, n_div_4_subtile) { |
| for (uint32_t n = 8; n <= 12; n += 4) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 2; m++) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_2X4__SCALAR, strided_cm_subtile) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 2; m++) { |
| for (uint32_t n = 1; n <= 4; n++) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(7) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_2X4__SCALAR, qmin) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(4) |
| .k(1) |
| .qmin(128) |
| .Test(xnn_f32_ppmm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_2X4__SCALAR, qmax) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(4) |
| .k(1) |
| .qmax(128) |
| .Test(xnn_f32_ppmm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_2X4__SCALAR, strided_cm) { |
| GemmMicrokernelTester() |
| .mr(2) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(2) |
| .n(4) |
| .k(1) |
| .cm_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_2x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| |
| TEST(F32_PPMM_4X4__SCALAR, k_eq_1) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(4) |
| .k(1) |
| .Test(xnn_f32_ppmm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_4X4__SCALAR, strided_cn) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(4) |
| .k(1) |
| .cn_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_4X4__SCALAR, k_eq_1_strided_a) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(4) |
| .k(1) |
| .a_stride(3) |
| .Test(xnn_f32_ppmm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_4X4__SCALAR, k_eq_1_subtile) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 4; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X4__SCALAR, k_eq_1_subtile_m) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(4) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_PPMM_4X4__SCALAR, k_eq_1_subtile_n) { |
| for (uint32_t n = 1; n <= 4; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_PPMM_4X4__SCALAR, k_gt_1) { |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(4) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_PPMM_4X4__SCALAR, k_gt_1_subtile) { |
| for (size_t k = 2; k < 10; k++) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 4; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X4__SCALAR, n_gt_4) { |
| for (uint32_t n = 5; n < 8; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(4) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X4__SCALAR, n_gt_4_strided_cn) { |
| for (uint32_t n = 5; n < 8; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(4) |
| .k(k) |
| .cn_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X4__SCALAR, n_gt_4_strided_a) { |
| for (uint32_t n = 5; n < 8; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X4__SCALAR, n_gt_4_subtile) { |
| for (uint32_t n = 5; n < 8; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X4__SCALAR, n_div_4) { |
| for (uint32_t n = 8; n <= 12; n += 4) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(4) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X4__SCALAR, n_div_4_strided_cn) { |
| for (uint32_t n = 8; n <= 12; n += 4) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .cn_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X4__SCALAR, n_div_4_strided_a) { |
| for (uint32_t n = 8; n <= 12; n += 4) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X4__SCALAR, n_div_4_subtile) { |
| for (uint32_t n = 8; n <= 12; n += 4) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X4__SCALAR, strided_cm_subtile) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 4; m++) { |
| for (uint32_t n = 1; n <= 4; n++) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(7) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_4X4__SCALAR, qmin) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(4) |
| .k(1) |
| .qmin(128) |
| .Test(xnn_f32_ppmm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_4X4__SCALAR, qmax) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(4) |
| .k(1) |
| .qmax(128) |
| .Test(xnn_f32_ppmm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_4X4__SCALAR, strided_cm) { |
| GemmMicrokernelTester() |
| .mr(4) |
| .nr(4) |
| .kr(1) |
| .sr(1) |
| .m(4) |
| .n(4) |
| .k(1) |
| .cm_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_4x4__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| |
| TEST(F32_PPMM_3X3__SCALAR, k_eq_1) { |
| GemmMicrokernelTester() |
| .mr(3) |
| .nr(3) |
| .kr(1) |
| .sr(1) |
| .m(3) |
| .n(3) |
| .k(1) |
| .Test(xnn_f32_ppmm_ukernel_3x3__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_3X3__SCALAR, strided_cn) { |
| GemmMicrokernelTester() |
| .mr(3) |
| .nr(3) |
| .kr(1) |
| .sr(1) |
| .m(3) |
| .n(3) |
| .k(1) |
| .cn_stride(5) |
| .Test(xnn_f32_ppmm_ukernel_3x3__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_3X3__SCALAR, k_eq_1_strided_a) { |
| GemmMicrokernelTester() |
| .mr(3) |
| .nr(3) |
| .kr(1) |
| .sr(1) |
| .m(3) |
| .n(3) |
| .k(1) |
| .a_stride(3) |
| .Test(xnn_f32_ppmm_ukernel_3x3__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_3X3__SCALAR, k_eq_1_subtile) { |
| for (uint32_t m = 1; m <= 3; m++) { |
| for (uint32_t n = 1; n <= 3; n++) { |
| GemmMicrokernelTester() |
| .mr(3) |
| .nr(3) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_3x3__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_3X3__SCALAR, k_eq_1_subtile_m) { |
| for (uint32_t m = 1; m <= 3; m++) { |
| GemmMicrokernelTester() |
| .mr(3) |
| .nr(3) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(3) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_3x3__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_PPMM_3X3__SCALAR, k_eq_1_subtile_n) { |
| for (uint32_t n = 1; n <= 3; n++) { |
| GemmMicrokernelTester() |
| .mr(3) |
| .nr(3) |
| .kr(1) |
| .sr(1) |
| .m(3) |
| .n(n) |
| .k(1) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_3x3__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_PPMM_3X3__SCALAR, k_gt_1) { |
| for (size_t k = 2; k < 10; k++) { |
| GemmMicrokernelTester() |
| .mr(3) |
| .nr(3) |
| .kr(1) |
| .sr(1) |
| .m(3) |
| .n(3) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_3x3__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| |
| TEST(F32_PPMM_3X3__SCALAR, k_gt_1_subtile) { |
| for (size_t k = 2; k < 10; k++) { |
| for (uint32_t m = 1; m <= 3; m++) { |
| for (uint32_t n = 1; n <= 3; n++) { |
| GemmMicrokernelTester() |
| .mr(3) |
| .nr(3) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_3x3__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_3X3__SCALAR, n_gt_3) { |
| for (uint32_t n = 4; n < 6; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(3) |
| .nr(3) |
| .kr(1) |
| .sr(1) |
| .m(3) |
| .n(3) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_3x3__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_3X3__SCALAR, n_gt_3_strided_cn) { |
| for (uint32_t n = 4; n < 6; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(3) |
| .nr(3) |
| .kr(1) |
| .sr(1) |
| .m(3) |
| .n(3) |
| .k(k) |
| .cn_stride(5) |
| .Test(xnn_f32_ppmm_ukernel_3x3__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_3X3__SCALAR, n_gt_3_strided_a) { |
| for (uint32_t n = 4; n < 6; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(3) |
| .nr(3) |
| .kr(1) |
| .sr(1) |
| .m(3) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_3x3__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_3X3__SCALAR, n_gt_3_subtile) { |
| for (uint32_t n = 4; n < 6; n++) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 3; m++) { |
| GemmMicrokernelTester() |
| .mr(3) |
| .nr(3) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_3x3__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_3X3__SCALAR, n_div_3) { |
| for (uint32_t n = 6; n <= 9; n += 3) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(3) |
| .nr(3) |
| .kr(1) |
| .sr(1) |
| .m(3) |
| .n(3) |
| .k(k) |
| .Test(xnn_f32_ppmm_ukernel_3x3__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_3X3__SCALAR, n_div_3_strided_cn) { |
| for (uint32_t n = 6; n <= 9; n += 3) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(3) |
| .nr(3) |
| .kr(1) |
| .sr(1) |
| .m(3) |
| .n(n) |
| .k(k) |
| .cn_stride(5) |
| .Test(xnn_f32_ppmm_ukernel_3x3__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_3X3__SCALAR, n_div_3_strided_a) { |
| for (uint32_t n = 6; n <= 9; n += 3) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| GemmMicrokernelTester() |
| .mr(3) |
| .nr(3) |
| .kr(1) |
| .sr(1) |
| .m(3) |
| .n(n) |
| .k(k) |
| .a_stride(7) |
| .Test(xnn_f32_ppmm_ukernel_3x3__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_3X3__SCALAR, n_div_3_subtile) { |
| for (uint32_t n = 6; n <= 9; n += 3) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 3; m++) { |
| GemmMicrokernelTester() |
| .mr(3) |
| .nr(3) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_3x3__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_3X3__SCALAR, strided_cm_subtile) { |
| for (size_t k = 1; k <= 5; k += 2) { |
| for (uint32_t m = 1; m <= 3; m++) { |
| for (uint32_t n = 1; n <= 3; n++) { |
| GemmMicrokernelTester() |
| .mr(3) |
| .nr(3) |
| .kr(1) |
| .sr(1) |
| .m(m) |
| .n(n) |
| .k(k) |
| .cm_stride(5) |
| .iterations(1) |
| .Test(xnn_f32_ppmm_ukernel_3x3__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| } |
| } |
| } |
| |
| TEST(F32_PPMM_3X3__SCALAR, qmin) { |
| GemmMicrokernelTester() |
| .mr(3) |
| .nr(3) |
| .kr(1) |
| .sr(1) |
| .m(3) |
| .n(3) |
| .k(1) |
| .qmin(128) |
| .Test(xnn_f32_ppmm_ukernel_3x3__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_3X3__SCALAR, qmax) { |
| GemmMicrokernelTester() |
| .mr(3) |
| .nr(3) |
| .kr(1) |
| .sr(1) |
| .m(3) |
| .n(3) |
| .k(1) |
| .qmax(128) |
| .Test(xnn_f32_ppmm_ukernel_3x3__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
| |
| TEST(F32_PPMM_3X3__SCALAR, strided_cm) { |
| GemmMicrokernelTester() |
| .mr(3) |
| .nr(3) |
| .kr(1) |
| .sr(1) |
| .m(3) |
| .n(3) |
| .k(1) |
| .cm_stride(5) |
| .Test(xnn_f32_ppmm_ukernel_3x3__scalar, GemmMicrokernelTester::Variant::Scalar); |
| } |
