| // 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/f16-vdiv-minmax.yaml |
| // Generator: tools/generate-vbinary-test.py |
| |
| |
| #include <gtest/gtest.h> |
| |
| #include <xnnpack/common.h> |
| #include <xnnpack/isa-checks.h> |
| |
| #include <xnnpack/params-init.h> |
| #include <xnnpack/vbinary.h> |
| #include "vbinary-microkernel-tester.h" |
| |
| |
| #if XNN_ARCH_ARM64 |
| TEST(F16_VDIV_MINMAX__NEONFP16ARITH_X8, batch_eq_8) { |
| TEST_REQUIRES_ARM_NEON_FP16_ARITH; |
| VBinaryMicrokernelTester() |
| .batch_size(8) |
| .Test(xnn_f16_vdiv_minmax_ukernel__neonfp16arith_x8, VBinaryMicrokernelTester::OpType::Div); |
| } |
| |
| TEST(F16_VDIV_MINMAX__NEONFP16ARITH_X8, batch_div_8) { |
| TEST_REQUIRES_ARM_NEON_FP16_ARITH; |
| for (size_t batch_size = 16; batch_size < 80; batch_size += 8) { |
| VBinaryMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_f16_vdiv_minmax_ukernel__neonfp16arith_x8, VBinaryMicrokernelTester::OpType::Div); |
| } |
| } |
| |
| TEST(F16_VDIV_MINMAX__NEONFP16ARITH_X8, batch_lt_8) { |
| TEST_REQUIRES_ARM_NEON_FP16_ARITH; |
| for (size_t batch_size = 1; batch_size < 8; batch_size++) { |
| VBinaryMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_f16_vdiv_minmax_ukernel__neonfp16arith_x8, VBinaryMicrokernelTester::OpType::Div); |
| } |
| } |
| |
| TEST(F16_VDIV_MINMAX__NEONFP16ARITH_X8, batch_gt_8) { |
| TEST_REQUIRES_ARM_NEON_FP16_ARITH; |
| for (size_t batch_size = 9; batch_size < 16; batch_size++) { |
| VBinaryMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_f16_vdiv_minmax_ukernel__neonfp16arith_x8, VBinaryMicrokernelTester::OpType::Div); |
| } |
| } |
| |
| TEST(F16_VDIV_MINMAX__NEONFP16ARITH_X8, inplace_a) { |
| TEST_REQUIRES_ARM_NEON_FP16_ARITH; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| VBinaryMicrokernelTester() |
| .batch_size(batch_size) |
| .inplace_a(true) |
| .Test(xnn_f16_vdiv_minmax_ukernel__neonfp16arith_x8, VBinaryMicrokernelTester::OpType::Div); |
| } |
| } |
| |
| TEST(F16_VDIV_MINMAX__NEONFP16ARITH_X8, inplace_b) { |
| TEST_REQUIRES_ARM_NEON_FP16_ARITH; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| VBinaryMicrokernelTester() |
| .batch_size(batch_size) |
| .inplace_b(true) |
| .Test(xnn_f16_vdiv_minmax_ukernel__neonfp16arith_x8, VBinaryMicrokernelTester::OpType::Div); |
| } |
| } |
| |
| TEST(F16_VDIV_MINMAX__NEONFP16ARITH_X8, inplace_a_and_b) { |
| TEST_REQUIRES_ARM_NEON_FP16_ARITH; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| VBinaryMicrokernelTester() |
| .batch_size(batch_size) |
| .inplace_a(true) |
| .inplace_b(true) |
| .Test(xnn_f16_vdiv_minmax_ukernel__neonfp16arith_x8, VBinaryMicrokernelTester::OpType::Div); |
| } |
| } |
| |
| TEST(F16_VDIV_MINMAX__NEONFP16ARITH_X8, qmin) { |
| TEST_REQUIRES_ARM_NEON_FP16_ARITH; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| VBinaryMicrokernelTester() |
| .batch_size(batch_size) |
| .qmin(128) |
| .Test(xnn_f16_vdiv_minmax_ukernel__neonfp16arith_x8, VBinaryMicrokernelTester::OpType::Div); |
| } |
| } |
| |
| TEST(F16_VDIV_MINMAX__NEONFP16ARITH_X8, qmax) { |
| TEST_REQUIRES_ARM_NEON_FP16_ARITH; |
| for (size_t batch_size = 1; batch_size <= 40; batch_size += 7) { |
| VBinaryMicrokernelTester() |
| .batch_size(batch_size) |
| .qmax(128) |
| .Test(xnn_f16_vdiv_minmax_ukernel__neonfp16arith_x8, VBinaryMicrokernelTester::OpType::Div); |
| } |
| } |
| #endif // XNN_ARCH_ARM64 |
| |
| |
| #if XNN_ARCH_ARM64 |
| TEST(F16_VDIV_MINMAX__NEONFP16ARITH_X16, batch_eq_16) { |
| TEST_REQUIRES_ARM_NEON_FP16_ARITH; |
| VBinaryMicrokernelTester() |
| .batch_size(16) |
| .Test(xnn_f16_vdiv_minmax_ukernel__neonfp16arith_x16, VBinaryMicrokernelTester::OpType::Div); |
| } |
| |
| TEST(F16_VDIV_MINMAX__NEONFP16ARITH_X16, batch_div_16) { |
| TEST_REQUIRES_ARM_NEON_FP16_ARITH; |
| for (size_t batch_size = 32; batch_size < 160; batch_size += 16) { |
| VBinaryMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_f16_vdiv_minmax_ukernel__neonfp16arith_x16, VBinaryMicrokernelTester::OpType::Div); |
| } |
| } |
| |
| TEST(F16_VDIV_MINMAX__NEONFP16ARITH_X16, batch_lt_16) { |
| TEST_REQUIRES_ARM_NEON_FP16_ARITH; |
| for (size_t batch_size = 1; batch_size < 16; batch_size++) { |
| VBinaryMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_f16_vdiv_minmax_ukernel__neonfp16arith_x16, VBinaryMicrokernelTester::OpType::Div); |
| } |
| } |
| |
| TEST(F16_VDIV_MINMAX__NEONFP16ARITH_X16, batch_gt_16) { |
| TEST_REQUIRES_ARM_NEON_FP16_ARITH; |
| for (size_t batch_size = 17; batch_size < 32; batch_size++) { |
| VBinaryMicrokernelTester() |
| .batch_size(batch_size) |
| .Test(xnn_f16_vdiv_minmax_ukernel__neonfp16arith_x16, VBinaryMicrokernelTester::OpType::Div); |
| } |
| } |
| |
| TEST(F16_VDIV_MINMAX__NEONFP16ARITH_X16, inplace_a) { |
| TEST_REQUIRES_ARM_NEON_FP16_ARITH; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VBinaryMicrokernelTester() |
| .batch_size(batch_size) |
| .inplace_a(true) |
| .Test(xnn_f16_vdiv_minmax_ukernel__neonfp16arith_x16, VBinaryMicrokernelTester::OpType::Div); |
| } |
| } |
| |
| TEST(F16_VDIV_MINMAX__NEONFP16ARITH_X16, inplace_b) { |
| TEST_REQUIRES_ARM_NEON_FP16_ARITH; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VBinaryMicrokernelTester() |
| .batch_size(batch_size) |
| .inplace_b(true) |
| .Test(xnn_f16_vdiv_minmax_ukernel__neonfp16arith_x16, VBinaryMicrokernelTester::OpType::Div); |
| } |
| } |
| |
| TEST(F16_VDIV_MINMAX__NEONFP16ARITH_X16, inplace_a_and_b) { |
| TEST_REQUIRES_ARM_NEON_FP16_ARITH; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VBinaryMicrokernelTester() |
| .batch_size(batch_size) |
| .inplace_a(true) |
| .inplace_b(true) |
| .Test(xnn_f16_vdiv_minmax_ukernel__neonfp16arith_x16, VBinaryMicrokernelTester::OpType::Div); |
| } |
| } |
| |
| TEST(F16_VDIV_MINMAX__NEONFP16ARITH_X16, qmin) { |
| TEST_REQUIRES_ARM_NEON_FP16_ARITH; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VBinaryMicrokernelTester() |
| .batch_size(batch_size) |
| .qmin(128) |
| .Test(xnn_f16_vdiv_minmax_ukernel__neonfp16arith_x16, VBinaryMicrokernelTester::OpType::Div); |
| } |
| } |
| |
| TEST(F16_VDIV_MINMAX__NEONFP16ARITH_X16, qmax) { |
| TEST_REQUIRES_ARM_NEON_FP16_ARITH; |
| for (size_t batch_size = 1; batch_size <= 80; batch_size += 15) { |
| VBinaryMicrokernelTester() |
| .batch_size(batch_size) |
| .qmax(128) |
| .Test(xnn_f16_vdiv_minmax_ukernel__neonfp16arith_x16, VBinaryMicrokernelTester::OpType::Div); |
| } |
| } |
| #endif // XNN_ARCH_ARM64 |