/external/FP16/test/ |
D | ieee-to-fp32x2-psimd.cc | 22 const psimd_u32x2 fp32 = in TEST() local 25 EXPECT_EQ(positive_infinity_f32, fp32.lo[0]) << in TEST() 28 "F32(F16) = 0x" << std::setw(8) << fp32.lo[0] << ", " << in TEST() 30 EXPECT_EQ(negative_infinity_f32, fp32.lo[1]) << in TEST() 33 "F32(F16) = 0x" << std::setw(8) << fp32.lo[1] << ", " << in TEST() 35 EXPECT_EQ(negative_infinity_f32, fp32.lo[2]) << in TEST() 38 "F32(F16) = 0x" << std::setw(8) << fp32.lo[2] << ", " << in TEST() 40 EXPECT_EQ(positive_infinity_f32, fp32.lo[3]) << in TEST() 43 "F32(F16) = 0x" << std::setw(8) << fp32.lo[3] << ", " << in TEST() 45 EXPECT_EQ(positive_infinity_f32, fp32.hi[0]) << in TEST() [all …]
|
D | ieee-to-fp32-psimd.cc | 20 const psimd_u32 fp32 = (psimd_u32) fp16_ieee_to_fp32_psimd(fp16); in TEST() local 22 EXPECT_EQ(positive_infinity_f32, fp32[0]) << in TEST() 25 "F32(F16) = 0x" << std::setw(8) << fp32[0] << ", " << in TEST() 27 EXPECT_EQ(negative_infinity_f32, fp32[1]) << in TEST() 30 "F32(F16) = 0x" << std::setw(8) << fp32[1] << ", " << in TEST() 32 EXPECT_EQ(negative_infinity_f32, fp32[2]) << in TEST() 35 "F32(F16) = 0x" << std::setw(8) << fp32[2] << ", " << in TEST() 37 EXPECT_EQ(positive_infinity_f32, fp32[3]) << in TEST() 40 "F32(F16) = 0x" << std::setw(8) << fp32[3] << ", " << in TEST() 52 const psimd_u32 fp32 = (psimd_u32) fp16_ieee_to_fp32_psimd(fp16); in TEST() local [all …]
|
D | alt-to-fp32x2-psimd.cc | 23 const psimd_u32x2 fp32 = in TEST() local 26 EXPECT_EQ(fp16_alt_to_fp32_bits(fp16[0]), fp32.lo[0]) << in TEST() 29 "F32(F16) = 0x" << std::setw(8) << fp32.lo[0] << ", " << in TEST() 31 EXPECT_EQ(fp16_alt_to_fp32_bits(fp16[1]), fp32.lo[1]) << in TEST() 34 "F32(F16) = 0x" << std::setw(8) << fp32.lo[1] << ", " << in TEST() 36 EXPECT_EQ(fp16_alt_to_fp32_bits(fp16[2]), fp32.lo[2]) << in TEST() 39 "F32(F16) = 0x" << std::setw(8) << fp32.lo[2] << ", " << in TEST() 41 EXPECT_EQ(fp16_alt_to_fp32_bits(fp16[3]), fp32.lo[3]) << in TEST() 44 "F32(F16) = 0x" << std::setw(8) << fp32.lo[3] << ", " << in TEST() 46 EXPECT_EQ(fp16_alt_to_fp32_bits(fp16[4]), fp32.hi[0]) << in TEST() [all …]
|
D | alt-to-fp32-psimd.cc | 19 const psimd_u32 fp32 = (psimd_u32) fp16_alt_to_fp32_psimd(fp16); in TEST() local 21 EXPECT_EQ(fp16_alt_to_fp32_bits(fp16[0]), fp32[0]) << in TEST() 24 "F32(F16) = 0x" << std::setw(8) << fp32[0] << ", " << in TEST() 26 EXPECT_EQ(fp16_alt_to_fp32_bits(fp16[1]), fp32[1]) << in TEST() 29 "F32(F16) = 0x" << std::setw(8) << fp32[1] << ", " << in TEST() 31 EXPECT_EQ(fp16_alt_to_fp32_bits(fp16[2]), fp32[2]) << in TEST() 34 "F32(F16) = 0x" << std::setw(8) << fp32[2] << ", " << in TEST() 36 EXPECT_EQ(fp16_alt_to_fp32_bits(fp16[3]), fp32[3]) << in TEST() 39 "F32(F16) = 0x" << std::setw(8) << fp32[3] << ", " << in TEST() 55 const psimd_u32 fp32 = (psimd_u32) fp16_alt_to_fp32_psimd(fp16); in TEST() local [all …]
|
D | alt-to-fp32-bits.cc | 218 const uint32_t fp32 = fp16::normalizedValues[h] + ((uint32_t) e << 23); in TEST() local 219 EXPECT_EQ(fp32, fp16_alt_to_fp32_bits(fp16)) << in TEST() 223 "F32 = 0x" << std::setw(8) << fp32; in TEST() 233 const uint32_t fp32 = (fp16::normalizedValues[h] + ((uint32_t) e << 23)) ^ UINT32_C(0x80000000); in TEST() local 234 EXPECT_EQ(fp32, fp16_alt_to_fp32_bits(fp16)) << in TEST() 238 "F32 = 0x" << std::setw(8) << fp32; in TEST() 256 const uint32_t fp32 = fp16::denormalizedValues[h] ^ UINT32_C(0x80000000); in TEST() local 257 EXPECT_EQ(fp32, fp16_alt_to_fp32_bits(fp16)) << in TEST() 261 "F32 = 0x" << std::setw(8) << fp32; in TEST()
|
D | ieee-to-fp32-bits.cc | 288 const uint32_t fp32 = fp16::normalizedValues[h] + ((uint32_t) e << 23); in TEST() local 289 EXPECT_EQ(fp32, fp16_ieee_to_fp32_bits(fp16)) << in TEST() 293 "F32 = 0x" << std::setw(8) << fp32; in TEST() 303 const uint32_t fp32 = (fp16::normalizedValues[h] + ((uint32_t) e << 23)) ^ UINT32_C(0x80000000); in TEST() local 304 EXPECT_EQ(fp32, fp16_ieee_to_fp32_bits(fp16)) << in TEST() 308 "F32 = 0x" << std::setw(8) << fp32; in TEST() 326 const uint32_t fp32 = fp16::denormalizedValues[h] ^ UINT32_C(0x80000000); in TEST() local 327 EXPECT_EQ(fp32, fp16_ieee_to_fp32_bits(fp16)) << in TEST() 331 "F32 = 0x" << std::setw(8) << fp32; in TEST()
|
/external/FP16/bench/ |
D | ieee-element.cc | 48 const uint32_t fp32 = fp16_ieee_to_fp32_bits(fp16); in fp16_ieee_to_fp32_bits() local 51 benchmark::DoNotOptimize(fp32); in fp16_ieee_to_fp32_bits() 59 const float fp32 = fp16_ieee_to_fp32_value(fp16); in fp16_ieee_to_fp32_value() local 62 benchmark::DoNotOptimize(fp32); in fp16_ieee_to_fp32_value() 71 const psimd_f32 fp32 = fp16_ieee_to_fp32_psimd(fp16); in fp16_ieee_to_fp32_psimd() local 74 benchmark::DoNotOptimize(fp32); in fp16_ieee_to_fp32_psimd() 83 const psimd_f32x2 fp32 = fp16_ieee_to_fp32x2_psimd(fp16); in fp16_ieee_to_fp32x2_psimd() local 86 benchmark::DoNotOptimize(fp32); in fp16_ieee_to_fp32x2_psimd() 96 float fp32; in TH_halfbits2float() local 97 TH_halfbits2float(&fp16, &fp32); in TH_halfbits2float() [all …]
|
D | to-ieee-array.cc | 35 std::vector<float> fp32(state.range(0)); in fp16_ieee_from_fp32_value() local 37 std::generate(fp32.begin(), fp32.end(), std::ref(rng)); in fp16_ieee_from_fp32_value() 40 float* input = fp32.data(); in fp16_ieee_from_fp32_value() 60 std::vector<float> fp32(state.range(0)); in hardware_mm_cvtps_ph() local 62 std::generate(fp32.begin(), fp32.end(), std::ref(rng)); in hardware_mm_cvtps_ph() 65 float* input = fp32.data(); in hardware_mm_cvtps_ph() 86 std::vector<float> fp32(state.range(0)); in hardware_mm256_cvtps_ph() local 88 std::generate(fp32.begin(), fp32.end(), std::ref(rng)); in hardware_mm256_cvtps_ph() 91 float* input = fp32.data(); in hardware_mm256_cvtps_ph() 114 std::vector<float> fp32(state.range(0)); in hardware_vcvt_f16_f32() local [all …]
|
D | alt-element.cc | 34 const uint32_t fp32 = fp16_alt_to_fp32_bits(fp16); in fp16_alt_to_fp32_bits() local 37 benchmark::DoNotOptimize(fp32); in fp16_alt_to_fp32_bits() 45 const float fp32 = fp16_alt_to_fp32_value(fp16); in fp16_alt_to_fp32_value() local 48 benchmark::DoNotOptimize(fp32); in fp16_alt_to_fp32_value() 57 const psimd_f32 fp32 = fp16_alt_to_fp32_psimd(fp16); in fp16_alt_to_fp32_psimd() local 60 benchmark::DoNotOptimize(fp32); in fp16_alt_to_fp32_psimd() 69 const psimd_f32x2 fp32 = fp16_alt_to_fp32x2_psimd(fp16); in fp16_alt_to_fp32x2_psimd() local 72 benchmark::DoNotOptimize(fp32); in fp16_alt_to_fp32x2_psimd() 79 uint32_t fp32 = UINT32_C(0x7F800000); in fp16_alt_from_fp32_value() local 81 const uint16_t fp16 = fp16_alt_from_fp32_value(fp32_from_bits(fp32)); in fp16_alt_from_fp32_value() [all …]
|
D | from-ieee-array.cc | 36 std::vector<uint32_t> fp32(state.range(0)); in fp16_ieee_to_fp32_bits() local 44 uint32_t* output = fp32.data(); in fp16_ieee_to_fp32_bits() 61 std::vector<float> fp32(state.range(0)); in fp16_ieee_to_fp32_value() local 69 float* output = fp32.data(); in fp16_ieee_to_fp32_value() 87 std::vector<float> fp32(state.range(0)); in fp16_ieee_to_fp32_psimd() local 95 float* output = fp32.data(); in fp16_ieee_to_fp32_psimd() 117 std::vector<float> fp32(state.range(0)); in fp16_ieee_to_fp32x2_psimd() local 125 float* output = fp32.data(); in fp16_ieee_to_fp32x2_psimd() 148 std::vector<float> fp32(state.range(0)); in hardware_mm_cvtph_ps() local 156 float* output = fp32.data(); in hardware_mm_cvtph_ps() [all …]
|
D | to-alt-array.cc | 27 std::vector<float> fp32(state.range(0)); in fp16_alt_from_fp32_value() local 29 std::generate(fp32.begin(), fp32.end(), std::ref(rng)); in fp16_alt_from_fp32_value() 32 float* input = fp32.data(); in fp16_alt_from_fp32_value() 52 std::vector<float> fp32(state.range(0)); in hardware_vcvt_f16_f32() local 54 std::generate(fp32.begin(), fp32.end(), std::ref(rng)); in hardware_vcvt_f16_f32() 57 float* input = fp32.data(); in hardware_vcvt_f16_f32()
|
D | from-alt-array.cc | 24 std::vector<uint32_t> fp32(state.range(0)); in fp16_alt_to_fp32_bits() local 32 uint32_t* output = fp32.data(); in fp16_alt_to_fp32_bits() 49 std::vector<float> fp32(state.range(0)); in fp16_alt_to_fp32_value() local 57 float* output = fp32.data(); in fp16_alt_to_fp32_value() 75 std::vector<float> fp32(state.range(0)); in fp16_alt_to_fp32_psimd() local 83 float* output = fp32.data(); in fp16_alt_to_fp32_psimd() 105 std::vector<float> fp32(state.range(0)); in fp16_alt_to_fp32x2_psimd() local 113 float* output = fp32.data(); in fp16_alt_to_fp32x2_psimd() 136 std::vector<float> fp32(state.range(0)); in hardware_vcvt_f32_f16() local 144 float* output = fp32.data(); in hardware_vcvt_f32_f16()
|
/external/FP16/test/peachpy/ |
D | alt-xmm-to-fp32-xmm-avx.cc | 8 extern "C" void fp16_alt_xmm_to_fp32_xmm_peachpy__avx(const uint16_t* fp16, uint32_t* fp32); 20 uint32_t fp32[vector_elements]; in TEST() local 21 fp16_alt_xmm_to_fp32_xmm_peachpy__avx(fp16, fp32); in TEST() 24 EXPECT_EQ(fp16_alt_to_fp32_bits(fp16[i]), fp32[i]) << in TEST() 27 "F32(F16) = 0x" << std::setw(8) << fp32[i] << ", " << in TEST() 43 uint32_t fp32[vector_elements]; in TEST() local 44 fp16_alt_xmm_to_fp32_xmm_peachpy__avx(fp16, fp32); in TEST() 47 EXPECT_EQ(fp16_alt_to_fp32_bits(fp16[i]), fp32[i]) << in TEST() 50 "F32(F16) = 0x" << std::setw(8) << fp32[i] << ", " << in TEST() 64 uint32_t fp32[vector_elements]; in TEST() local [all …]
|
D | alt-xmm-to-fp32-ymm-avx2.cc | 8 extern "C" void fp16_alt_xmm_to_fp32_ymm_peachpy__avx2(const uint16_t* fp16, uint32_t* fp32); 20 uint32_t fp32[vector_elements]; in TEST() local 21 fp16_alt_xmm_to_fp32_ymm_peachpy__avx2(fp16, fp32); in TEST() 24 EXPECT_EQ(fp16_alt_to_fp32_bits(fp16[i]), fp32[i]) << in TEST() 27 "F32(F16) = 0x" << std::setw(8) << fp32[i] << ", " << in TEST() 43 uint32_t fp32[vector_elements]; in TEST() local 44 fp16_alt_xmm_to_fp32_ymm_peachpy__avx2(fp16, fp32); in TEST() 47 EXPECT_EQ(fp16_alt_to_fp32_bits(fp16[i]), fp32[i]) << in TEST() 50 "F32(F16) = 0x" << std::setw(8) << fp32[i] << ", " << in TEST() 64 uint32_t fp32[vector_elements]; in TEST() local [all …]
|
/external/FP16/ |
D | CMakeLists.txt | 109 ADD_EXECUTABLE(ieee-to-fp32-bits-test test/ieee-to-fp32-bits.cc test/tables.cc) 110 TARGET_INCLUDE_DIRECTORIES(ieee-to-fp32-bits-test PRIVATE test) 111 TARGET_LINK_LIBRARIES(ieee-to-fp32-bits-test fp16 gtest gtest_main) 112 ADD_TEST(ieee-to-fp32-bits ieee-to-fp32-bits-test) 114 ADD_EXECUTABLE(ieee-to-fp32-value-test test/ieee-to-fp32-value.cc test/tables.cc) 115 TARGET_INCLUDE_DIRECTORIES(ieee-to-fp32-value-test PRIVATE test) 116 TARGET_LINK_LIBRARIES(ieee-to-fp32-value-test fp16 gtest gtest_main) 117 ADD_TEST(ieee-to-fp32-value ieee-to-fp32-value-test) 119 ADD_EXECUTABLE(ieee-from-fp32-value-test test/ieee-from-fp32-value.cc test/tables.cc) 120 TARGET_INCLUDE_DIRECTORIES(ieee-from-fp32-value-test PRIVATE test) [all …]
|
D | Android.bp | 58 "test/alt-from-fp32-value.cc", 69 "test/alt-to-fp32-bits.cc", 80 "test/alt-to-fp32-value.cc", 102 "test/ieee-from-fp32-value.cc", 113 "test/ieee-to-fp32-bits.cc", 124 "test/ieee-to-fp32-value.cc",
|
/external/llvm-project/llvm/test/CodeGen/MIR/AMDGPU/ |
D | machine-function-info.ll | 33 ; CHECK-NEXT: fp32-input-denormals: true 34 ; CHECK-NEXT: fp32-output-denormals: true 66 ; CHECK-NEXT: fp32-input-denormals: true 67 ; CHECK-NEXT: fp32-output-denormals: true 96 ; CHECK-NEXT: fp32-input-denormals: true 97 ; CHECK-NEXT: fp32-output-denormals: true 126 ; CHECK-NEXT: fp32-input-denormals: true 127 ; CHECK-NEXT: fp32-output-denormals: true 140 ; CHECK-NEXT: fp32-input-denormals: true 141 ; CHECK-NEXT: fp32-output-denormals: true [all …]
|
D | machine-function-info-no-ir.mir | 30 # FULL-NEXT: fp32-input-denormals: true 31 # FULL-NEXT: fp32-output-denormals: true 100 # FULL-NEXT: fp32-input-denormals: true 101 # FULL-NEXT: fp32-output-denormals: true 141 # FULL-NEXT: fp32-input-denormals: true 142 # FULL-NEXT: fp32-output-denormals: true 183 # FULL-NEXT: fp32-input-denormals: true 184 # FULL-NEXT: fp32-output-denormals: true 254 # ALL-NEXT: fp32-input-denormals: false 255 # ALL-NEXT: fp32-output-denormals: false [all …]
|
/external/llvm-project/llvm/test/CodeGen/AMDGPU/GlobalISel/ |
D | inst-select-fcanonicalize.mir | 66 fp32-input-denormals: true 67 fp32-output-denormals: true 93 fp32-input-denormals: false 94 fp32-output-denormals: false 228 fp32-input-denormals: true 229 fp32-output-denormals: true 256 fp32-input-denormals: false 257 fp32-output-denormals: false 283 fp32-input-denormals: true 284 fp32-output-denormals: true [all …]
|
/external/tensorflow/tensorflow/core/kernels/ |
D | conv_grad_filter_ops_benchmark_test.cc | 157 using fp32 = float; typedef 160 BENCHMARK_DTYPE(NHWC, 8, fp32, cpu); 161 BENCHMARK_DTYPE(NHWC, 16, fp32, cpu); 162 BENCHMARK_DTYPE(NHWC, 32, fp32, cpu); 171 BENCHMARK_DTYPE(NHWC, 32, fp32, gpu); 172 BENCHMARK_DTYPE(NCHW, 32, fp32, gpu); 177 BENCHMARK_DTYPE(NHWC, 64, fp32, gpu); 178 BENCHMARK_DTYPE(NCHW, 64, fp32, gpu);
|
D | fused_batch_norm_op_test.cc | 199 using fp32 = float; typedef 304 BM_FusedBatchNorm(64, 14, 14, 256, fp32, false, NHWC, cpu); 307 BM_FusedBatchNorm(64, 14, 14, 256, fp32, true, NHWC, cpu); 311 BM_FusedBatchNorm(64, 14, 14, 256, fp32, false, NHWC, gpu); 314 BM_FusedBatchNorm(64, 14, 14, 256, fp32, false, NCHW, gpu); 317 BM_FusedBatchNorm(64, 14, 14, 256, fp32, true, NHWC, gpu); 320 BM_FusedBatchNorm(64, 14, 14, 256, fp32, true, NCHW, gpu); 355 BM_FusedBatchNormGradResnetShapes(fp32, true, NHWC, cpu); 356 BM_FusedBatchNormGradResnetShapes(fp32, false, NHWC, cpu); 359 BM_FusedBatchNormGradResnetShapes(fp32, true, NHWC, gpu); [all …]
|
D | conv_grad_input_ops_benchmark_test.cc | 130 using fp32 = float; typedef 158 BENCHMARK_DTYPE(NHWC, 8, fp32, cpu); 159 BENCHMARK_DTYPE(NHWC, 16, fp32, cpu); 160 BENCHMARK_DTYPE(NHWC, 32, fp32, cpu); 169 BENCHMARK_DTYPE(NHWC, 32, fp32, gpu); 170 BENCHMARK_DTYPE(NCHW, 32, fp32, gpu); 175 BENCHMARK_DTYPE(NHWC, 64, fp32, gpu); 176 BENCHMARK_DTYPE(NCHW, 64, fp32, gpu);
|
/external/FP16/include/fp16/ |
D | bitcasts.h | 33 } fp32 = { w }; in fp32_from_bits() 34 return fp32.as_value; in fp32_from_bits() 51 } fp32 = { f }; in fp32_to_bits() 52 return fp32.as_bits; in fp32_to_bits()
|
/external/llvm/test/CodeGen/AMDGPU/ |
D | hsa-fp-mode.ll | 65 attributes #2 = { nounwind "target-features"="-fp32-denormals,+fp64-denormals" } 66 attributes #3 = { nounwind "target-features"="+fp32-denormals,-fp64-denormals" } 67 attributes #4 = { nounwind "target-features"="+fp32-denormals,+fp64-denormals" } 68 attributes #5 = { nounwind "target-features"="-fp32-denormals,-fp64-denormals" }
|
/external/swiftshader/src/System/ |
D | Half.hpp | 161 static inline unsigned short float32ToFloat11(float fp32) in float32ToFloat11() argument 180 const unsigned int float32Bits = *reinterpret_cast<unsigned int *>(&fp32); in float32ToFloat11() 240 static inline unsigned short float32ToFloat10(float fp32) in float32ToFloat10() argument 259 const unsigned int float32Bits = *reinterpret_cast<unsigned int *>(&fp32); in float32ToFloat10()
|