/external/ruy/ruy/ |
D | kernel_avx512.cc | 291 const __m512i lhs_data = _mm512_loadu_si512(lhs_ptr); 313 _mm512_cvtepi8_epi16(_mm512_cvtepi32_epi16(lhs_data)); 316 _mm512_cvtepi32_epi16(_mm512_srli_epi32(lhs_data, 16))); 730 const __m512i lhs_data = _mm512_loadu_si512(lhs_ptr); 745 _mm512_cvtepi8_epi16(_mm512_cvtepi32_epi16(lhs_data)); 748 _mm512_cvtepi32_epi16(_mm512_srli_epi32(lhs_data, 16))); 974 const __m512 lhs_data = _mm512_loadu_ps(lhs_ptr); 992 _mm512_fmadd_ps(lhs_data, dup_rhs_element_j0, accum_data_v0); 995 _mm512_fmadd_ps(lhs_data, dup_rhs_element_j1, accum_data_v1); 998 _mm512_fmadd_ps(lhs_data, dup_rhs_element_j2, accum_data_v2); [all …]
|
D | kernel_x86.h | 609 const __m256 lhs_data = _mm256_loadu_ps(lhs_ptr); 621 lhs_data, dup_rhs_element_0, accum_data_v[0]); 625 lhs_data, dup_rhs_element_1, accum_data_v[1]); 629 lhs_data, dup_rhs_element_2, accum_data_v[2]); 633 lhs_data, dup_rhs_element_3, accum_data_v[3]); 637 lhs_data, dup_rhs_element_4, accum_data_v[4]); 641 lhs_data, dup_rhs_element_5, accum_data_v[5]); 645 lhs_data, dup_rhs_element_6, accum_data_v[6]); 649 lhs_data, dup_rhs_element_7, accum_data_v[7]); 709 const __m256 lhs_data = _mm256_loadu_ps(lhs_ptr); [all …]
|
D | perchannel_buffers_reallocation_test.cc | 55 const LhsScalar lhs_data[kMatrixSize * kMatrixSize] = {0}; in TestPerChannelBuffersReallocation() local 56 lhs.set_data(lhs_data); in TestPerChannelBuffersReallocation()
|
D | prepacked_cache_test.cc | 235 const float lhs_data[] = {1, 2, 3, 4}; in TestCachePolicies() local 241 lhs.set_data(lhs_data); in TestCachePolicies() 276 const float lhs_data[] = {1, 2, 3, 4}; in TEST() local 282 lhs.set_data(lhs_data); in TEST()
|
D | kernel_avx2_fma.cc | 256 const __m256i lhs_data = 279 _mm256_shuffle_epi8(lhs_data, splitter_idx); 812 const __m256i lhs_data = 828 _mm256_shuffle_epi8(lhs_data, splitter_idx);
|
D | kernel_avx.cc | 603 const __m256i lhs_data = 628 intrin_utils::mm256_shuffle_epi8<path>(lhs_data, splitter_idx); 1253 const __m256i lhs_data = 1269 intrin_utils::mm256_shuffle_epi8<path>(lhs_data, splitter_idx);
|
/external/ruy/example/ |
D | example.cc | 22 const float lhs_data[] = {1, 2, 3, 4}; in ExampleMulFloat() local 28 lhs.set_data(lhs_data); in ExampleMulFloat() 46 const float lhs_data[] = {1, 2, 3, 4}; in ExampleMulFloatWithBiasAddAndClamp() local 53 lhs.set_data(lhs_data); in ExampleMulFloatWithBiasAddAndClamp() 74 const std::uint8_t lhs_data[] = {124, 125, 126, 127}; in ExampleMulUint8AsymmetricQuantized() local 80 lhs.set_data(lhs_data); in ExampleMulUint8AsymmetricQuantized() 103 const std::int8_t lhs_data[] = {1, 2, 3, 4}; in ExampleMulInt8PerChannelQuantized() local 111 lhs.set_data(lhs_data); in ExampleMulInt8PerChannelQuantized() 130 const std::int8_t lhs_data[] = {1, 2, 3, 4}; in ExampleMulInt8GetRawAccumulators() local 136 lhs.set_data(lhs_data); in ExampleMulInt8GetRawAccumulators()
|
/external/tensorflow/tensorflow/lite/kernels/ |
D | cpu_backend_gemm_x86.h | 39 const MatrixParams<LhsScalar>& lhs_params, const LhsScalar* lhs_data, in Run() 50 quantization_flavor>::Run(lhs_params, lhs_data, rhs_params, rhs_data, in Run() 57 quantization_flavor>::Run(lhs_params, lhs_data, in Run() 68 static void Run(const MatrixParams<float>& lhs_params, const float* lhs_data, 74 GemmImplUsingEigen::Run(lhs_params, lhs_data, rhs_params, rhs_data,
|
D | cpu_backend_gemm.h | 116 void Gemm(const MatrixParams<LhsScalar>& lhs_params, const LhsScalar* lhs_data, in Gemm() argument 144 quantization_flavor>::Run(lhs_params, lhs_data, in Gemm() 156 if (detail::CustomGemv(lhs_params, lhs_data, rhs_params, rhs_data, in Gemm() 163 quantization_flavor>::Run(lhs_params, lhs_data, rhs_params, rhs_data, in Gemm() 171 void Gemm(const MatrixParams<LhsScalar>& lhs_params, const LhsScalar* lhs_data, in Gemm() argument 183 quantization_flavor>::Run(lhs_params, lhs_data, in Gemm()
|
D | cpu_backend_gemm_test.cc | 254 const std::vector<LhsScalar>& lhs_data, in PerformGemmThenCompareResultsThenAgainWithClamping() argument 262 Gemm(lhs_params, lhs_data.data(), rhs_params, rhs_data.data(), dst_params, in PerformGemmThenCompareResultsThenAgainWithClamping() 275 Gemm(lhs_params, lhs_data.data(), rhs_params, rhs_data.data(), dst_params, in PerformGemmThenCompareResultsThenAgainWithClamping() 284 Gemm(lhs_params, lhs_data.data(), rhs_params, rhs_data.data(), dst_params, in PerformGemmThenCompareResultsThenAgainWithClamping() 312 const std::vector<LhsScalar>& lhs_data, in BisectReasonableMultiplierExponent() argument 331 Gemm(lhs_params, lhs_data.data(), rhs_params, rhs_data.data(), dst_params, in BisectReasonableMultiplierExponent() 336 bisect_mid + 1, bisect_max, lhs_params, lhs_data, rhs_params, rhs_data, in BisectReasonableMultiplierExponent() 340 bisect_min, bisect_mid, lhs_params, lhs_data, rhs_params, rhs_data, in BisectReasonableMultiplierExponent() 348 const MatrixParams<LhsScalar>& lhs_params, const LhsScalar* lhs_data, in ReferenceGemm() argument 356 cpu_backend_gemm::detail::MakeRuyMatrix(lhs_params, lhs_data, &ruy_lhs); in ReferenceGemm() [all …]
|
D | batch_matmul.cc | 302 const TfLiteTensor* lhs_data; in Prepare() local 304 GetInputSafe(context, node, kInputLHSTensor, &lhs_data)); in Prepare() 314 if (lhs_data->type == kTfLiteInt8 || lhs_data->type == kTfLiteInt16) { in Prepare() 317 context, lhs_data, rhs_data, output, &real_multiplier)); in Prepare() 324 if (lhs_data->type == kTfLiteInt8) { in Prepare() 333 if (lhs_data->type == kTfLiteInt16) { in Prepare() 334 TF_LITE_ENSURE_EQ(context, lhs_data->params.zero_point, 0); in Prepare() 339 TF_LITE_ENSURE(context, lhs_data->type == kTfLiteFloat32 || in Prepare() 340 lhs_data->type == kTfLiteInt8 || in Prepare() 341 lhs_data->type == kTfLiteInt16); in Prepare() [all …]
|
D | cpu_backend_gemm_gemmlowp.h | 81 const MatrixParams<SrcScalar>& lhs_params, const SrcScalar* lhs_data, 89 gemmlowp_lhs(lhs_data, lhs_params.rows, lhs_params.cols); 138 const MatrixParams<SrcScalar>& lhs_params, const SrcScalar* lhs_data, 149 gemmlowp_lhs(lhs_data, lhs_params.rows, lhs_params.cols); 183 Run(lhs_params, lhs_data, rhs_params, rhs_data, dst_params, dst_data,
|
D | cpu_backend_gemm_eigen.cc | 38 const MatrixParams<float>& lhs_params, const float* lhs_data, in Run() argument 54 EigenMatrixMapRowMajorConst eigen_lhs(lhs_data, lhs_params.rows, in Run()
|
D | cpu_backend_gemm_custom_gemv.h | 89 const MatrixParams<LhsScalar>& lhs_params, const LhsScalar* lhs_data, in Run() 102 const MatrixParams<LhsScalar>& lhs_params, const LhsScalar* lhs_data, in CustomGemvTask() argument 108 lhs_data_(lhs_data), in CustomGemvTask() 148 const MatrixParams<LhsScalar>& lhs_params, const LhsScalar* lhs_data, in CustomGemv() argument 168 Impl::Run(lhs_params, lhs_data, rhs_params, rhs_data, dst_params, dst_data, in CustomGemv() 180 tasks.emplace_back(lhs_params, lhs_data, rhs_params, rhs_data, dst_params, in CustomGemv() 323 const MatrixParams<LhsScalar>& lhs_params, const LhsScalar* lhs_data, 337 const LhsScalar* filter_ptr = lhs_data + row * lhs_params.cols; 635 static void Run(const MatrixParams<float>& lhs_params, const float* lhs_data, 649 const float* filter_ptr = lhs_data + row * lhs_params.cols;
|
D | cpu_backend_gemm_eigen.h | 29 static void Run(const MatrixParams<float>& lhs_params, const float* lhs_data,
|
D | cpu_backend_gemm_ruy.h | 126 const MatrixParams<LhsScalar>& lhs_params, const LhsScalar* lhs_data, 134 MakeRuyMatrix(lhs_params, lhs_data, &ruy_lhs, context->use_caching());
|
/external/tensorflow/tensorflow/lite/kernels/internal/reference/ |
D | batch_matmul.h | 53 inline void BatchMatMul(const RuntimeShape& lhs_shape, const float* lhs_data, in BatchMatMul() argument 81 const float* lhs_ptr0 = lhs_data + (b0 * lhs_ext0); in BatchMatMul() 108 inline void BatchMatMul(const RuntimeShape& lhs_shape, const int8_t* lhs_data, in BatchMatMul() argument 151 lhs_data, row_sums, num_weights_matrices * lhs_rows, accum_depth); in BatchMatMul() 158 const int8_t* lhs_ptr0 = lhs_data + (b0 * lhs_ext0); in BatchMatMul() 200 const RuntimeShape& lhs_shape, const T* lhs_data, in BatchMatMul() argument 237 const T* lhs_ptr0 = lhs_data + (b0 * lhs_ext0); in BatchMatMul()
|
/external/tensorflow/tensorflow/compiler/xla/service/gpu/ |
D | gemm_thunk.cc | 64 se::DeviceMemoryBase lhs_data = get_device_address(lhs_buffer_); in ExecuteOnStream() local 67 return RunGemm(config_, lhs_data, rhs_data, output_data, params.stream, in ExecuteOnStream() 117 se::DeviceMemory<Element> lhs_data(lhs_matrix.data); in DoGemmWithAlgorithm() local 135 /*alpha=*/static_cast<Element>(alpha), lhs_data, in DoGemmWithAlgorithm() 152 /*alpha=*/alpha, lhs_data, in DoGemmWithAlgorithm() 165 lhs_data, /*leading dim of LHS=*/lhs_matrix.num_rows, rhs_data, in DoGemmWithAlgorithm()
|
/external/tensorflow/tensorflow/lite/kernels/internal/optimized/ |
D | batch_matmul.h | 28 inline void BatchMatMul(const RuntimeShape& lhs_shape, const float* lhs_data, in BatchMatMul() argument 96 const float* lhs_ptr0 = lhs_data + (b0 * lhs_ext0); in BatchMatMul() 115 inline void BatchMatMul(const RuntimeShape& lhs_shape, const int8_t* lhs_data, in BatchMatMul() argument 184 lhs_data, row_sums, num_weights_matrices * lhs_rows, accum_depth); in BatchMatMul() 206 const int8_t* lhs_ptr0 = lhs_data + (b0 * lhs_ext0); in BatchMatMul() 272 const RuntimeShape& lhs_shape, const int8_t* lhs_data, in BatchMatMul() argument 353 const int8_t* lhs_ptr0 = lhs_data + (b0 * lhs_ext0); in BatchMatMul()
|
/external/tensorflow/tensorflow/core/kernels/ |
D | matmul_op_test.cc | 36 std::function<void(const Tensor& lhs_data, const Tensor& rhs_data, 100 void RunMatMulWithBias(const Tensor& lhs_data, const Tensor& rhs_data, in RunMatMulWithBias() argument 108 ops::Const(root.WithOpName("lhs"), Input::Initializer(lhs_data)), in RunMatMulWithBias() 120 const Tensor& lhs_data, const Tensor& rhs_data, const Tensor& bias_data, in RunMatMulWithBiasAndActivation() argument 127 ops::Const(root.WithOpName("lhs"), Input::Initializer(lhs_data)), in RunMatMulWithBiasAndActivation() 150 void RunFusedMatMulOp(const Tensor& lhs_data, const Tensor& rhs_data, in RunFusedMatMulOp() argument 161 ops::Const(root.WithOpName("lhs"), Input::Initializer(lhs_data)); in RunFusedMatMulOp()
|
/external/tensorflow/tensorflow/core/kernels/linalg/ |
D | tridiagonal_solve_op_gpu.cu.cc | 297 const Scalar* lhs_data = lhs_transposed.flat<Scalar>().data(); in ComputeWithGtsvBatched() local 298 const Scalar* superdiag = lhs_data; in ComputeWithGtsvBatched() 299 const Scalar* diag = lhs_data + matrix_size * batch_size; in ComputeWithGtsvBatched() 300 const Scalar* subdiag = lhs_data + 2 * matrix_size * batch_size; in ComputeWithGtsvBatched()
|
/external/llvm-project/lldb/examples/python/ |
D | mach_o.py | 920 lhs_data = lhs_section.get_contents(self) 922 if lhs_data and rhs_data: 923 if lhs_data == rhs_data: 926 lhs_data_len = len(lhs_data) 947 elif lhs_data and not rhs_data: 952 elif not lhs_data and rhs_data:
|
/external/tensorflow/tensorflow/lite/toco/ |
D | tooling_util.cc | 778 const auto& lhs_data = lhs_array.GetBuffer<A>().data; in CompareArrayBuffers() local 780 CHECK_EQ(lhs_data.size(), rhs_data.size()) in CompareArrayBuffers() 782 for (int i = 0; i < lhs_data.size(); ++i) { in CompareArrayBuffers() 783 if (lhs_data[i] != rhs_data[i]) { in CompareArrayBuffers()
|
/external/tensorflow/tensorflow/compiler/xla/tests/ |
D | array_elementwise_ops_test.cc | 263 std::unique_ptr<GlobalData> lhs_data = in XLA_TEST_F() local 287 ComputeAndCompareR1<uint64>(&b, expected, {lhs_data.get(), rhs_data.get()}); in XLA_TEST_F() 303 std::unique_ptr<GlobalData> lhs_data = in XLA_TEST_F() local 326 ComputeAndCompareR1<int64>(&b, expected, {lhs_data.get(), rhs_data.get()}); in XLA_TEST_F()
|
/external/llvm-project/openmp/runtime/src/ |
D | kmp.h | 3827 void *reduce_data, void (*reduce_func)(void *lhs_data, void *rhs_data), 3833 void *reduce_data, void (*reduce_func)(void *lhs_data, void *rhs_data), 3842 void *reduce_data, void (*reduce_func)(void *lhs_data, void *rhs_data),
|