Home
last modified time | relevance | path

Searched refs:blas (Results 1 – 25 of 78) sorted by relevance

1234

/external/tensorflow/tensorflow/stream_executor/
Dblas.h66 namespace blas {
523 virtual bool DoBlasGbmv(Stream *stream, blas::Transpose trans, uint64 m,
528 virtual bool DoBlasGbmv(Stream *stream, blas::Transpose trans, uint64 m,
533 virtual bool DoBlasGbmv(Stream *stream, blas::Transpose trans, uint64 m,
540 virtual bool DoBlasGbmv(Stream *stream, blas::Transpose trans, uint64 m,
559 virtual bool DoBlasGemv(Stream *stream, blas::Transpose trans, uint64 m,
563 virtual bool DoBlasGemv(Stream *stream, blas::Transpose trans, uint64 m,
567 virtual bool DoBlasGemv(Stream *stream, blas::Transpose trans, uint64 m,
573 virtual bool DoBlasGemv(Stream *stream, blas::Transpose trans, uint64 m,
581 Stream *stream, blas::Transpose trans, uint64 m, uint64 n, float alpha,
[all …]
Dstream.cc73 std::string ToVlogString(blas::Transpose t) { return blas::TransposeString(t); } in ToVlogString()
75 std::string ToVlogString(blas::UpperLower ul) { in ToVlogString()
76 return blas::UpperLowerString(ul); in ToVlogString()
79 std::string ToVlogString(blas::Diagonal d) { return blas::DiagonalString(d); } in ToVlogString()
81 std::string ToVlogString(blas::Side s) { return blas::SideString(s); } in ToVlogString()
83 std::string ToVlogString(blas::ComputationType ty) { in ToVlogString()
84 return blas::ComputationTypeString(ty); in ToVlogString()
1232 bool (blas::BlasSupport::*blas_func)(Stream *, Args...), in operator ()()
1240 bool (blas::BlasSupport::*blas_func)(Stream *, Args...),
1246 Stream *stream, bool (blas::BlasSupport::*blas_func)(Stream *, Args...), in Run()
[all …]
Dstream.h976 Stream &ThenBlasGbmv(blas::Transpose trans, uint64 m, uint64 n, uint64 kl,
980 Stream &ThenBlasGbmv(blas::Transpose trans, uint64 m, uint64 n, uint64 kl,
984 Stream &ThenBlasGbmv(blas::Transpose trans, uint64 m, uint64 n, uint64 kl,
990 Stream &ThenBlasGbmv(blas::Transpose trans, uint64 m, uint64 n, uint64 kl,
998 Stream &ThenBlasGemv(blas::Transpose trans, uint64 m, uint64 n, float alpha,
1002 Stream &ThenBlasGemv(blas::Transpose trans, uint64 m, uint64 n, double alpha,
1006 Stream &ThenBlasGemv(blas::Transpose trans, uint64 m, uint64 n,
1012 Stream &ThenBlasGemv(blas::Transpose trans, uint64 m, uint64 n,
1019 Stream &ThenBlasGemvWithProfiling(blas::Transpose trans, uint64 m, uint64 n,
1024 blas::ProfileResult *output_profile_result);
[all …]
DBUILD60 "blas.h",
114 "blas.h",
192 "blas.h",
278 name = "blas",
279 srcs = ["blas.cc"],
280 hdrs = ["blas.h"],
332 "blas.h",
389 ":blas",
515 ":blas",
551 "blas.h",
[all …]
Dplugin_registry.cc47 blas(kNullPlugin), dnn(kNullPlugin), fft(kNullPlugin), rng(kNullPlugin) { } in DefaultFactories()
128 default_factories_[platform_id].blas = plugin_id; in SetDefaultFactory()
153 return factories.blas.find(plugin_id) != factories.blas.end(); in HasFactory()
246 EMIT_PLUGIN_SPECIALIZATIONS(BlasFactory, blas, "BLAS");
Dplugin.cc33 PluginConfig& PluginConfig::SetBlas(PluginId blas) { in SetBlas() argument
34 blas_ = blas; in SetBlas()
Dplugin.h71 PluginConfig& SetBlas(PluginId blas);
76 PluginId blas() const { return blas_; } in blas() function
Dplugin_registry.h52 typedef blas::BlasSupport* (*BlasFactory)(internal::StreamExecutorInterface*);
107 std::map<PluginId, BlasFactory> blas; member
117 PluginId blas, dnn, fft, rng; member
/external/tensorflow/tensorflow/stream_executor/cuda/
Dcuda_blas.cc289 cublasOperation_t CUDABlasTranspose(blas::Transpose trans) { in CUDABlasTranspose()
291 case blas::Transpose::kNoTranspose: in CUDABlasTranspose()
293 case blas::Transpose::kTranspose: in CUDABlasTranspose()
295 case blas::Transpose::kConjugateTranspose: in CUDABlasTranspose()
302 cublasFillMode_t CUDABlasUpperLower(blas::UpperLower uplo) { in CUDABlasUpperLower()
304 case blas::UpperLower::kUpper: in CUDABlasUpperLower()
306 case blas::UpperLower::kLower: in CUDABlasUpperLower()
313 cublasDiagType_t CUDABlasDiagonal(blas::Diagonal diag) { in CUDABlasDiagonal()
315 case blas::Diagonal::kUnit: in CUDABlasDiagonal()
317 case blas::Diagonal::kNonUnit: in CUDABlasDiagonal()
[all …]
Dcuda_blas.h56 class CUDABlas : public blas::BlasSupport {
107 FuncT cublas_func, Stream *stream, blas::Transpose transa,
108 blas::Transpose transb, uint64 m, uint64 n, uint64 k, Scalar alpha,
117 Stream *stream, blas::Transpose transa, blas::Transpose transb, uint64 m,
120 DeviceMemory<T> *c, int ldc, blas::ProfileResult *output_profile_result);
124 bool DoBlasGemvWithProfilingImpl(Stream *stream, blas::Transpose trans,
129 blas::ProfileResult *output_profile_result);
133 const blas::IBlasLtMatmulPlan *plan,
139 const blas::IBlasLtMatmulAlgorithm *algorithm,
143 port::StatusOr<std::vector<std::unique_ptr<blas::IBlasLtMatmulAlgorithm>>>
[all …]
/external/tensorflow/tensorflow/stream_executor/rocm/
Drocm_blas.cc138 rocblas_operation ROCMBlasTranspose(blas::Transpose trans) { in ROCMBlasTranspose()
140 case blas::Transpose::kNoTranspose: in ROCMBlasTranspose()
142 case blas::Transpose::kTranspose: in ROCMBlasTranspose()
144 case blas::Transpose::kConjugateTranspose: in ROCMBlasTranspose()
151 rocblas_fill ROCMBlasUpperLower(blas::UpperLower uplo) { in ROCMBlasUpperLower()
153 case blas::UpperLower::kUpper: in ROCMBlasUpperLower()
155 case blas::UpperLower::kLower: in ROCMBlasUpperLower()
162 rocblas_diagonal ROCMBlasDiagonal(blas::Diagonal diag) { in ROCMBlasDiagonal()
164 case blas::Diagonal::kUnit: in ROCMBlasDiagonal()
166 case blas::Diagonal::kNonUnit: in ROCMBlasDiagonal()
[all …]
Drocm_blas.h74 class ROCMBlas : public blas::BlasSupport {
166 FuncT rocblas_func, Stream *stream, blas::Transpose transa,
167 blas::Transpose transb, uint64 m, uint64 n, uint64 k, T alpha,
176 Stream *stream, blas::Transpose transa, blas::Transpose transb, uint64 m,
179 DeviceMemory<T> *c, int ldc, blas::ProfileResult *output_profile_result);
183 bool DoBlasGemvWithProfilingImpl(Stream *stream, blas::Transpose trans,
188 blas::ProfileResult *output_profile_result);
/external/tensorflow/tensorflow/compiler/xla/service/gpu/
Dtriangular_solve_thunk.cc40 uplo_(options.lower() ? se::blas::UpperLower::kLower in TriangularSolveThunk()
41 : se::blas::UpperLower::kUpper), in TriangularSolveThunk()
42 side_(options.left_side() ? se::blas::Side::kLeft in TriangularSolveThunk()
43 : se::blas::Side::kRight), in TriangularSolveThunk()
44 unit_diagonal_(options.unit_diagonal() ? se::blas::Diagonal::kUnit in TriangularSolveThunk()
45 : se::blas::Diagonal::kNonUnit), in TriangularSolveThunk()
57 return se::blas::Transpose::kNoTranspose; in TriangularSolveThunk()
59 return se::blas::Transpose::kTranspose; in TriangularSolveThunk()
61 return se::blas::Transpose::kConjugateTranspose; in TriangularSolveThunk()
65 return se::blas::Transpose::kNoTranspose; in TriangularSolveThunk()
[all …]
Dgemm_thunk.cc75 se::blas::Transpose transpose;
81 return transpose == se::blas::Transpose::kTranspose ? num_rows : num_cols; in reduced_dim()
93 static absl::optional<se::blas::ComputationType> ComputationTypeFromPrimitive( in ComputationTypeFromPrimitive()
98 return se::blas::ComputationType::kF32; in ComputationTypeFromPrimitive()
100 return se::blas::ComputationType::kF32; in ComputationTypeFromPrimitive()
102 return se::blas::ComputationType::kF64; in ComputationTypeFromPrimitive()
104 return se::blas::ComputationType::kComplexF32; in ComputationTypeFromPrimitive()
106 return se::blas::ComputationType::kComplexF64; in ComputationTypeFromPrimitive()
108 return se::blas::ComputationType::kI32; in ComputationTypeFromPrimitive()
118 se::Stream *stream, se::blas::AlgorithmType algorithm, in DoGemmWithAlgorithm()
[all …]
Dcusolver_context.cc133 cublasFillMode_t GpuBlasUpperLower(se::blas::UpperLower uplo) { in GpuBlasUpperLower()
135 case se::blas::UpperLower::kUpper: in GpuBlasUpperLower()
137 case se::blas::UpperLower::kLower: in GpuBlasUpperLower()
176 rocblas_fill GpuBlasUpperLower(se::blas::UpperLower uplo) { in GpuBlasUpperLower()
178 case se::blas::UpperLower::kUpper: in GpuBlasUpperLower()
180 case se::blas::UpperLower::kLower: in GpuBlasUpperLower()
284 se::blas::UpperLower uplo, in PotrfBufferSize()
323 se::blas::UpperLower uplo, int n, se::DeviceMemory<T> A, int lda, \
334 se::blas::UpperLower uplo, int n, se::DeviceMemory<T> A, int lda, \
Dtriangular_solve_thunk.h54 const se::blas::UpperLower uplo_;
55 const se::blas::Side side_;
56 const se::blas::Diagonal unit_diagonal_;
57 se::blas::Transpose transpose_a_;
Dgemm_algorithm_picker.cc50 absl::optional<se::blas::AlgorithmType>>();
60 static StatusOr<absl::optional<se::blas::AlgorithmType>> DoUncachedGemmAutotune( in DoUncachedGemmAutotune()
112 std::vector<se::blas::AlgorithmType> algorithms; in DoUncachedGemmAutotune()
115 absl::optional<se::blas::AlgorithmType> first_algorithm; in DoUncachedGemmAutotune()
120 for (se::blas::AlgorithmType algorithm : algorithms) { in DoUncachedGemmAutotune()
128 se::blas::ProfileResult profile_result; in DoUncachedGemmAutotune()
214 static StatusOr<absl::optional<se::blas::AlgorithmType>> DoGemmAutotune( in DoGemmAutotune()
245 TF_ASSIGN_OR_RETURN(absl::optional<se::blas::AlgorithmType> result, in DoGemmAutotune()
264 TF_ASSIGN_OR_RETURN(absl::optional<se::blas::AlgorithmType> gemm_algorithm, in RunOnInstruction()
Dcusolver_context.h60 Status Potrf(se::blas::UpperLower uplo, int n, se::DeviceMemory<T> dev_A,
66 StatusOr<int64> PotrfBufferSize(PrimitiveType type, se::blas::UpperLower uplo,
83 se::blas::UpperLower uplo, int n, se::DeviceMemory<T> A, int lda, \
Dcholesky_thunk.cc45 uplo_(options.lower() ? se::blas::UpperLower::kLower in CholeskyThunk()
46 : se::blas::UpperLower::kUpper), in CholeskyThunk()
57 << " uplo=" << se::blas::UpperLowerString(uplo_) in ExecuteOnStream()
Dgemm_thunk.h82 se::blas::ProfileResult* profile_result = nullptr,
83 absl::optional<se::blas::AlgorithmType> algorithm = absl::nullopt);
Dcusolver_rewriter.cc66 se::blas::UpperLower uplo = options.lower() ? se::blas::UpperLower::kLower in CreateCholesky()
67 : se::blas::UpperLower::kUpper; in CreateCholesky()
/external/tensorflow/tensorflow/compiler/mlir/tfrt/tests/lhlo_to_tfrt/
Dlhlo_gpu_to_tfrt_gpu.mlir30 // CHECK: [[ALGO:%[0-9]+]] = tfrt_gpu.blas.gemm.algo CUBLAS_GEMM_DEFAULT
31 // CHECK: [[HANDLE:%[0-9]+]] = tfrt_gpu.blas.create %arg1
33 // CHECK: [[CHAIN:%[0-9]+]] = tfrt_gpu.blas.gemm [[HANDLE]],
76 // CHECK: [[ALGO:%[0-9]+]] = tfrt_gpu.blas.gemm.algo CUBLAS_GEMM_DEFAULT
77 // CHECK: [[HANDLE:%[0-9]+]] = tfrt_gpu.blas.create %arg1
83 // CHECK: [[CHAIN:%[0-9]+]] = tfrt_gpu.blas.gemm.batch [[HANDLE]],
128 // CHECK: [[ALGO:%[0-9]+]] = tfrt_gpu.blas.gemm.algo CUBLAS_GEMM_DEFAULT
129 // CHECK: [[HANDLE:%[0-9]+]] = tfrt_gpu.blas.create %arg1
131 // CHECK: [[CHAIN:%[0-9]+]] = tfrt_gpu.blas.gemm [[HANDLE]],
203 // CHECK: tfrt_gpu.blas.gemm
[all …]
/external/tensorflow/tensorflow/core/kernels/rnn/
Dblas_gemm.cc45 se::blas::Transpose trans[] = {se::blas::Transpose::kNoTranspose, in operator ()()
46 se::blas::Transpose::kTranspose}; in operator ()()
/external/eigen/blas/testing/
DCMakeLists.txt17 …add_test(${testname} "${Eigen_SOURCE_DIR}/blas/testing/runblastest.sh" "${testname}" "${Eigen_SOUR…
/external/tensorflow/tensorflow/core/kernels/
Dmatmul_op_impl.h327 se::blas::Transpose trans[] = {se::blas::Transpose::kNoTranspose,
328 se::blas::Transpose::kTranspose,
329 se::blas::Transpose::kConjugateTranspose};
413 blas_transpose_b != se::blas::Transpose::kConjugateTranspose &&
414 blas_transpose_a != se::blas::Transpose::kConjugateTranspose) {
420 auto gemv_trans_a = blas_transpose_a == se::blas::Transpose::kTranspose
421 ? se::blas::Transpose::kNoTranspose
422 : se::blas::Transpose::kTranspose;
480 se::blas::Transpose trans[] = {se::blas::Transpose::kNoTranspose,
481 se::blas::Transpose::kTranspose,
[all …]

1234