Home
last modified time | relevance | path

Searched refs:blas (Results 1 – 25 of 77) sorted by relevance

1234

/external/tensorflow/tensorflow/stream_executor/
Dblas.h66 namespace blas {
492 virtual bool DoBlasGbmv(Stream *stream, blas::Transpose trans, uint64 m,
497 virtual bool DoBlasGbmv(Stream *stream, blas::Transpose trans, uint64 m,
502 virtual bool DoBlasGbmv(Stream *stream, blas::Transpose trans, uint64 m,
509 virtual bool DoBlasGbmv(Stream *stream, blas::Transpose trans, uint64 m,
528 virtual bool DoBlasGemv(Stream *stream, blas::Transpose trans, uint64 m,
532 virtual bool DoBlasGemv(Stream *stream, blas::Transpose trans, uint64 m,
536 virtual bool DoBlasGemv(Stream *stream, blas::Transpose trans, uint64 m,
542 virtual bool DoBlasGemv(Stream *stream, blas::Transpose trans, uint64 m,
550 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()
1434 bool (blas::BlasSupport::*blas_func)(Stream *, Args...), in operator ()()
1442 bool (blas::BlasSupport::*blas_func)(Stream *, Args...),
1448 Stream *stream, bool (blas::BlasSupport::*blas_func)(Stream *, Args...), in Run()
[all …]
Dstream.h918 Stream &ThenBlasGbmv(blas::Transpose trans, uint64 m, uint64 n, uint64 kl,
922 Stream &ThenBlasGbmv(blas::Transpose trans, uint64 m, uint64 n, uint64 kl,
926 Stream &ThenBlasGbmv(blas::Transpose trans, uint64 m, uint64 n, uint64 kl,
932 Stream &ThenBlasGbmv(blas::Transpose trans, uint64 m, uint64 n, uint64 kl,
940 Stream &ThenBlasGemv(blas::Transpose trans, uint64 m, uint64 n, float alpha,
944 Stream &ThenBlasGemv(blas::Transpose trans, uint64 m, uint64 n, double alpha,
948 Stream &ThenBlasGemv(blas::Transpose trans, uint64 m, uint64 n,
954 Stream &ThenBlasGemv(blas::Transpose trans, uint64 m, uint64 n,
961 Stream &ThenBlasGemvWithProfiling(blas::Transpose trans, uint64 m, uint64 n,
966 blas::ProfileResult *output_profile_result);
[all …]
DBUILD61 "blas.h",
115 "blas.h",
193 "blas.h",
279 name = "blas",
280 srcs = ["blas.cc"],
281 hdrs = ["blas.h"],
332 "blas.h",
389 ":blas",
515 ":blas",
550 "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
Dstream_executor_pimpl.h396 bool GetBlasGemmAlgorithms(std::vector<blas::AlgorithmType> *out_algorithms);
402 port::StatusOr<std::unique_ptr<blas::IBlasLtMatmulPlan>>
403 CreateBlasLtMatmulPlan(const blas::BlasLtMatmulPlanParams &params);
409 port::StatusOr<std::vector<std::unique_ptr<blas::IBlasLtMatmulAlgorithm>>>
410 GetBlasLtMatmulAlgorithms(const blas::IBlasLtMatmulPlan *plan,
507 blas::BlasSupport *AsBlas();
684 std::unique_ptr<blas::BlasSupport> blas_ TF_GUARDED_BY(mu_);
/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 {
108 FuncT cublas_func, Stream *stream, blas::Transpose transa,
109 blas::Transpose transb, uint64 m, uint64 n, uint64 k, Scalar alpha,
118 Stream *stream, blas::Transpose transa, blas::Transpose transb, uint64 m,
122 blas::ComputationType computation_type, blas::AlgorithmType algorithm,
123 blas::ProfileResult *output_profile_result);
128 Stream *stream, blas::Transpose transa, blas::Transpose transb, uint64 m,
131 DeviceMemory<T> *c, int ldc, blas::ProfileResult *output_profile_result);
135 bool DoBlasGemvWithProfilingImpl(Stream *stream, blas::Transpose trans,
140 blas::ProfileResult *output_profile_result);
[all …]
/external/tensorflow/tensorflow/stream_executor/rocm/
Drocm_blas.cc377 rocblas_operation ROCMBlasTranspose(blas::Transpose trans) { in ROCMBlasTranspose()
379 case blas::Transpose::kNoTranspose: in ROCMBlasTranspose()
381 case blas::Transpose::kTranspose: in ROCMBlasTranspose()
383 case blas::Transpose::kConjugateTranspose: in ROCMBlasTranspose()
390 rocblas_fill ROCMBlasUpperLower(blas::UpperLower uplo) { in ROCMBlasUpperLower()
392 case blas::UpperLower::kUpper: in ROCMBlasUpperLower()
394 case blas::UpperLower::kLower: in ROCMBlasUpperLower()
401 rocblas_diagonal ROCMBlasDiagonal(blas::Diagonal diag) { in ROCMBlasDiagonal()
403 case blas::Diagonal::kUnit: in ROCMBlasDiagonal()
405 case blas::Diagonal::kNonUnit: in ROCMBlasDiagonal()
[all …]
Drocm_blas.h73 class ROCMBlas : public blas::BlasSupport {
155 FuncT rocblas_func, Stream *stream, blas::Transpose transa,
156 blas::Transpose transb, uint64 m, uint64 n, uint64 k, T alpha,
170 Stream *stream, blas::Transpose transa, blas::Transpose transb, uint64 m,
173 DeviceMemory<OutT> *c, int ldc, blas::ComputationType computation_type,
174 blas::AlgorithmType algorithm,
175 blas::ProfileResult *output_profile_result);
180 Stream *stream, blas::Transpose transa, blas::Transpose transb, uint64 m,
183 DeviceMemory<T> *c, int ldc, blas::ProfileResult *output_profile_result);
187 bool DoBlasGemvWithProfilingImpl(Stream *stream, blas::Transpose trans,
[all …]
/external/tensorflow/tensorflow/compiler/xla/service/gpu/
Dtriangular_solve_thunk.cc41 uplo_(options.lower() ? se::blas::UpperLower::kLower in TriangularSolveThunk()
42 : se::blas::UpperLower::kUpper), in TriangularSolveThunk()
43 side_(options.left_side() ? se::blas::Side::kLeft in TriangularSolveThunk()
44 : se::blas::Side::kRight), in TriangularSolveThunk()
45 unit_diagonal_(options.unit_diagonal() ? se::blas::Diagonal::kUnit in TriangularSolveThunk()
46 : se::blas::Diagonal::kNonUnit), in TriangularSolveThunk()
58 return se::blas::Transpose::kNoTranspose; in TriangularSolveThunk()
60 return se::blas::Transpose::kTranspose; in TriangularSolveThunk()
62 return se::blas::Transpose::kConjugateTranspose; in TriangularSolveThunk()
66 return se::blas::Transpose::kNoTranspose; in TriangularSolveThunk()
[all …]
Dgemm_thunk.cc85 se::Stream *stream, absl::optional<se::blas::AlgorithmType> algorithm, in DoGemmWithAlgorithm()
86 se::blas::ProfileResult *output_profile_result) { in DoGemmWithAlgorithm()
94 se::blas::ComputationType computation_type; in DoGemmWithAlgorithm()
99 computation_type = se::blas::ComputationType::kF32; in DoGemmWithAlgorithm()
102 computation_type = se::blas::ComputationType::kF32; in DoGemmWithAlgorithm()
105 computation_type = se::blas::ComputationType::kF64; in DoGemmWithAlgorithm()
108 computation_type = se::blas::ComputationType::kComplexF32; in DoGemmWithAlgorithm()
111 computation_type = se::blas::ComputationType::kComplexF64; in DoGemmWithAlgorithm()
121 auto lhs_transpose = lhs_matrix.transpose ? se::blas::Transpose::kTranspose in DoGemmWithAlgorithm()
122 : se::blas::Transpose::kNoTranspose; in DoGemmWithAlgorithm()
[all …]
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()
231 static StatusOr<absl::optional<se::blas::AlgorithmType>> DoGemmAutotune( in DoGemmAutotune()
263 absl::optional<se::blas::AlgorithmType> result; in DoGemmAutotune()
289 TF_ASSIGN_OR_RETURN(absl::optional<se::blas::AlgorithmType> gemm_algorithm, in RunOnInstruction()
Dcusolver_context.h58 Status Potrf(se::blas::UpperLower uplo, int n, se::DeviceMemory<T> dev_A,
64 StatusOr<int64> PotrfBufferSize(PrimitiveType type, se::blas::UpperLower uplo,
81 se::blas::UpperLower uplo, int n, se::DeviceMemory<T> A, int lda, \
103 Status Potrf(se::blas::UpperLower uplo, int n, se::DeviceMemory<T> dev_A,
109 StatusOr<int64> PotrfBufferSize(PrimitiveType type, se::blas::UpperLower uplo,
Dcusolver_context.cc44 cublasFillMode_t CUDABlasUpperLower(se::blas::UpperLower uplo) { in CUDABlasUpperLower()
46 case se::blas::UpperLower::kUpper: in CUDABlasUpperLower()
48 case se::blas::UpperLower::kLower: in CUDABlasUpperLower()
140 se::blas::UpperLower uplo, in PotrfBufferSize()
174 se::blas::UpperLower uplo, int n, se::DeviceMemory<T> A, int lda, \
Dcholesky_thunk.cc41 uplo_(options.lower() ? se::blas::UpperLower::kLower in CholeskyThunk()
42 : se::blas::UpperLower::kUpper), in CholeskyThunk()
53 << " uplo=" << se::blas::UpperLowerString(uplo_) in ExecuteOnStream()
Dgemm_thunk.h85 se::blas::ProfileResult* profile_result = nullptr,
86 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/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;
499 se::blas::Transpose trans[] = {se::blas::Transpose::kNoTranspose,
500 se::blas::Transpose::kTranspose,
[all …]

1234