/external/gemmlowp/internal/ |
D | simd_wrappers.h | 64 static constexpr int kRows = tRows; member 66 static constexpr int kScalarCount = kRows * kCols; 98 (LhsType::kRows < RhsType::kRows)); 132 static constexpr int kRows = 133 Lhs::kRows > Rhs::kRows ? Lhs::kRows : Rhs::kRows; 142 using Type = RegisterBlock<ScalarType, Shape::kRows, Shape::kCols>; 151 static constexpr int Rows = ResultBlockType::kRows; 153 static constexpr int LhsRows = Lhs::kRows; 155 static constexpr int RhsRows = Rhs::kRows; 200 static constexpr int Rows = ResultBlockType::kRows; [all …]
|
D | block_params.h | 65 std::max(1, RoundUp<KernelFormat::kRows>(rows) / num_threads); in FindL2BlockSizes() 87 l2_rows = RoundUp<KernelFormat::kRows>(per_thread_rows); in FindL2BlockSizes() 94 l2_rows = RoundUp<KernelFormat::kRows>( in FindL2BlockSizes() 112 assert(rows % KernelFormat::kRows == 0); in FindL1BlockSizes() 122 1, (l1_bytes_to_use - 4 * KernelFormat::kRows * KernelFormat::kCols) / in FindL1BlockSizes() 123 (KernelFormat::kRows + KernelFormat::kCols)); in FindL1BlockSizes() 136 RoundUp<KernelFormat::kRows>(CeilQuotient(rows, min_l1_rows_blocks)); in FindL1BlockSizes()
|
D | kernel_reference.h | 52 std::int32_t accumulator[Format::kRows * Format::kCols]; in Run() 84 (ci + cc * Format::Rhs::Cell::kWidth) * Format::kRows; in Run() 97 for (int r = 0; r < Format::kRows; r++) { in Run() 100 accumulator[r + c * Format::kRows]; in Run() 106 for (int r = 0; r < Format::kRows; r++) { in Run() 109 accumulator[r + c * Format::kRows]; in Run()
|
D | compute.h | 84 start_row, start_col, Format::kRows, Format::kCols); in ComputeRun() 93 assert(rows % Format::kRows == 0); in ComputeL1() 98 for (int r = 0; r < rows; r += Format::kRows) { in ComputeL1()
|
D | output.h | 44 static constexpr int kRows = InputType::kRows; member 51 using OutputType = RegisterBlock<OutputScalarType, kRows, kCols>;
|
D | kernel.h | 165 static const int kRows = Lhs::Cell::kWidth * Lhs::kCells; member
|
D | multi_thread_gemm.h | 630 const int thread_count = HowManyThreads<KernelFormat::kRows>( in MultiThreadGemm() 670 rows, RoundUp<KernelFormat::kRows>(rows * (n + 1) / task_count)); in MultiThreadGemm()
|
/external/skqp/gm/ |
D | complexclip2.cpp | 39 fTotalHeight = kRows * fHeight + SK_Scalar1 * (kRows + 1) * kPadY; in ComplexClip2GM() 96 for (int i = 0; i < kRows; ++i) { in onOnceBeforeDraw() 99 fOps[j*kRows+i][k] = ops[r.nextU() % SK_ARRAY_COUNT(ops)]; in onOnceBeforeDraw() 105 static constexpr int kRows = 5; member in skiagm::ComplexClip2GM 148 for (int i = 0; i < kRows; ++i) { in onDraw() 176 fOps[j*kRows+i][k], in onDraw() 181 fOps[j*kRows+i][k], in onDraw() 186 fOps[j*kRows+i][k], in onDraw() 203 SkClipOp fOps[kRows * kCols][5];
|
/external/skia/gm/ |
D | complexclip2.cpp | 39 fTotalHeight = kRows * fHeight + SK_Scalar1 * (kRows + 1) * kPadY; in ComplexClip2GM() 96 for (int i = 0; i < kRows; ++i) { in onOnceBeforeDraw() 99 fOps[j*kRows+i][k] = ops[r.nextU() % SK_ARRAY_COUNT(ops)]; in onOnceBeforeDraw() 105 static constexpr int kRows = 5; member in skiagm::ComplexClip2GM 148 for (int i = 0; i < kRows; ++i) { in onDraw() 176 fOps[j*kRows+i][k], in onDraw() 181 fOps[j*kRows+i][k], in onDraw() 186 fOps[j*kRows+i][k], in onDraw() 203 SkClipOp fOps[kRows * kCols][5];
|
/external/tensorflow/tensorflow/core/kernels/ |
D | scatter_nd_op_test.cc | 256 const int kRows = 10000000 / embedding_size; in BM_ScatterNdHelper() local 258 values.reserve(kRows); in BM_ScatterNdHelper() 259 for (int i = 0; i < kRows * embedding_size; i++) { in BM_ScatterNdHelper() 268 indices.push_back(rnd.Uniform(kRows)); in BM_ScatterNdHelper() 276 bm.AddInputFromArray<float>(TensorShape({kRows, embedding_size}), values); in BM_ScatterNdHelper()
|
D | dynamic_partition_op_test.cc | 176 const int kRows = ((128 << 20) / sizeof(T)) / dim; in DynamicPartition() local 177 Tensor data(DataTypeToEnum<T>::value, TensorShape({kRows, dim})); in DynamicPartition() 182 Tensor partitions(DT_INT32, TensorShape({kRows})); in DynamicPartition() 183 for (int i = 0; i < kRows; i++) { in DynamicPartition()
|
D | scatter_op_test.cc | 246 const int kRows = 10000000 / embedding_size; in BM_ScatterHelper() local 248 values.reserve(kRows); in BM_ScatterHelper() 249 for (int i = 0; i < kRows * embedding_size; i++) { in BM_ScatterHelper() 258 indices.push_back(rnd.Uniform(kRows)); in BM_ScatterHelper() 266 bm.AddInputFromArray<float>(TensorShape({kRows, embedding_size}), values); in BM_ScatterHelper()
|
D | gather_op_test.cc | 185 const int kRows = ((512 << 20) / sizeof(float)) / dim; in Gather() local 186 Tensor params(DT_FLOAT, TensorShape({kRows, dim})); in Gather() 194 indices_vec.push_back(rnd.Uniform(kRows)); in Gather()
|
D | cwise_ops_test.cc | 39 const int kRows = 100000; variable 41 int RowsAndColsArg(int r, int c) { return r * kRows + c; } in RowsAndColsArg() 42 int RowsFromArg(int arg) { return (arg / kRows); } in RowsFromArg() 43 int ColsFromArg(int arg) { return (arg % kRows); } in ColsFromArg()
|
/external/gemmlowp/standalone/ |
D | neon-gemm-kernel-benchmark.cc | 239 static const int kRows = Lhs::Cell::kWidth * Lhs::kCells; member 4865 (ci + cc * Format::Rhs::Cell::kWidth) * Format::kRows; in Run()
|