Home
last modified time | relevance | path

Searched refs:kRows (Results 1 – 24 of 24) sorted by relevance

/external/gemmlowp/internal/
Dsimd_wrappers.h64 static constexpr int kRows = tRows; member
66 static constexpr int kScalarCount = kRows * kCols;
98 (LhsType::kRows < RhsType::kRows));
136 static constexpr int kRows =
137 Lhs::kRows > Rhs::kRows ? Lhs::kRows : Rhs::kRows;
146 using Type = RegisterBlock<ScalarType, Shape::kRows, Shape::kCols>;
155 static constexpr int Rows = ResultBlockType::kRows;
157 static constexpr int LhsRows = Lhs::kRows;
159 static constexpr int RhsRows = Rhs::kRows;
204 static constexpr int Rows = ResultBlockType::kRows;
[all …]
Dblock_params.h65 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()
Dkernel_reference.h52 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()
Dcompute.h84 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()
Dkernel.h177 static constexpr int kRows = Lhs::Cell::kWidth * Lhs::kCells; member
Doutput.h45 static constexpr int kRows = InputType::kRows; member
52 using OutputType = RegisterBlock<OutputScalarType, kRows, kCols>;
Dmulti_thread_gemm.h658 const int thread_count = HowManyThreads<KernelFormat::kRows>( in MultiThreadGemm()
698 rows, RoundUp<KernelFormat::kRows>(rows * (n + 1) / task_count)); in MultiThreadGemm()
/external/skqp/gm/
Dcomplexclip2.cpp39 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/
Dcomplexclip2.cpp46 fTotalHeight = kRows * fHeight + SK_Scalar1 * (kRows + 1) * kPadY; in ComplexClip2GM()
99 for (int i = 0; i < kRows; ++i) { in onOnceBeforeDraw()
102 fOps[j*kRows+i][k] = ops[r.nextU() % SK_ARRAY_COUNT(ops)]; in onOnceBeforeDraw()
108 static constexpr int kRows = 5; member in skiagm::ComplexClip2GM
151 for (int i = 0; i < kRows; ++i) { in onDraw()
179 fOps[j*kRows+i][k], in onDraw()
184 fOps[j*kRows+i][k], in onDraw()
189 fOps[j*kRows+i][k], in onDraw()
206 SkClipOp fOps[kRows * kCols][5];
Dlabyrinth.cpp25 constexpr static bool kRows[11][12] = { in draw_labyrinth() local
56 for (size_t y = 0; y < SK_ARRAY_COUNT(kRows); ++y) { in draw_labyrinth()
57 for (size_t x = 0; x < SK_ARRAY_COUNT(kRows[0]); ++x) { in draw_labyrinth()
58 if (kRows[y][x]) { in draw_labyrinth()
/external/angle/src/tests/gl_tests/
DShaderStorageBufferTest.cpp626 constexpr unsigned int kRows = 3; in TEST_P() local
631 MatrixCase matrixCase(kRows, kColumns, kMatrixStride, kComputeShaderSource, kInputDada); in TEST_P()
657 constexpr unsigned int kRows = 3; in TEST_P() local
662 constexpr float kInputDada[kRows * (kMatrixStride / kBytesPerComponent)] = { in TEST_P()
664 MatrixCase matrixCase(kColumns, kRows, kMatrixStride, kComputeShaderSource, kInputDada); in TEST_P()
697 constexpr unsigned int kRows = 3; in TEST_P() local
702 constexpr float kInputDada[kRows * (kMatrixStride / kBytesPerComponent)] = { in TEST_P()
704 MatrixCase matrixCase(kColumns, kRows, kMatrixStride, kComputeShaderSource, kInputDada); in TEST_P()
729 constexpr unsigned int kRows = 3; in TEST_P() local
734 constexpr float kInputDada[kRows * (kMatrixStride / kBytesPerComponent)] = { in TEST_P()
[all …]
/external/tensorflow/tensorflow/core/kernels/
Dscatter_op_test.cc205 const int kRows = 1; in TEST_F() local
206 std::vector<int32> values(kRows, 0); in TEST_F()
210 AddInputFromArray<int32>(TensorShape({kRows}), values); in TEST_F()
285 const int kRows = 10000000 / embedding_size; in BM_ScatterHelper() local
287 values.reserve(kRows); in BM_ScatterHelper()
288 for (int i = 0; i < kRows * embedding_size; i++) { in BM_ScatterHelper()
297 indices.push_back(rnd.Uniform(kRows)); in BM_ScatterHelper()
305 bm.AddInputFromArray<float>(TensorShape({kRows, embedding_size}), values); in BM_ScatterHelper()
Dscatter_nd_op_test.cc259 const int kRows = 10000000 / embedding_size; in BM_ScatterNdHelper() local
261 values.reserve(kRows); in BM_ScatterNdHelper()
262 for (int i = 0; i < kRows * embedding_size; i++) { in BM_ScatterNdHelper()
271 indices.push_back(rnd.Uniform(kRows)); in BM_ScatterNdHelper()
279 bm.AddInputFromArray<float>(TensorShape({kRows, embedding_size}), values); in BM_ScatterNdHelper()
Ddynamic_partition_op_test.cc176 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()
Dgather_op_test.cc200 const int kRows = ((512 << 20) / sizeof(float)) / dim; in Gather() local
201 Tensor params(DT_FLOAT, TensorShape({kRows, dim})); in Gather()
209 indices_vec.push_back(rnd.Uniform(kRows)); in Gather()
Dcwise_ops_test.cc39 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/ruy/ruy/
Dkernel.h100 static_assert(LhsLayout::kRows == RhsLayout::kRows, "");
108 static_assert(is_pot(LhsLayout::kRows), "");
110 static_assert(is_pot(RhsLayout::kRows), "");
Dpack_avx512.cc70 RUY_DCHECK_EQ(Layout::kRows, 4);
76 for (int j = 0; j < (kHalfLayoutCols * Layout::kRows); ++j) {
77 packed_ptr[Layout::kCols * Layout::kRows * k + j] = packed_zero_point;
109 RUY_DCHECK_EQ(Layout::kRows, 4);
113 constexpr int kNumChunkedSrcRows = kNumRowChunks * Layout::kRows;
643 RUY_DCHECK_EQ(kHalfBlockOffset * 2, Layout::kRows * Layout::kCols);
649 RUY_DCHECK_EQ(Layout::kRows, 4);
658 kNumRowChunks * Layout::kCols * Layout::kRows;
686 constexpr int kChunkedRowMask = kNumRowChunks * Layout::kRows - 1;
Dpack_x86.h84 Scalar zerobuf[Layout::kCols * Layout::kRows];
86 Layout::kCols * Layout::kRows * sizeof(Scalar));
131 Scalar zerobuf[Layout::kCols * Layout::kRows];
133 Layout::kCols * Layout::kRows * sizeof(Scalar));
252 Scalar zerobuf[kHalfLayoutCols * Layout::kRows];
254 kHalfLayoutCols * Layout::kRows * sizeof(Scalar));
472 RUY_DCHECK_EQ(PackImpl::Layout::kRows, 1);
Dmat.h467 static constexpr int kRows = tRows; member
475 ret.rows = FixedKernelLayout::kRows; in ToKernelLayout()
487 constexpr int FixedKernelLayout<tOrder, tRows, tCols>::kRows;
Dpack_avx2_fma.cc73 RUY_DCHECK_EQ(Layout::kRows, 4);
77 constexpr int kNumChunkedSrcRows = kNumRowChunks * Layout::kRows;
553 RUY_DCHECK_EQ(Layout::kRows, 4);
562 kNumRowChunks * Layout::kCols * Layout::kRows;
570 constexpr int kChunkedRowMask = kNumRowChunks * Layout::kRows - 1;
Dpack_avx.cc223 RUY_DCHECK_EQ(Layout::kRows, 4);
227 constexpr int kNumChunkedSrcRows = kNumRowChunks * Layout::kRows;
695 RUY_DCHECK_EQ(Layout::kRows, 4);
704 kNumRowChunks * Layout::kCols * Layout::kRows;
712 constexpr int kChunkedRowMask = kNumRowChunks * Layout::kRows - 1;
Dtrace.h566 LhsKernelLayout::kRows, LhsKernelLayout::kCols, \
569 RhsKernelLayout::kRows, RhsKernelLayout::kCols, \
/external/gemmlowp/standalone/
Dneon-gemm-kernel-benchmark.cc239 static const int kRows = Lhs::Cell::kWidth * Lhs::kCells; member
5635 (ci + cc * Format::Rhs::Cell::kWidth) * Format::kRows; in Run()