/dports/math/armadillo/armadillo-10.7.1/include/armadillo_bits/ |
H A D | glue_kron_meat.hpp | 36 const uword B_cols = B.n_cols; in direct_kron() local 38 out.set_size(A_rows*B_rows, A_cols*B_cols); in direct_kron() 46 out.submat(i*B_rows, j*B_cols, (i+1)*B_rows-1, (j+1)*B_cols-1) = A.at(i,j) * B; in direct_kron() 68 const uword B_cols = B.n_cols; in direct_kron() local 70 out.set_size(A_rows*B_rows, A_cols*B_cols); in direct_kron() 80 out.submat(i*B_rows, j*B_cols, (i+1)*B_rows-1, (j+1)*B_cols-1) = A.at(i,j) * tmp_B; in direct_kron() 100 const uword B_cols = B.n_cols; in direct_kron() local 102 out.set_size(A_rows*B_rows, A_cols*B_cols); in direct_kron() 110 out.submat(i*B_rows, j*B_cols, (i+1)*B_rows-1, (j+1)*B_cols-1) = A.at(i,j) * B; in direct_kron()
|
/dports/math/R-cran-RcppArmadillo/RcppArmadillo/inst/include/armadillo_bits/ |
H A D | glue_kron_meat.hpp | 36 const uword B_cols = B.n_cols; in direct_kron() local 38 out.set_size(A_rows*B_rows, A_cols*B_cols); in direct_kron() 46 out.submat(i*B_rows, j*B_cols, (i+1)*B_rows-1, (j+1)*B_cols-1) = A.at(i,j) * B; in direct_kron() 68 const uword B_cols = B.n_cols; in direct_kron() local 70 out.set_size(A_rows*B_rows, A_cols*B_cols); in direct_kron() 80 out.submat(i*B_rows, j*B_cols, (i+1)*B_rows-1, (j+1)*B_cols-1) = A.at(i,j) * tmp_B; in direct_kron() 100 const uword B_cols = B.n_cols; in direct_kron() local 102 out.set_size(A_rows*B_rows, A_cols*B_cols); in direct_kron() 110 out.submat(i*B_rows, j*B_cols, (i+1)*B_rows-1, (j+1)*B_cols-1) = A.at(i,j) * B; in direct_kron()
|
/dports/www/firefox/firefox-99.0/third_party/intgemm/ |
H A D | example.cc | 16 const Index B_cols = 8; in main() local 22 AlignedVector<float> B(width * B_cols); in main() 38 top_left_reference += A[w] * B[w * B_cols]; in main() 51 intgemm::Int16::PrepareB(B.begin(), B_prepared.begin(), quant_mult, width, B_cols); in main() 53 AlignedVector<float> C(A_rows * B_cols); in main() 55 …intgemm::Int16::Multiply(A_prepared.begin(), B_prepared.begin(), A_rows, width, B_cols, intgemm::c… in main() 70 intgemm::Int8::PrepareB(B.begin(), B_prepared.begin(), quant_mult, width, B_cols); in main() 72 AlignedVector<float> C(A_rows * B_cols); in main() 74 …intgemm::Int8::Multiply(A_prepared.begin(), B_prepared.begin(), A_rows, width, B_cols, intgemm::ca… in main()
|
H A D | README.md | 38 * B is width x B_cols. 43 intgemm::Int16::PrepareB(B.begin(), B_prepared.begin(), quant_mult, width, B_cols); 45 intgemm::Int16::Multiply(A_prepared.begin(), B_prepared.begin(), A_rows, width, B_cols, intgemm::ca… 64 * B is width x B_cols. 73 intgemm::Int8Shift::PrepareB(B.begin(), B_prepared.begin(), quant_mult, width, B_cols); 76 intgemm::Int8Shift::PrepareBias(B_prepared.begin(), width, B_cols, callbacks::UnquantizeAndAddBiasA… 78 intgemm::Int8Shift::Multiply(A_prepared.begin(), B_prepared.begin(), A_rows, width, B_cols, callbac…
|
/dports/www/firefox/firefox-99.0/third_party/intgemm/test/ |
H A D | add127_test.cc | 91 AlignedVector<float> B(width * B_cols); in TestMultiplyBiasNew() 92 AlignedVector<float> bias(B_cols); in TestMultiplyBiasNew() 112 Routine::PrepareB(B.begin(), B_prep.begin(), quant_mult, width, B_cols); in TestMultiplyBiasNew() 114 AlignedVector<float> test_C(A_rows * B_cols); in TestMultiplyBiasNew() 154 AlignedVector<float> B(width * B_cols); in TestMultiplyShiftNonShift() 155 AlignedVector<float> bias(B_cols); in TestMultiplyShiftNonShift() 179 AlignedVector<float> test_C(A_rows * B_cols); in TestMultiplyShiftNonShift() 210 AlignedVector<float> B(width * B_cols); in TestMultiplyShiftInt() 211 AlignedVector<float> bias(B_cols); in TestMultiplyShiftInt() 235 AlignedVector<float> test_C(A_rows * B_cols); in TestMultiplyShiftInt() [all …]
|
H A D | multiply_test.cc | 276 AlignedVector<float> B(width * B_cols); in TestMultiply() 292 Routine::PrepareB(B.begin(), B_prep.begin(), quant_mult, width, B_cols); in TestMultiply() 294 AlignedVector<float> test_C(A_rows * B_cols); in TestMultiply() 326 AlignedVector<float> B(width * B_cols); in TestMultiplyRelu() 344 AlignedVector<float> test_C(A_rows * B_cols); in TestMultiplyRelu() 379 AlignedVector<float> B(width * B_cols); in TestMultiplyBias() 380 AlignedVector<float> bias(B_cols); in TestMultiplyBias() 401 AlignedVector<float> test_C(A_rows * B_cols); in TestMultiplyBias() 430 AlignedVector<float> B(width * B_cols); in TestMultiplyBiasRelu() 431 AlignedVector<float> bias(B_cols); in TestMultiplyBiasRelu() [all …]
|
H A D | prepare_b_transposed.cc | 30 bool Test(const AlignedVector<float>& input, Index B_rows, Index B_cols, float quant_mult) { in Test() argument 34 Backend::PrepareBTransposed(input.begin(), output.begin(), quant_mult, B_rows, B_cols); in Test() 37 PrepareBTransposedRef<Backend>(input.begin(), reference.begin(), quant_mult, B_rows, B_cols); in Test() 50 bool TestMany(Index B_rows, Index B_cols, float quant_mult) { in TestMany() argument 51 AlignedVector<float> input(B_rows * B_cols); in TestMany() 59 return Test<Backend>(input, B_rows, B_cols, quant_mult); in TestMany()
|
H A D | prepare_b_quantized_transposed.cc | 29 bool Test(const AlignedVector<typename Backend::Integer>& input, Index B_rows, Index B_cols) { in Test() argument 33 Backend::PrepareBQuantizedTransposed(input.begin(), output.begin(), B_rows, B_cols); in Test() 36 PrepareBQuantizedTransposedRef<Backend>(input.begin(), reference.begin(), B_rows, B_cols); in Test() 49 bool TestMany(Index B_rows, Index B_cols) { in TestMany() argument 50 AlignedVector<typename Backend::Integer> input(B_rows * B_cols); in TestMany() 58 return Test<Backend>(input, B_rows, B_cols); in TestMany()
|
H A D | test.h | 93 void Multiply(const TypeA* A, const TypeB* B, TypeC* C, Index A_rows, Index width, Index B_cols, La… in Multiply() argument 97 for (Index c = 0; c < B_cols; ++c) { in Multiply() 100 sum += IntermediateType(A[r * width + k]) * IntermediateType(B[k * B_cols + c]); in Multiply() 102 C[r * B_cols + c] = callback(sum, {r, c, A_rows, B_cols}); in Multiply()
|
/dports/www/firefox/firefox-99.0/third_party/intgemm/intgemm/ |
H A D | intgemm.h | 207 …static void Multiply(const int8_t *A, const int8_t *B, Index A_rows, Index width, Index B_cols, Ca… in Multiply() 208 MultiplyImpl<Callback>::run(A, B, A_rows, width, B_cols, callback); in Multiply() 216 …static void (*run)(const int8_t *A, const int8_t *B, Index A_rows, Index width, Index B_cols, Call… 255 …static void Multiply(const int8_t *A, const int8_t *B, Index A_rows, Index width, Index B_cols, Ca… in Multiply() 256 MultiplyImpl<Callback>::run((const uint8_t *)A, B, A_rows, width, B_cols, callback); in Multiply() 265 static void PrepareBias(const int8_t *B, Index width, Index B_cols, Callback callback) { in PrepareBias() 266 PrepareBiasImpl<Callback>::run(B, width, B_cols, callback); in PrepareBias() 274 …static void (*run)(const uint8_t *A, const int8_t *B, Index A_rows, Index width, Index B_cols, Cal… 279 static void (*run)(const int8_t *B, Index width, Index B_cols, Callback callback); 335 MultiplyImpl<Callback>::run(A, B, A_rows, width, B_cols, callback); in Multiply() [all …]
|
H A D | avx512vnni_gemm.h | 23 …tiply(const int8_t *A, const int8_t *B, Index A_rows, Index width, Index B_cols, Callback callback… in Multiply() 25 assert(B_cols % 8 == 0); in Multiply() 33 for (Index B0_colidx = 0; B0_colidx < B_cols; B0_colidx += 8) { in Multiply() 78 callback_impl.Run(total, callbacks::OutputBufferInfo(A_rowidx, B0_colidx, A_rows, B_cols)); in Multiply() 86 assert(B_cols % 8 == 0); in Multiply8Shift() 94 for (Index B0_colidx = 0; B0_colidx < B_cols; B0_colidx += 8) { in Multiply8Shift() 119 callback_impl.Run(total, callbacks::OutputBufferInfo(A_rowidx, B0_colidx, A_rows, B_cols)); in Multiply8Shift() 125 …INTGEMM_AVX512VNNI static void PrepareBias(const int8_t *B, Index width, Index B_cols, Callback ca… in PrepareBias() 127 assert(B_cols % 8 == 0); in PrepareBias() 135 for (Index B0_colidx = 0; B0_colidx < B_cols; B0_colidx += 8) { in PrepareBias() [all …]
|
H A D | multiply.h | 164 assert(B_cols % 8 == 0); \ 170 for (Index B0_colidx = 0; B0_colidx < B_cols; B0_colidx += 8) { \ 213 RunCallback(callback_impl, total, A_rowidx, B0_colidx, A_rows, B_cols); \ 222 assert(B_cols % 8 == 0); \ 228 for (Index B0_colidx = 0; B0_colidx < B_cols; B0_colidx += 8) { \ 288 RunCallback(callback_impl, total, 0, B0_colidx, 1, B_cols); \ 296 assert(B_cols % 8 == 0); \ 302 for (Index B0_colidx = 0; B0_colidx < B_cols; B0_colidx += 8) { \ 543 assert(B_cols % 8 == 0); \ 549 for (Index B0_colidx = 0; B0_colidx < B_cols; B0_colidx += 8) { \ [all …]
|
H A D | avx512_gemm.h | 289 …tiply(const int8_t *A, const int8_t *B, Index A_rows, Index width, Index B_cols, Callback callback… in Multiply() 292 assert(B_cols % 8 == 0); in Multiply() 302 for (Index B0_colidx = 0; B0_colidx < B_cols; B0_colidx += 8) { in Multiply() 394 callback_impl.Run(total, callbacks::OutputBufferInfo(A_rowidx, B0_colidx, A_rows, B_cols)); in Multiply()
|
/dports/www/firefox/firefox-99.0/third_party/intgemm/benchmarks/ |
H A D | biasmultiply.cc | 14 std::chrono::duration<double> testNew(Index A_rows, Index width, Index B_cols) { in testNew() argument 16 AlignedVector<float> B(width * B_cols); in testNew() 17 AlignedVector<float> bias(B_cols); in testNew() 37 Routine::PrepareB(B.begin(), B_prep.begin(), quant_mult, width, B_cols); in testNew() 39 AlignedVector<float> test_C(A_rows * B_cols); in testNew() 55 AlignedVector<float> B(width * B_cols); in testOld() 56 AlignedVector<float> bias(B_cols); in testOld() 76 Routine::PrepareB(B.begin(), B_prep.begin(), quant_mult, width, B_cols); in testOld() 78 AlignedVector<float> test_C(A_rows * B_cols); in testOld() 92 AlignedVector<float> B(width * B_cols); in testOld_nobias() [all …]
|
H A D | benchmark.cc | 27 A_rows(A_rows_in), width(width_in), B_cols(B_cols_in), in RandomMatrices() 28 A(A_rows * width), B(width * B_cols) { in RandomMatrices() 41 const Index A_rows, width, B_cols; member 51 AlignedVector<Integer> B_prepared(m.width * m.B_cols); in Run() 52 Backend::PrepareB(m.B.begin(), B_prepared.begin(), quant_mult, m.width, m.B_cols); in Run() 53 AlignedVector<float> output(m.A_rows * m.B_cols); in Run() 55 …Backend::Multiply(A_prepared.begin(), B_prepared.begin(), m.A_rows, m.width, m.B_cols, callbacks::… in Run() 57 …Backend::Multiply(A_prepared.begin(), B_prepared.begin(), m.A_rows, m.width, m.B_cols, callbacks::… in Run() 196 …< matrices[i].A_rows << '\t' << matrices[i].width << '\t' << matrices[i].B_cols << '\t' << "Sample… in main()
|
/dports/misc/py-mxnet/incubator-mxnet-1.9.0/src/operator/contrib/intgemm/ |
H A D | intgemm_fully_connected_op.cc | 184 size_t B_cols = B.shape_.ProdShape(0, B.shape_.ndim() - 1); in IntgemmFullyConnectedOpForwardCPU() local 186 CHECK_EQ(C.shape_.Size(), A_rows * B_cols); in IntgemmFullyConnectedOpForwardCPU() 195 CHECK_EQ(B_cols % ::intgemm::Int8::tile_info.b_cols, 0) << in IntgemmFullyConnectedOpForwardCPU() 244 ::intgemm::Int8::Multiply(A_quant, B_quant, A_rows, inner, B_cols, cb); in IntgemmFullyConnectedOpForwardCPU() 252 ::intgemm::Int8::Multiply(A_quant, B_quant, A_rows, inner, B_cols, cb); in IntgemmFullyConnectedOpForwardCPU() 257 ::intgemm::Int8::Multiply(A_quant, B_quant, A_rows, inner, B_cols, cb); in IntgemmFullyConnectedOpForwardCPU() 261 ::intgemm::Int8::Multiply(A_quant, B_quant, A_rows, inner, B_cols, cb); in IntgemmFullyConnectedOpForwardCPU()
|
H A D | prepare_weight_op.cc | 111 size_t B_cols = in.shape_.ProdShape(0, in.shape_.ndim() - 1); in PrepareWeightOpForwardCPU() local 115 CHECK_EQ(B_cols % ::intgemm::Int8::tile_info.b_cols, 0) << in PrepareWeightOpForwardCPU() 128 ::intgemm::Int8::PrepareBQuantizedTransposed(B, quantB, inner, B_cols); in PrepareWeightOpForwardCPU() 138 B_cols); in PrepareWeightOpForwardCPU()
|
H A D | take_weight_op.cc | 103 size_t B_cols = indices.shape_[0]; in TakeWeightOpForwardCPU() local 107 CHECK_EQ(B_cols % ::intgemm::Int8::tile_info.b_cols, 0) << in TakeWeightOpForwardCPU() 121 index + B_cols); in TakeWeightOpForwardCPU()
|
/dports/misc/mxnet/incubator-mxnet-1.9.0/src/operator/contrib/intgemm/ |
H A D | intgemm_fully_connected_op.cc | 184 size_t B_cols = B.shape_.ProdShape(0, B.shape_.ndim() - 1); in IntgemmFullyConnectedOpForwardCPU() local 186 CHECK_EQ(C.shape_.Size(), A_rows * B_cols); in IntgemmFullyConnectedOpForwardCPU() 195 CHECK_EQ(B_cols % ::intgemm::Int8::tile_info.b_cols, 0) << in IntgemmFullyConnectedOpForwardCPU() 244 ::intgemm::Int8::Multiply(A_quant, B_quant, A_rows, inner, B_cols, cb); in IntgemmFullyConnectedOpForwardCPU() 252 ::intgemm::Int8::Multiply(A_quant, B_quant, A_rows, inner, B_cols, cb); in IntgemmFullyConnectedOpForwardCPU() 257 ::intgemm::Int8::Multiply(A_quant, B_quant, A_rows, inner, B_cols, cb); in IntgemmFullyConnectedOpForwardCPU() 261 ::intgemm::Int8::Multiply(A_quant, B_quant, A_rows, inner, B_cols, cb); in IntgemmFullyConnectedOpForwardCPU()
|
H A D | prepare_weight_op.cc | 111 size_t B_cols = in.shape_.ProdShape(0, in.shape_.ndim() - 1); in PrepareWeightOpForwardCPU() local 115 CHECK_EQ(B_cols % ::intgemm::Int8::tile_info.b_cols, 0) << in PrepareWeightOpForwardCPU() 128 ::intgemm::Int8::PrepareBQuantizedTransposed(B, quantB, inner, B_cols); in PrepareWeightOpForwardCPU() 138 B_cols); in PrepareWeightOpForwardCPU()
|
H A D | take_weight_op.cc | 103 size_t B_cols = indices.shape_[0]; in TakeWeightOpForwardCPU() local 107 CHECK_EQ(B_cols % ::intgemm::Int8::tile_info.b_cols, 0) << in TakeWeightOpForwardCPU() 121 index + B_cols); in TakeWeightOpForwardCPU()
|
/dports/math/py-sympy/sympy-1.9/sympy/matrices/ |
H A D | solvers.py | 561 B_cols = B.cols 562 row, col = aug[:, :-B_cols].shape 566 A, v = A[:, :-B_cols], A[:, -B_cols:] 588 tau = Matrix([next(gen) for k in range((col - rank)*B_cols)]).reshape( 589 col - rank, B_cols) 597 sol = zeros(col, B_cols)
|
/dports/math/stanmath/math-4.2.0/stan/math/opencl/kernels/ |
H A D | rep_matrix.hpp | 26 unsigned int B_rows, unsigned int B_cols, 33 for (int j = gid_j; j < B_cols; j += gsize_j) {
|
/dports/www/qt5-webengine/qtwebengine-everywhere-src-5.15.2/src/3rdparty/chromium/third_party/SPIRV-Tools/src/source/val/ |
H A D | validate_arithmetics.cpp | 489 A_rows, B_rows, C_rows, D_rows, A_cols, B_cols, C_cols, D_cols; in ArithmeticsPass() local 502 B_cols = _.EvalInt32IfConst(B->GetOperandAs<uint32_t>(4)); in ArithmeticsPass() 527 if (notEqual(B_cols, C_cols) || notEqual(B_cols, D_cols) || in ArithmeticsPass()
|
/dports/devel/intel-graphics-compiler/SPIRV-Tools/source/val/ |
H A D | validate_arithmetics.cpp | 489 A_rows, B_rows, C_rows, D_rows, A_cols, B_cols, C_cols, D_cols; in ArithmeticsPass() local 502 B_cols = _.EvalInt32IfConst(B->GetOperandAs<uint32_t>(4)); in ArithmeticsPass() 527 if (notEqual(B_cols, C_cols) || notEqual(B_cols, D_cols) || in ArithmeticsPass()
|