Home
last modified time | relevance | path

Searched +full:k +full:- +full:block (Results 1 – 25 of 1060) sorted by relevance

12345678910>>...43

/external/XNNPACK/test/
Dqs8-gemm-minmax.yaml3 # This source code is licensed under the BSD-style license found in the
5 - name: xnn_qs8_gemm_minmax_ukernel_2x8c8__aarch64_neon_mlal_padal
6 k-block: 16
7 - name: xnn_qs8_gemm_minmax_ukernel_2x8c8__aarch64_neon_mull_padal
8 k-block: 8
9 - name: xnn_qs8_gemm_minmax_ukernel_1x8__neon_mlal_lane
10 k-block: 8
11 - name: xnn_qs8_gemm_minmax_ukernel_2x8__neon_mlal_lane
12 k-block: 8
13 - name: xnn_qs8_gemm_minmax_ukernel_4x8__neon_mlal_lane
[all …]
Df32-gemm-minmax.yaml3 # This source code is licensed under the BSD-style license found in the
5 - name: xnn_f32_gemm_minmax_ukernel_1x8__aarch64_neonfma_ld64
6 k-block: 2
8 - name: xnn_f32_gemm_minmax_ukernel_1x8__aarch64_neonfma_cortex_a53
9 k-block: 8
12 - name: xnn_f32_gemm_minmax_ukernel_1x8__aarch64_neonfma_cortex_a57
13 k-block: 8
16 - name: xnn_f32_gemm_minmax_ukernel_1x8__aarch64_neonfma_cortex_a75
17 k-block: 8
20 - name: xnn_f32_gemm_minmax_ukernel_4x8__aarch64_neonfma_cortex_a53
[all …]
Dqs8-igemm-minmax.yaml3 # This source code is licensed under the BSD-style license found in the
5 - name: xnn_qs8_igemm_minmax_ukernel_2x8c8__aarch64_neon_mlal_padal
6 k-block: 16
7 - name: xnn_qs8_igemm_minmax_ukernel_2x8c16__aarch64_neon_mlal_padal
8 k-block: 16
9 - name: xnn_qs8_igemm_minmax_ukernel_1x8__neon_mlal_lane
10 k-block: 8
11 - name: xnn_qs8_igemm_minmax_ukernel_2x8__neon_mlal_lane
12 k-block: 8
13 - name: xnn_qs8_igemm_minmax_ukernel_4x8__neon_mlal_lane
[all …]
Df32-igemm-minmax.yaml3 # This source code is licensed under the BSD-style license found in the
5 - name: xnn_f32_igemm_minmax_ukernel_1x8__aarch64_neonfma_cortex_a53
6 k-block: 8
9 - name: xnn_f32_igemm_minmax_ukernel_1x8__aarch64_neonfma_cortex_a57
10 k-block: 8
13 - name: xnn_f32_igemm_minmax_ukernel_1x8__aarch64_neonfma_cortex_a75
14 k-block: 8
17 - name: xnn_f32_igemm_minmax_ukernel_4x8__aarch64_neonfma_cortex_a53
18 k-block: 4
21 - name: xnn_f32_igemm_minmax_ukernel_4x8__aarch64_neonfma_cortex_a55
[all …]
Df32-gemminc-minmax.yaml3 # This source code is licensed under the BSD-style license found in the
5 - name: xnn_f32_gemminc_minmax_ukernel_1x8__aarch64_neonfma_cortex_a53
6 k-block: 8
9 - name: xnn_f32_gemminc_minmax_ukernel_1x8__aarch64_neonfma_cortex_a57
10 k-block: 8
13 - name: xnn_f32_gemminc_minmax_ukernel_1x8__aarch64_neonfma_cortex_a75
14 k-block: 8
17 - name: xnn_f32_gemminc_minmax_ukernel_4x8__aarch64_neonfma_cortex_a53
18 k-block: 4
21 - name: xnn_f32_gemminc_minmax_ukernel_4x8__aarch64_neonfma_cortex_a55
[all …]
Df32-spmm-minmax.yaml3 # This source code is licensed under the BSD-style license found in the
5 - name: xnn_f32_spmm_minmax_ukernel_4x1__neon
6 k-block: 1
7 - name: xnn_f32_spmm_minmax_ukernel_4x1__neon_pipelined
8 k-block: 1
9 - name: xnn_f32_spmm_minmax_ukernel_4x1__neon_x2
10 k-block: 2
11 - name: xnn_f32_spmm_minmax_ukernel_8x1__neon
12 k-block: 1
13 - name: xnn_f32_spmm_minmax_ukernel_8x1__neon_pipelined
[all …]
Df16-gemm-minmax.yaml3 # This source code is licensed under the BSD-style license found in the
5 - name: xnn_f16_gemm_minmax_ukernel_1x8__neonfp16arith_ld64
6 k-block: 4
8 - aarch64
9 - name: xnn_f16_gemm_minmax_ukernel_4x8__neonfp16arith_ld64
10 k-block: 4
12 - aarch64
13 - name: xnn_f16_gemm_minmax_ukernel_6x8__neonfp16arith_ld64
14 k-block: 4
16 - aarch64
[all …]
Df32-gemm.yaml3 # This source code is licensed under the BSD-style license found in the
5 - name: xnn_f32_gemm_ukernel_4x4__aarch32_vfp_ld64
6 k-block: 2
8 - name: xnn_f32_gemm_ukernel_1x8__wasmsimd_splat
9 k-block: 4
10 - name: xnn_f32_gemm_ukernel_4x8__wasmsimd_splat
11 k-block: 4
12 - name: xnn_f32_gemm_ukernel_5x8__wasmsimd_splat
13 k-block: 4
14 - name: xnn_f32_gemm_ukernel_4x2c4__wasmsimd
[all …]
Df32-gemm-relu.yaml3 # This source code is licensed under the BSD-style license found in the
5 - name: xnn_f32_gemm_relu_ukernel_1x8__wasmsimd_splat
6 k-block: 4
7 - name: xnn_f32_gemm_relu_ukernel_4x8__wasmsimd_splat
8 k-block: 4
9 - name: xnn_f32_gemm_relu_ukernel_5x8__wasmsimd_splat
10 k-block: 4
11 - name: xnn_f32_gemm_relu_ukernel_4x2c4__wasmsimd
12 k-block: 4
13 - name: xnn_f32_gemm_relu_ukernel_1x4__wasm
[all …]
Df32-igemm.yaml3 # This source code is licensed under the BSD-style license found in the
5 - name: xnn_f32_igemm_ukernel_1x8__wasmsimd_splat
6 k-block: 4
7 - name: xnn_f32_igemm_ukernel_4x8__wasmsimd_splat
8 k-block: 4
9 - name: xnn_f32_igemm_ukernel_5x8__wasmsimd_splat
10 k-block: 4
11 - name: xnn_f32_igemm_ukernel_4x2c4__wasmsimd
12 k-block: 4
13 - name: xnn_f32_igemm_ukernel_1x4__wasm
[all …]
Df32-igemm-relu.yaml3 # This source code is licensed under the BSD-style license found in the
5 - name: xnn_f32_igemm_relu_ukernel_1x8__wasmsimd_splat
6 k-block: 4
7 - name: xnn_f32_igemm_relu_ukernel_4x8__wasmsimd_splat
8 k-block: 4
9 - name: xnn_f32_igemm_relu_ukernel_5x8__wasmsimd_splat
10 k-block: 4
11 - name: xnn_f32_igemm_relu_ukernel_4x2c4__wasmsimd
12 k-block: 4
13 - name: xnn_f32_igemm_relu_ukernel_1x4__wasm
[all …]
Df32-ppmm-minmax.yaml3 # This source code is licensed under the BSD-style license found in the
5 - name: xnn_f32_ppmm_minmax_ukernel_4x8__neon
6 k-block: 1
7 - name: xnn_f32_ppmm_minmax_ukernel_4x8__neonfma
8 k-block: 1
9 - name: xnn_f32_ppmm_minmax_ukernel_8x8__neon
10 k-block: 1
11 - name: xnn_f32_ppmm_minmax_ukernel_8x8__neonfma
12 k-block: 1
13 - name: xnn_f32_ppmm_minmax_ukernel_4x8__sse
[all …]
Df16-igemm-minmax.yaml3 # This source code is licensed under the BSD-style license found in the
5 - name: xnn_f16_igemm_minmax_ukernel_1x8__neonfp16arith_ld64
6 k-block: 4
8 - aarch64
9 - name: xnn_f16_igemm_minmax_ukernel_4x8__neonfp16arith_ld64
10 k-block: 4
12 - aarch64
13 - name: xnn_f16_igemm_minmax_ukernel_6x8__neonfp16arith_ld64
14 k-block: 4
16 - aarch64
[all …]
Df16-spmm-minmax.yaml3 # This source code is licensed under the BSD-style license found in the
5 - name: xnn_f16_spmm_minmax_ukernel_8x1__neonfp16arith
6 k-block: 1
8 - aarch64
9 - name: xnn_f16_spmm_minmax_ukernel_8x1__neonfp16arith_x2
10 k-block: 2
12 - aarch64
13 - name: xnn_f16_spmm_minmax_ukernel_16x1__neonfp16arith
14 k-block: 1
16 - aarch64
[all …]
/external/bzip2/
Dbzip2.14 bzip2, bunzip2 \- a block-sorting file compressor, v1.0.6
6 bzcat \- decompresses files to stdout
8 bzip2recover \- recovers data from damaged bzip2 files
13 .RB [ " \-cdfkqstvzVL123456789 " ]
20 .RB [ " \-fkvsVL " ]
26 .RB [ " \-s " ]
36 compresses files using the Burrows-Wheeler block sorting
39 LZ77/LZ78-based compressors, and approaches the performance of the PPM
42 The command-line options are deliberately very similar to
49 command-line flags. Each file is replaced by a compressed version of
[all …]
Dbzip2.txt3 bzip2, bunzip2 - a block-sorting file compressor, v1.0.6
4 bzcat - decompresses files to stdout
5 bzip2recover - recovers data from damaged bzip2 files
9 bzip2 [ -cdfkqstvzVL123456789 ] [ filenames ... ]
10 bunzip2 [ -fkvsVL ] [ filenames ... ]
11 bzcat [ -s ] [ filenames ... ]
16 bzip2 compresses files using the Burrows-Wheeler block
19 achieved by more conventional LZ77/LZ78-based compressors,
20 and approaches the performance of the PPM family of sta-
23 The command-line options are deliberately very similar to
[all …]
Dblocksort.c2 /*-------------------------------------------------------------*/
3 /*--- Block sorting machinery ---*/
4 /*--- blocksort.c ---*/
5 /*-------------------------------------------------------------*/
7 /* ------------------------------------------------------------------
9 lossless, block-sorting data compression.
12 Copyright (C) 1996-2010 Julian Seward <jseward@bzip.org>
19 ------------------------------------------------------------------ */
24 /*---------------------------------------------*/
25 /*--- Fallback O(N log(N)^2) sorting ---*/
[all …]
/external/mesa3d/src/gallium/drivers/llvmpipe/
Dlp_rast_debug.c67 const struct cmd_block *block, in get_variant() argument
68 int k ) in get_variant() argument
73 if (block->cmd[k] == LP_RAST_OP_SHADE_TILE || in get_variant()
74 block->cmd[k] == LP_RAST_OP_SHADE_TILE_OPAQUE || in get_variant()
75 block->cmd[k] == LP_RAST_OP_TRIANGLE_1 || in get_variant()
76 block->cmd[k] == LP_RAST_OP_TRIANGLE_2 || in get_variant()
77 block->cmd[k] == LP_RAST_OP_TRIANGLE_3 || in get_variant()
78 block->cmd[k] == LP_RAST_OP_TRIANGLE_4 || in get_variant()
79 block->cmd[k] == LP_RAST_OP_TRIANGLE_5 || in get_variant()
80 block->cmd[k] == LP_RAST_OP_TRIANGLE_6 || in get_variant()
[all …]
/external/tensorflow/tensorflow/core/kernels/
Deigen_spatial_convolutions.h7 http://www.apache.org/licenses/LICENSE-2.0
23 #include "tensorflow/core/kernels/eigen_spatial_convolutions-inl.h"
36 EIGEN_ALWAYS_INLINE static Index finalize(Scalar* block,
40 const Index num_coeffs = max_depth - depth;
44 *block = pad ? Scalar(0) : rhs.coeffNoPadding(depth, base_idx);
45 ++block;
55 EIGEN_ALWAYS_INLINE static Index finalize(Scalar* block,
59 Index num_coeffs = max_depth - depth;
66 internal::pstoreu(block, p, mask<Packet>(0, num_coeffs));
72 // Pack a block of the right input matrix (in our case it's always a
[all …]
/external/libjpeg-turbo/simd/x86_64/
Djcphuff-sse2.asm2 ; jcphuff-sse2.asm - prepare data for progressive Huffman encoding
3 ; (64-bit SSE2)
8 ; Copyright (C) 1999-2006, MIYASAKA Masaru.
22 ; --------------------------------------------------------------------------
26 ; --------------------------------------------------------------------------
36 pinsrw X0, word [BLOCK + T0 * 2], 0
37 pinsrw X1, word [BLOCK + T1 * 2], 0
41 pinsrw X0, word [BLOCK + T0 * 2], 1
42 pinsrw X1, word [BLOCK + T1 * 2], 1
46 pinsrw X0, word [BLOCK + T0 * 2], 2
[all …]
/external/tensorflow/tensorflow/core/kernels/linalg/
Dcholesky_grad.cc7 http://www.apache.org/licenses/LICENSE-2.0
64 MatrixMap output_matrix = outputs->at(0); in ComputeMatrix()
76 block_end -= kMaxBlockSize) { in ComputeMatrix()
77 /* This shows the block structure. in ComputeMatrix()
87 const int64 block_begin = std::max(int64{0}, block_end - kMaxBlockSize); in ComputeMatrix()
88 const int64 block_size = block_end - block_begin; in ComputeMatrix()
89 const int64 trailing_size = kMatrixSize - block_end; in ComputeMatrix()
91 auto B = input_matrix_l.block(block_end, 0, trailing_size, block_begin); in ComputeMatrix()
93 output_matrix.block(block_end, 0, trailing_size, block_begin); in ComputeMatrix()
95 auto C = input_matrix_l.block(block_end, block_begin, trailing_size, in ComputeMatrix()
[all …]
/external/libjpeg-turbo/simd/i386/
Djcphuff-sse2.asm2 ; jcphuff-sse2.asm - prepare data for progressive Huffman encoding (SSE2)
7 ; Copyright (C) 1999-2006, MIYASAKA Masaru.
21 ; --------------------------------------------------------------------------
25 ; --------------------------------------------------------------------------
35 pinsrw X0, word [BLOCK + T0 * 2], 0
36 pinsrw X1, word [BLOCK + T1 * 2], 0
40 pinsrw X0, word [BLOCK + T0 * 2], 1
41 pinsrw X1, word [BLOCK + T1 * 2], 1
45 pinsrw X0, word [BLOCK + T0 * 2], 2
46 pinsrw X1, word [BLOCK + T1 * 2], 2
[all …]
/external/apache-commons-math/src/main/java/org/apache/commons/math/linear/
DBlockFieldMatrix.java9 * http://www.apache.org/licenses/LICENSE-2.0
30 * Cache-friendly implementation of FieldMatrix using a flat arrays to store
33 * This implementation is specially designed to be cache-friendly. Square blocks are
35 * and columns major direction, one block at a time. This greatly increases performances
52 * As an example, for a block size of 36x36, a 100x60 matrix would be stored in 6 blocks.
53 * Block 0 would be a Field[1296] array holding the upper left 36x36 square, block 1 would be
54 * a Field[1296] array holding the upper center 36x36 square, block 2 would be a Field[1008]
55 * array holding the upper right 36x28 rectangle, block 3 would be a Field[864] array holding
56 * the lower left 24x36 rectangle, block 4 would be a Field[864] array holding the lower center
57 * 24x36 rectangle and block 5 would be a Field[672] array holding the lower right 24x28
[all …]
DBlockRealMatrix.java9 * http://www.apache.org/licenses/LICENSE-2.0
29 * Cache-friendly implementation of RealMatrix using a flat arrays to store
32 * This implementation is specially designed to be cache-friendly. Square blocks are
34 * and columns major direction, one block at a time. This greatly increases performances
42 * for processors with 64k L1 cache (one block holds 2704 values or 21632 bytes). This value
43 * could be lowered to 36x36 for processors with 32k L1 cache.
53 * As an example, for a block size of 52x52, a 100x60 matrix would be stored in 4 blocks.
54 * Block 0 would be a double[2704] array holding the upper left 52x52 square, block 1 would be
55 * a double[416] array holding the upper right 52x8 rectangle, block 2 would be a double[2496]
56 * array holding the lower left 48x52 rectangle and block 3 would be a double[384] array
[all …]
/external/eigen/Eigen/src/SVD/
DUpperBidiagonalization.h5 // Copyright (C) 2013-2014 Gael Guennebaud <gael.guennebaud@inria.fr>
80 .setLength(m_householder.cols()-1) in householderV()
111 for (Index k = 0; /* breaks at k==cols-1 below */ ; ++k) variable
113 Index remainingRows = rows - k;
114 Index remainingCols = cols - k - 1;
116 // construct left householder transform in-place in A
117 mat.col(k).tail(remainingRows)
118 .makeHouseholderInPlace(mat.coeffRef(k,k), diagonal[k]);
121 .applyHouseholderOnTheLeft(mat.col(k).tail(remainingRows-1), mat.coeff(k,k), tempData);
123 if(k == cols-1) break;
[all …]

12345678910>>...43