Georgios Pinitas | 421405b | 2018-10-26 19:05:32 +0100 | [diff] [blame] | 1 | /* |
| 2 | * Copyright (c) 2017-2018 ARM Limited. |
| 3 | * |
| 4 | * SPDX-License-Identifier: MIT |
| 5 | * |
| 6 | * Permission is hereby granted, free of charge, to any person obtaining a copy |
| 7 | * of this software and associated documentation files (the "Software"), to |
| 8 | * deal in the Software without restriction, including without limitation the |
| 9 | * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or |
| 10 | * sell copies of the Software, and to permit persons to whom the Software is |
| 11 | * furnished to do so, subject to the following conditions: |
| 12 | * |
| 13 | * The above copyright notice and this permission notice shall be included in all |
| 14 | * copies or substantial portions of the Software. |
| 15 | * |
| 16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
| 17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
| 18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE |
| 19 | * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
| 20 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
| 21 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
| 22 | * SOFTWARE. |
| 23 | */ |
| 24 | #pragma once |
| 25 | |
| 26 | #include "mergeresults.hpp" |
| 27 | #include "transform.hpp" |
| 28 | |
| 29 | namespace arm_gemm { |
| 30 | |
| 31 | /* |
| 32 | * Define "standard" transforms for the blocked GEMMs for SVE. |
| 33 | * |
| 34 | * This assumes that A is interleaved 'height' ways, B is interleaved |
| 35 | * 'width'xVL ways and transposed, and that the merge needs to work in |
| 36 | * 'height' x 'width'xVL blocks. |
| 37 | * |
| 38 | * The optional 'block' parameter is for kernels using dot-product type |
| 39 | * instructions like UDOT and SDOT. |
| 40 | */ |
| 41 | template<typename TOperand, typename TResult, unsigned int height, unsigned int width_vectors, unsigned int block=1, unsigned int mmla=1> |
| 42 | class StdTransformsSVE |
| 43 | { |
| 44 | public: |
| 45 | template<typename TIn> |
| 46 | void PrepareA(TOperand *out, const TIn *in, const int stride, const int y0, |
| 47 | const int ymax, const int k0, const int kmax, bool transposed) { |
| 48 | if (transposed) { |
| 49 | Transform<height, block, true>(out, in, stride, y0, ymax, k0, kmax); |
| 50 | } else { |
| 51 | Transform<height, block, false>(out, in, stride, y0, ymax, k0, kmax); |
| 52 | } |
| 53 | } |
| 54 | |
| 55 | template<typename TIn> |
| 56 | void PrepareB(TOperand *out, const TIn *in, const int stride, const int x0, |
| 57 | const int xmax, const int k0, const int kmax, bool transposed) { |
| 58 | if (transposed) { |
| 59 | Transform<width_vectors, block, false, true>(out, in, stride, x0, xmax, k0, kmax); |
| 60 | } else { |
| 61 | Transform<width_vectors, block, true, true>(out, in, stride, x0, xmax, k0, kmax); |
| 62 | } |
| 63 | } |
| 64 | |
| 65 | template<typename TOut> |
| 66 | void Merge(TOut *out, const TResult *in, int stride, int y0, int ymax, int x0, int xmax, const TOut alpha, const TOut beta) { |
| 67 | MergeResults<width_vectors / mmla, height, true>(out, in, stride, y0, ymax, x0, xmax, alpha, beta); |
| 68 | } |
| 69 | }; |
| 70 | |
| 71 | } // namespace arm_gemm |