diff options
| author | David Mansell <David.Mansell@arm.com> | 2018-07-06 14:52:52 +0100 |
|---|---|---|
| committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:54:10 +0000 |
| commit | d93991e290618a685b67506c78090350e6aee43f (patch) | |
| tree | 1d5c3b3017cfccd3f0ec3f24e8e11334cf977ce3 /src/core/NEON/kernels/arm_gemm/merges | |
| parent | dec32a9edd4b3c6dc55c60d7436e79af6be58c3d (diff) | |
| download | ComputeLibrary-d93991e290618a685b67506c78090350e6aee43f.tar.gz | |
COMPMID-1380: Pre-work for SVE support.
This patch makes the needed infrastructure changes to allow SVE
kernels to be added later on.
Change-Id: Ide5bccac2f47278e93fff3d648231aee2d5f8c2e
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/139070
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Tested-by: Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/NEON/kernels/arm_gemm/merges')
4 files changed, 4 insertions, 4 deletions
diff --git a/src/core/NEON/kernels/arm_gemm/merges/a32_merge_float_8x6.hpp b/src/core/NEON/kernels/arm_gemm/merges/a32_merge_float_8x6.hpp index 2b833937a8..f4485bcbb1 100644 --- a/src/core/NEON/kernels/arm_gemm/merges/a32_merge_float_8x6.hpp +++ b/src/core/NEON/kernels/arm_gemm/merges/a32_merge_float_8x6.hpp @@ -28,7 +28,7 @@ #include <arm_neon.h> template<> -inline void MergeResults<8, 6>(float *out, const float *in, const int ldout, const int y0, const int ymax, const int x0, const int xmax, const float alpha, const float beta) { +inline void MergeResults<8, 6, false>(float *out, const float *in, const int ldout, const int y0, const int ymax, const int x0, const int xmax, const float alpha, const float beta) { const float *inptr = in; prefetch_6x(inptr); prefetch_6x(inptr + 96); diff --git a/src/core/NEON/kernels/arm_gemm/merges/a64_merge_float_12x8.hpp b/src/core/NEON/kernels/arm_gemm/merges/a64_merge_float_12x8.hpp index f6befa2d14..be23978b80 100644 --- a/src/core/NEON/kernels/arm_gemm/merges/a64_merge_float_12x8.hpp +++ b/src/core/NEON/kernels/arm_gemm/merges/a64_merge_float_12x8.hpp @@ -26,7 +26,7 @@ #ifdef __aarch64__ template<> -inline void MergeResults<12, 8>(float *out, const float *in, const int ldout, const int y0, const int ymax, const int x0, const int xmax, const float alpha, const float beta) { +inline void MergeResults<12, 8, false>(float *out, const float *in, const int ldout, const int y0, const int ymax, const int x0, const int xmax, const float alpha, const float beta) { const float *inptr = in; prefetch_6x(inptr); prefetch_6x(inptr + 96); diff --git a/src/core/NEON/kernels/arm_gemm/merges/a64_merge_float_to_half_12x8.hpp b/src/core/NEON/kernels/arm_gemm/merges/a64_merge_float_to_half_12x8.hpp index e7a7521823..9e5eb88dc1 100644 --- a/src/core/NEON/kernels/arm_gemm/merges/a64_merge_float_to_half_12x8.hpp +++ b/src/core/NEON/kernels/arm_gemm/merges/a64_merge_float_to_half_12x8.hpp @@ -29,7 +29,7 @@ #include <arm_neon.h> template<> -inline void MergeResults<12,8>(__fp16 *out, const float *in, int ldout, int y0, int ymax, int x0, int xmax, const __fp16 alpha, const __fp16 beta) { +inline void MergeResults<12,8,false>(__fp16 *out, const float *in, int ldout, int y0, int ymax, int x0, int xmax, const __fp16 alpha, const __fp16 beta) { const float *inptr = in; prefetch_6x(inptr); prefetch_6x(inptr + 24); diff --git a/src/core/NEON/kernels/arm_gemm/merges/a64_merge_int32_12x8.hpp b/src/core/NEON/kernels/arm_gemm/merges/a64_merge_int32_12x8.hpp index 1a51505a25..ee32ce7630 100644 --- a/src/core/NEON/kernels/arm_gemm/merges/a64_merge_int32_12x8.hpp +++ b/src/core/NEON/kernels/arm_gemm/merges/a64_merge_int32_12x8.hpp @@ -26,7 +26,7 @@ #ifdef __aarch64__ template<> -inline void MergeResults<12, 8>(int32_t *out, const int32_t *in, const int ldout, const int y0, const int ymax, const int x0, const int xmax, const int32_t alpha, const int32_t beta) { +inline void MergeResults<12, 8, false>(int32_t *out, const int32_t *in, const int ldout, const int y0, const int ymax, const int x0, const int xmax, const int32_t alpha, const int32_t beta) { const int32_t *inptr = in; prefetch_6x(inptr); prefetch_6x(inptr + 96); |