diff options
author | Moritz Pflanzer <moritz.pflanzer@arm.com> | 2017-08-31 14:56:32 +0100 |
---|---|---|
committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:35:24 +0000 |
commit | beabe3bdf47306d0940ddf2ddf52ada6903a0875 (patch) | |
tree | 97afa72f2d60858898ab2dadb95e4cda7176e88b /src/core/NEON/kernels/NEGEMMInterleave4x4Kernel.cpp | |
parent | 7655a67384895868c0afa72bfda9a9b2fcfdf323 (diff) | |
download | ComputeLibrary-beabe3bdf47306d0940ddf2ddf52ada6903a0875.tar.gz |
COMPMID-481: Add AArch64 GEMM
Change-Id: I34f94f99cb05f0eabafee13c5e623ee779b72360
Reviewed-on: http://mpd-gerrit.cambridge.arm.com/83741
Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Reviewed-by: Pablo Tello <pablo.tello@arm.com>
Diffstat (limited to 'src/core/NEON/kernels/NEGEMMInterleave4x4Kernel.cpp')
-rw-r--r-- | src/core/NEON/kernels/NEGEMMInterleave4x4Kernel.cpp | 16 |
1 files changed, 8 insertions, 8 deletions
diff --git a/src/core/NEON/kernels/NEGEMMInterleave4x4Kernel.cpp b/src/core/NEON/kernels/NEGEMMInterleave4x4Kernel.cpp index c76c39aa4b..ae5d456141 100644 --- a/src/core/NEON/kernels/NEGEMMInterleave4x4Kernel.cpp +++ b/src/core/NEON/kernels/NEGEMMInterleave4x4Kernel.cpp @@ -85,10 +85,10 @@ void gemm_interleave_16bit_elements(const ITensor *input, ITensor *output, const const uint16x4x4_t data = { { - vld1_u16(reinterpret_cast<uint16_t *>(in.ptr() + 0 * in_stride)), - vld1_u16(reinterpret_cast<uint16_t *>(in.ptr() + 1 * in_stride)), - vld1_u16(reinterpret_cast<uint16_t *>(in.ptr() + 2 * in_stride)), - vld1_u16(reinterpret_cast<uint16_t *>(in.ptr() + 3 * in_stride)), + vld1_u16(reinterpret_cast<const uint16_t *>(in.ptr() + 0 * in_stride)), + vld1_u16(reinterpret_cast<const uint16_t *>(in.ptr() + 1 * in_stride)), + vld1_u16(reinterpret_cast<const uint16_t *>(in.ptr() + 2 * in_stride)), + vld1_u16(reinterpret_cast<const uint16_t *>(in.ptr() + 3 * in_stride)), } }; vst4_u16(reinterpret_cast<uint16_t *>(out.ptr()), data); @@ -113,10 +113,10 @@ void gemm_interleave_32bit_elements(const ITensor *input, ITensor *output, const const uint32x4x4_t data = { { - vld1q_u32(reinterpret_cast<uint32_t *>(in.ptr() + 0 * in_stride)), - vld1q_u32(reinterpret_cast<uint32_t *>(in.ptr() + 1 * in_stride)), - vld1q_u32(reinterpret_cast<uint32_t *>(in.ptr() + 2 * in_stride)), - vld1q_u32(reinterpret_cast<uint32_t *>(in.ptr() + 3 * in_stride)) + vld1q_u32(reinterpret_cast<const uint32_t *>(in.ptr() + 0 * in_stride)), + vld1q_u32(reinterpret_cast<const uint32_t *>(in.ptr() + 1 * in_stride)), + vld1q_u32(reinterpret_cast<const uint32_t *>(in.ptr() + 2 * in_stride)), + vld1q_u32(reinterpret_cast<const uint32_t *>(in.ptr() + 3 * in_stride)) } }; vst4q_u32(reinterpret_cast<uint32_t *>(out.ptr()), data); |