diff options
| author | Adnan AlSinan <adnan.alsinan@arm.com> | 2021-12-10 12:34:02 +0000 |
|---|---|---|
| committer | SiCong Li <sicong.li@arm.com> | 2021-12-23 17:01:14 +0000 |
| commit | 3e155a52f19db046f83e30c25182460b00d108c7 (patch) | |
| tree | f3ec747af21c8ee7b95d6dc88bf6ea4b4aa44659 /src/core/CL/cl_kernels/helpers.h | |
| parent | f76a502a73ca628e2a2556abeaa60ed17bb68d97 (diff) | |
| download | ComputeLibrary-3e155a52f19db046f83e30c25182460b00d108c7.tar.gz | |
Rework gemm_reshape_lhs_ with new macros
Resolves COMPMID-4892
Signed-off-by: Adnan AlSinan <adnan.alsinan@arm.com>
Change-Id: I52f23ca293506fc693ae829daccc6e889a050752
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6833
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: SiCong Li <sicong.li@arm.com>
Reviewed-by: Giorgio Arena <giorgio.arena@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/CL/cl_kernels/helpers.h')
| -rw-r--r-- | src/core/CL/cl_kernels/helpers.h | 8 |
1 files changed, 4 insertions, 4 deletions
diff --git a/src/core/CL/cl_kernels/helpers.h b/src/core/CL/cl_kernels/helpers.h index 88a7665eeb..bfb693e376 100644 --- a/src/core/CL/cl_kernels/helpers.h +++ b/src/core/CL/cl_kernels/helpers.h @@ -392,18 +392,18 @@ #define vload_partial_12(DATA, OFFSET, PTR) \ vload_partial_8(DATA.s01234567, OFFSET, PTR); \ vload_partial_4(DATA.s89AB, OFFSET, PTR + 8); - +// For vload_partial_{13,14,15}, an 8-vector size has been passed, because vectors size of size 5,6,7 are not supported #define vload_partial_13(DATA, OFFSET, PTR) \ vload_partial_8(DATA.s01234567, OFFSET, PTR); \ - vload_partial_5(DATA.s89ABC, OFFSET, PTR + 8); + vload_partial_5(DATA.s89ABCDEF, OFFSET, PTR + 8); #define vload_partial_14(DATA, OFFSET, PTR) \ vload_partial_8(DATA.s01234567, OFFSET, PTR); \ - vload_partial_6(DATA.s89ABCD, OFFSET, PTR + 8); + vload_partial_6(DATA.s89ABCDEF, OFFSET, PTR + 8); #define vload_partial_15(DATA, OFFSET, PTR) \ vload_partial_8(DATA.s01234567, OFFSET, PTR); \ - vload_partial_7(DATA.s89ABCDE, OFFSET, PTR + 8); + vload_partial_7(DATA.s89ABCDEF, OFFSET, PTR + 8); #define vload_partial_16(DATA, OFFSET, PTR) \ DATA = vload16(OFFSET, PTR); |