diff options
| author | Gunes Bayir <gunes.bayir@arm.com> | 2023-03-17 13:52:21 +0000 |
|---|---|---|
| committer | Gunes Bayir <gunes.bayir@arm.com> | 2023-03-20 14:49:51 +0000 |
| commit | 8918b23073851417e8be6e5e53c6380dbdedf201 (patch) | |
| tree | ad0eb38aa7086adb71a444802009a04de3e34929 /src/core/CL/cl_kernels/tile_helpers.h | |
| parent | 14d7b535d48620f009efca576cc70fb6ea9ff20d (diff) | |
| download | ComputeLibrary-8918b23073851417e8be6e5e53c6380dbdedf201.tar.gz | |
Implement OpenCL MatMul for Lhs T Rhs T/NT FP32/16
- Implement opencl kernel for LHS transposed and RHS non-transposed
- Implement opencl kernel for LHS transposed and RHS transposed
- Add validation tests
Resolves: COMPMID-5953, COMPMID-5955
Change-Id: I55589acbffe86c44e29807574975978a1ec09bad
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9345
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/CL/cl_kernels/tile_helpers.h')
| -rw-r--r-- | src/core/CL/cl_kernels/tile_helpers.h | 36 |
1 files changed, 36 insertions, 0 deletions
diff --git a/src/core/CL/cl_kernels/tile_helpers.h b/src/core/CL/cl_kernels/tile_helpers.h index 5d397ad333..872f4c0b57 100644 --- a/src/core/CL/cl_kernels/tile_helpers.h +++ b/src/core/CL/cl_kernels/tile_helpers.h @@ -1297,6 +1297,42 @@ }) \ } +#define T_MMUL_T_NT(LHS_DATA_TYPE, RHS_DATA_TYPE, DST_DATA_TYPE, M0, N0, K0, lhs, rhs, dst) T_MMUL_T_NT_##LHS_DATA_TYPE##_##RHS_DATA_TYPE##_##DST_DATA_TYPE(LHS_DATA_TYPE, RHS_DATA_TYPE, DST_DATA_TYPE, M0, N0, K0, lhs, rhs, dst) +#define T_MMUL_T_NT_float_float_float(LHS_DATA_TYPE, RHS_DATA_TYPE, DST_DATA_TYPE, M0, N0, K0, lhs, rhs, dst) T_MMUL_T_NT_FLOAT(LHS_DATA_TYPE, RHS_DATA_TYPE, DST_DATA_TYPE, M0, N0, K0, lhs, rhs, dst) +#define T_MMUL_T_NT_half_half_float(LHS_DATA_TYPE, RHS_DATA_TYPE, DST_DATA_TYPE, M0, N0, K0, lhs, rhs, dst) T_MMUL_T_NT_FLOAT(LHS_DATA_TYPE, RHS_DATA_TYPE, DST_DATA_TYPE, M0, N0, K0, lhs, rhs, dst) +#define T_MMUL_T_NT_half_half_half(LHS_DATA_TYPE, RHS_DATA_TYPE, DST_DATA_TYPE, M0, N0, K0, lhs, rhs, dst) T_MMUL_T_NT_FLOAT(LHS_DATA_TYPE, RHS_DATA_TYPE, DST_DATA_TYPE, M0, N0, K0, lhs, rhs, dst) +#define T_MMUL_T_NT_FLOAT(LHS_DATA_TYPE, RHS_DATA_TYPE, DST_DATA_TYPE, M0, N0, K0, lhs, rhs, dst) \ + { \ + LOOP_UNROLLING(int, _m, 0, 1, M0, \ + { \ + LOOP_UNROLLING(int, _n, 0, 1, N0, \ + { \ + LOOP_UNROLLING(int, _k, 0, 1, K0, \ + { \ + dst[_m].s[_n] = fma((DST_DATA_TYPE)(lhs[_k].s[_m]), (DST_DATA_TYPE)(rhs[_k].s[_n]), dst[_m].s[_n]); \ + }) \ + }) \ + }) \ + } + +#define T_MMUL_T_T(LHS_DATA_TYPE, RHS_DATA_TYPE, DST_DATA_TYPE, M0, N0, K0, lhs, rhs, dst) T_MMUL_T_T_##LHS_DATA_TYPE##_##RHS_DATA_TYPE##_##DST_DATA_TYPE(LHS_DATA_TYPE, RHS_DATA_TYPE, DST_DATA_TYPE, M0, N0, K0, lhs, rhs, dst) +#define T_MMUL_T_T_float_float_float(LHS_DATA_TYPE, RHS_DATA_TYPE, DST_DATA_TYPE, M0, N0, K0, lhs, rhs, dst) T_MMUL_T_T_FLOAT(LHS_DATA_TYPE, RHS_DATA_TYPE, DST_DATA_TYPE, M0, N0, K0, lhs, rhs, dst) +#define T_MMUL_T_T_half_half_float(LHS_DATA_TYPE, RHS_DATA_TYPE, DST_DATA_TYPE, M0, N0, K0, lhs, rhs, dst) T_MMUL_T_T_FLOAT(LHS_DATA_TYPE, RHS_DATA_TYPE, DST_DATA_TYPE, M0, N0, K0, lhs, rhs, dst) +#define T_MMUL_T_T_half_half_half(LHS_DATA_TYPE, RHS_DATA_TYPE, DST_DATA_TYPE, M0, N0, K0, lhs, rhs, dst) T_MMUL_T_T_FLOAT(LHS_DATA_TYPE, RHS_DATA_TYPE, DST_DATA_TYPE, M0, N0, K0, lhs, rhs, dst) +#define T_MMUL_T_T_FLOAT(LHS_DATA_TYPE, RHS_DATA_TYPE, DST_DATA_TYPE, M0, N0, K0, lhs, rhs, dst) \ + { \ + LOOP_UNROLLING(int, _m, 0, 1, M0, \ + { \ + LOOP_UNROLLING(int, _n, 0, 1, N0, \ + { \ + LOOP_UNROLLING(int, _k, 0, 1, K0, \ + { \ + dst[_m].s[_n] = fma((DST_DATA_TYPE)(lhs[_k].s[_m]), (DST_DATA_TYPE)(rhs[_n].s[_k]), dst[_m].s[_n]); \ + }) \ + }) \ + }) \ + } + #define T_MMUL_NT_T_INTEGER8(LHS_DATA_TYPE, RHS_DATA_TYPE, DST_DATA_TYPE, M0, N0, K0, lhs, rhs, dst) \ ({ \ LOOP_UNROLLING(int, _m, 0, 1, M0, \ |