COMPMID-1054 Update RSH's GEMM to add batch+multi support
Change-Id: Ib9d91b77f1d51976da4449fa1e6eeeffae307353
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/127876
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Pablo Tello <pablo.tello@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
diff --git a/src/core/NEON/kernels/arm_gemm/gemm_uint8.cpp b/src/core/NEON/kernels/arm_gemm/gemm_uint8.cpp
index 9ec479c..0e9f3f2 100644
--- a/src/core/NEON/kernels/arm_gemm/gemm_uint8.cpp
+++ b/src/core/NEON/kernels/arm_gemm/gemm_uint8.cpp
@@ -34,17 +34,18 @@
{
template <>
UniqueGemmCommon<uint8_t, uint32_t> gemm<uint8_t, uint32_t>(const CPUInfo &ci, const unsigned int M, const unsigned int N, const unsigned int K,
+ const unsigned int nbatches, const unsigned int nmulti,
const bool trA, const bool trB, const uint32_t alpha, const uint32_t beta,
const int maxthreads, const bool pretransposed_hint)
{
if(ci.has_dotprod())
{
// Dot product supporting CPUs. This family has a special version for A55r1.
- return UniqueGemmCommon<uint8_t, uint32_t>(new GemmInterleaved<gemm_u8_12x8, uint8_t, uint32_t>(&ci, M, N, K, trA, trB, alpha, beta, maxthreads, pretransposed_hint));
+ return UniqueGemmCommon<uint8_t, uint32_t>(new GemmInterleaved<gemm_u8_12x8, uint8_t, uint32_t>(&ci, M, N, K, nbatches, nmulti, trA, trB, alpha, beta, maxthreads, pretransposed_hint));
}
// Non dot-product code.
- return UniqueGemmCommon<uint8_t, uint32_t>(new GemmInterleaved<gemm_u8_4x4, uint8_t, uint32_t>(&ci, M, N, K, trA, trB, alpha, beta, maxthreads, pretransposed_hint));
+ return UniqueGemmCommon<uint8_t, uint32_t>(new GemmInterleaved<gemm_u8_4x4, uint8_t, uint32_t>(&ci, M, N, K, nbatches, nmulti, trA, trB, alpha, beta, maxthreads, pretransposed_hint));
// TODO: There's a better approach for A53, but it doesn't work
// well on heterogeneous systems as the required data formats