commit 00474e99260da69c5abd14277d0dd0b6de209904
author Gunes Bayir <gunes.bayir@arm.com> Mon Jun 19 21:33:51 2023 +0100
committer Gunes Bayir <gunes.bayir@arm.com> Thu Jun 29 13:23:45 2023 +0000
tree 28238ebbf4721d7aca6fbf6a23658fbe056da055
parent 7a698a38c625047bd558027d4cbc493f063739f5

Implement FP32/16 MatMul Lhs T Rhs T/NT kernel using MMUL extension

Resolves: COMPMID-6196, COMPMID-6197
Change-Id: I22a1c32686eb70e7676c8b4d64a76dbaeb638cb3
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9798
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>

src/gpu/cl/kernels/ClMatMulNativeMMULKernel.cpp

diff --git a/src/gpu/cl/kernels/ClMatMulNativeMMULKernel.cpp b/src/gpu/cl/kernels/ClMatMulNativeMMULKernel.cpp
index 06a0bde..4630ec0 100644
--- a/src/gpu/cl/kernels/ClMatMulNativeMMULKernel.cpp
+++ b/src/gpu/cl/kernels/ClMatMulNativeMMULKernel.cpp
@@ -64,14 +64,17 @@
     const int n0 = matmul_kernel_info.n0;
     const int k0 = matmul_kernel_info.k0;
 
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG((adj_lhs), "adj_lhs is not supported yet");
-
     // Validate M0
     ARM_COMPUTE_RETURN_ERROR_ON_MSG(m0 < 1, "Only positive integers are supported for M0");
 
+    if(adj_lhs)
+    {
+        ARM_COMPUTE_RETURN_ERROR_ON_MSG((m0 != 1) && (m0 != 2) && (m0 != 3) && (m0 != 4) && (m0 != 8) && (m0 != 16), "Only 1,2,3,4,8,16 are supported for M0 for Lhs transposed");
+    }
+
     // Validate N0
     ARM_COMPUTE_RETURN_ERROR_ON_MSG(n0 < 1, "Only positive integers are supported for N0");
-    ARM_COMPUTE_RETURN_ERROR_ON_MSG(((n0 & (n0 - 1)) && (n0 != 3)) || (n0 > 16), "Only 1,2,3,4,8,16 are supported for N0");
+    ARM_COMPUTE_RETURN_ERROR_ON_MSG((n0 != 1) && (n0 != 2) && (n0 != 3) && (n0 != 4) && (n0 != 8) && (n0 != 16), "Only 1,2,3,4,8,16 are supported for N0");
 
     // Validate K0
     ARM_COMPUTE_RETURN_ERROR_ON_MSG((k0 != 1), "Only 1 is supported for k0");
@@ -81,8 +84,7 @@
 
 Status validate_input_shapes(const TensorShape &lhs_shape, const TensorShape &rhs_shape, const MatMulKernelInfo &matmul_kernel_info)
 {
-    ARM_COMPUTE_UNUSED(matmul_kernel_info);
-    const size_t lhs_k = lhs_shape.x();
+    const size_t lhs_k = matmul_kernel_info.adj_lhs ? lhs_shape.y() : lhs_shape.x();
     const size_t rhs_k = matmul_kernel_info.adj_rhs ? rhs_shape.x() : rhs_shape.y();
 
     ARM_COMPUTE_RETURN_ERROR_ON_MSG(lhs_k != rhs_k, "K dimension in Lhs and Rhs matrices must match.");