COMPMID-1607 - (Nightly) CLGEMMLowpMatrixMultiplyCore errors and mismatches
COMPMID-1608 - (Nightly) CLGEMMConvolutionLayer QASYMM8 errors and mismatches
COMPMID-1609 - (Nightly) CLFullyConnectedLayer QASYMM8 mismatches
Change-Id: I84c0d4f468be892f437f9f38b964dc7dfb66663a
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/150869
Reviewed-by: Michele DiGiorgio <michele.digiorgio@arm.com>
Tested-by: bsgcomp <bsgcomp@arm.com>
diff --git a/src/core/CL/cl_kernels/gemmlowp.cl b/src/core/CL/cl_kernels/gemmlowp.cl
index e8124e7..0fc3868 100644
--- a/src/core/CL/cl_kernels/gemmlowp.cl
+++ b/src/core/CL/cl_kernels/gemmlowp.cl
@@ -91,7 +91,7 @@
// src_addr_a = address of matrix A
// src_addr_b = address of matrix B
- __global uchar *src_addr_a = (__global uchar *)(src0_ptr + y * src0_stride_y + src0_offset_first_element_in_bytes);
+ __global uchar *src_addr_a = (__global uchar *)(src0_ptr + z * src0_stride_z + y * src0_stride_y + src0_offset_first_element_in_bytes);
__global uchar *src_addr_b = (__global uchar *)(src1_ptr + x * src1_stride_y + src1_offset_first_element_in_bytes);
#if defined(MATRIX_B_DEPTH)
@@ -251,7 +251,7 @@
// src_addr_a = address of matrix A
// src_addr_b = address of matrix B
- __global uchar *src_addr_a = (__global uchar *)(src0_ptr + y * src0_stride_y + src0_offset_first_element_in_bytes);
+ __global uchar *src_addr_a = (__global uchar *)(src0_ptr + z * src0_stride_z + y * src0_stride_y + src0_offset_first_element_in_bytes);
__global uchar *src_addr_b = (__global uchar *)(src1_ptr + x * src1_stride_y + src1_offset_first_element_in_bytes);
#if defined(MATRIX_B_DEPTH)
@@ -610,7 +610,7 @@
// src_addr_a = address of matrix A
// src_addr_b = address of matrix B
- __global uchar *src_addr_a = (__global uchar *)(src0_ptr + y * src0_stride_y + src0_offset_first_element_in_bytes);
+ __global uchar *src_addr_a = (__global uchar *)(src0_ptr + z * src0_stride_z + y * src0_stride_y + src0_offset_first_element_in_bytes);
__global uchar *src_addr_b = (__global uchar *)(src1_ptr + x * src1_stride_y + src1_offset_first_element_in_bytes);
#if defined(MATRIX_B_DEPTH)
diff --git a/src/core/CL/kernels/CLGEMMLowpMatrixMultiplyKernel.cpp b/src/core/CL/kernels/CLGEMMLowpMatrixMultiplyKernel.cpp
index cf66ebd..ee364e5 100644
--- a/src/core/CL/kernels/CLGEMMLowpMatrixMultiplyKernel.cpp
+++ b/src/core/CL/kernels/CLGEMMLowpMatrixMultiplyKernel.cpp
@@ -325,8 +325,15 @@
ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
+ if(_input1->info()->num_dimensions() < 3)
+ {
+ // The stride_z for matrix B must be zero if we do not slice
+ ARM_COMPUTE_ERROR_ON(_input1->info()->strides_in_bytes()[3] != 0);
+ }
+
Window slice = window.first_slice_window_3D();
Window slice_matrix_b = slice;
+
slice_matrix_b.set(Window::DimX, Window::Dimension(0, 1, 1));
slice_matrix_b.set(Window::DimY, Window::Dimension(0, 1, 1));
@@ -350,8 +357,8 @@
{
Window slice_b = slice;
// Don't slice matrix B along the z dimension if matrix B has just 2 dimensions and matrix A more than 2
- // This scenario can happen when the the matrix multiplication is used to perform a convolution operation
- if(_slide_matrix_b)
+ // This scenario can happen when the matrix multiplication is used to perform a convolution operation
+ if(!_slide_matrix_b)
{
slice_b = slice_matrix_b;
}
diff --git a/src/core/CL/kernels/CLGEMMMatrixMultiplyKernel.cpp b/src/core/CL/kernels/CLGEMMMatrixMultiplyKernel.cpp
index a3cf18a..715edae 100644
--- a/src/core/CL/kernels/CLGEMMMatrixMultiplyKernel.cpp
+++ b/src/core/CL/kernels/CLGEMMMatrixMultiplyKernel.cpp
@@ -384,7 +384,7 @@
if(_reinterpret_input_as_3d)
{
- // Pass bottom paddings to the kernel if the output has to be reinterpreted as 3D tensor
+ // Pass bottom paddings to the kernel if the input has to be reinterpreted as 3D tensor
const unsigned int idx0 = 3 * num_arguments_per_2D_tensor() + 3;
const unsigned int total_cross_plane_pad = _input0->info()->padding().top + _input0->info()->padding().bottom;
_kernel.setArg<cl_uint>(idx0, static_cast<unsigned int>(total_cross_plane_pad));