From f6f08dac6d57770c191d1bc77123f0ddd2363d3f Mon Sep 17 00:00:00 2001 From: Michele Di Giorgio Date: Thu, 26 Apr 2018 10:24:30 +0100 Subject: COMPMID-1044: Optimizing GCGEMM - Support for not reshaped GEMM on GLES Change-Id: I22fe80393ec70e4501a4f9f9cad14014029d035d Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/129134 Tested-by: Jenkins Reviewed-by: Gian Marco Iodice --- src/core/CL/cl_kernels/gemm.cl | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src/core/CL/cl_kernels/gemm.cl') diff --git a/src/core/CL/cl_kernels/gemm.cl b/src/core/CL/cl_kernels/gemm.cl index 69bc09f023..9ed3af8adc 100644 --- a/src/core/CL/cl_kernels/gemm.cl +++ b/src/core/CL/cl_kernels/gemm.cl @@ -1129,7 +1129,7 @@ __kernel void gemm_mm_interleaved_transposed_qs16(IMAGE_DECLARATION(src0), #if defined(COLS_A) && defined(NUM_ELEMS_PROCESSED_PER_THREAD_X) && (NUM_ELEMS_PROCESSED_PER_THREAD_Y) #if defined(DATA_TYPE) #define VECTOR_TYPE VEC_DATA_TYPE(DATA_TYPE, NUM_ELEMS_PROCESSED_PER_THREAD_X) -/** This OpenCL kernel computes the matrix by matrix multiplication between the matrix A (src0) and matrix B (src1) in case both matrices have not beed reshaped +/** This OpenCL kernel computes the matrix by matrix multiplication between the matrix A (src0) and matrix B (src1) in case both matrices have not been reshaped * * @note This OpenCL kernel works with floating point data types (F16/F32) * @note The floating point data type must be passed at compile time using -DDATA_TYPE (e.g. -DDATA_TYPE=float) @@ -1304,7 +1304,7 @@ __kernel void gemm_mm_floating_point(IMAGE_DECLARATION(src0), } #endif // defined(DATA_TYPE) -/** This OpenCL kernel computes the matrix by matrix multiplication between the matrix A (src0) and matrix B (src1) in case both matrices have not beed reshaped +/** This OpenCL kernel computes the matrix by matrix multiplication between the matrix A (src0) and matrix B (src1) in case both matrices have not been reshaped * * @note This OpenCL kernel works with the 32-bit floating point data type (float) and uses the fma units. * @note The number of elements processed along the x and y directions must be passed at compile time using -DNUM_ELEMS_PROCESSED_PER_THREAD_X and -DNUM_ELEMS_PROCESSED_PER_THREAD_Y. -- cgit v1.2.1