aboutsummaryrefslogtreecommitdiff
path: root/src/core/CL/cl_kernels/gemmlowp.cl
diff options
context:
space:
mode:
Diffstat (limited to 'src/core/CL/cl_kernels/gemmlowp.cl')
-rw-r--r--src/core/CL/cl_kernels/gemmlowp.cl6
1 files changed, 3 insertions, 3 deletions
diff --git a/src/core/CL/cl_kernels/gemmlowp.cl b/src/core/CL/cl_kernels/gemmlowp.cl
index 033b4b4942..18ccb65aaf 100644
--- a/src/core/CL/cl_kernels/gemmlowp.cl
+++ b/src/core/CL/cl_kernels/gemmlowp.cl
@@ -35,7 +35,7 @@
#if defined(COLS_B) && defined(MULT_INTERLEAVE4X4_HEIGHT) && defined(TRANSPOSE1XW_WIDTH_STEP)
/** This OpenCL kernel computes the matrix multiplication between matrix A (src0) and matrix B (src1)
- * Matrix A and matrix B must be reshaped respectively with @ref CLGEMMInterleave4x4Kernel and @ref CLGEMMTranspose1xWKernel before running the matrix multiplication
+ * Matrix A and matrix B must be reshaped respectively with @ref CLGEMMReshapeLHSMatrixKernel and @ref CLGEMMReshapeRHSMatrixKernel before running the matrix multiplication
*
* @note The number of matrix B columns needs to be passed at compile time using -DCOLS_B: e.g. -DCOLS_B=1024
* @note The transposition width step (mult_transpose1xW_width * 4) must be passed at compile time using -DTRANSPOSE1XW_WIDTH_STEP (i.e. -DTRANSPOSE1XW_WIDTH_STEP=2)
@@ -195,7 +195,7 @@ __kernel void gemmlowp_mm_interleaved_transposed_midgard(IMAGE_DECLARATION(src0)
}
/** This OpenCL kernel is optimized for Bifrost and computes the matrix multiplication between matrix A (src0) and matrix B (src1)
- * Matrix A and matrix B must be reshaped respectively with @ref CLGEMMInterleave4x4Kernel and @ref CLGEMMTranspose1xWKernel before running the matrix multiplication
+ * Matrix A and matrix B must be reshaped respectively with @ref CLGEMMReshapeLHSMatrixKernel and @ref CLGEMMReshapeRHSMatrixKernel before running the matrix multiplication
*
* @attention The number of matrix B columns needs to be passed at compile time using -DCOLS_B
* @note The transposition width step (mult_transpose1xW_width * 4) must be passed at compile time using -DTRANSPOSE1XW_WIDTH_STEP (i.e. -DTRANSPOSE1XW_WIDTH_STEP=2)
@@ -554,7 +554,7 @@ __kernel void gemmlowp_mm_interleaved_transposed_bifrost(IMAGE_DECLARATION(src0)
#if defined(ARM_COMPUTE_OPENCL_DOT8_ENABLED) && defined(cl_arm_integer_dot_product_int8)
/** This OpenCL kernel is optimized for Bifrost and computes the matrix multiplication between matrix A (src0) and matrix B (src1)
- * Matrix A and matrix B must be reshaped respectively with @ref CLGEMMInterleave4x4Kernel and @ref CLGEMMTranspose1xWKernel before running the matrix multiplication
+ * Matrix A and matrix B must be reshaped respectively with @ref CLGEMMReshapeLHSMatrixKernel and @ref CLGEMMReshapeRHSMatrixKernel before running the matrix multiplication
*
* @attention The number of matrix B columns needs to be passed at compile time using -DCOLS_B
* @note The transposition width step (mult_transpose1xW_width * 4) must be passed at compile time using -DTRANSPOSE1XW_WIDTH_STEP (i.e. -DTRANSPOSE1XW_WIDTH_STEP=2)