cv::transpose

This commit is contained in:
Ilya Lavrenov
2014-06-27 12:44:32 +04:00
parent cbd0511a56
commit 9c8b9fc733
2 changed files with 5 additions and 3 deletions

View File

@@ -53,7 +53,7 @@
#define TSIZE ((int)sizeof(T1)*3)
#endif
#define LDS_STEP TILE_DIM
#define LDS_STEP (TILE_DIM + 1)
__kernel void transpose(__global const uchar * srcptr, int src_step, int src_offset, int src_rows, int src_cols,
__global uchar * dstptr, int dst_step, int dst_offset)
@@ -90,6 +90,7 @@ __kernel void transpose(__global const uchar * srcptr, int src_step, int src_off
{
int index_src = mad24(y, src_step, mad24(x, TSIZE, src_offset));
#pragma unroll
for (int i = 0; i < TILE_DIM; i += BLOCK_ROWS)
if (y + i < src_rows)
{
@@ -103,6 +104,7 @@ __kernel void transpose(__global const uchar * srcptr, int src_step, int src_off
{
int index_dst = mad24(y_index, dst_step, mad24(x_index, TSIZE, dst_offset));
#pragma unroll
for (int i = 0; i < TILE_DIM; i += BLOCK_ROWS)
if ((y_index + i) < src_cols)
{