mirror/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2025-06-07 09:25:45 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Merge pull request #24740 from opencv-pushbot:gitee/alalek/ocl_fix_kernel_compilation

Browse Source 
ocl: fix kernels compilation
This commit is contained in:
Alexander Smorkalov 2023-12-22 09:20:47 +03:00 committed by GitHub
commit 35e2ef8019
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
3 changed files with 45 additions and 45 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

34
modules/dnn/src/opencl/conv_layer_spatial.cl
View File

@ -95,23 +95,23 @@
#define __CAT(x, y) x##y
#define CAT(x, y) __CAT(x, y)
#define LOOP0(VAR, STMT)
#define LOOP1(VAR, STMT) (STMT); (VAR)++;
#define LOOP2(VAR, STMT) LOOP1(VAR, STMT); (STMT); (VAR)++;
#define LOOP3(VAR, STMT) LOOP2(VAR, STMT); (STMT); (VAR)++;
#define LOOP4(VAR, STMT) LOOP3(VAR, STMT); (STMT); (VAR)++;
#define LOOP5(VAR, STMT) LOOP4(VAR, STMT); (STMT); (VAR)++;
#define LOOP6(VAR, STMT) LOOP5(VAR, STMT); (STMT); (VAR)++;
#define LOOP7(VAR, STMT) LOOP6(VAR, STMT); (STMT); (VAR)++;
#define LOOP8(VAR, STMT) LOOP7(VAR, STMT); (STMT); (VAR)++;
#define LOOP9(VAR, STMT) LOOP8(VAR, STMT); (STMT); (VAR)++;
#define LOOP10(VAR, STMT) LOOP9(VAR, STMT); (STMT); (VAR)++;
#define LOOP11(VAR, STMT) LOOP10(VAR, STMT); (STMT); (VAR)++;
#define LOOP12(VAR, STMT) LOOP11(VAR, STMT); (STMT); (VAR)++;
#define LOOP13(VAR, STMT) LOOP12(VAR, STMT); (STMT); (VAR)++;
#define LOOP14(VAR, STMT) LOOP13(VAR, STMT); (STMT); (VAR)++;
#define LOOP15(VAR, STMT) LOOP14(VAR, STMT); (STMT); (VAR)++;
#define LOOP16(VAR, STMT) LOOP15(VAR, STMT); (STMT); (VAR)++;
#define LOOP(N, VAR, STMT) CAT(LOOP, N)((VAR), (STMT))
#define LOOP1(VAR, STMT) STMT; (VAR)++;
#define LOOP2(VAR, STMT) LOOP1(VAR, STMT); STMT; (VAR)++;
#define LOOP3(VAR, STMT) LOOP2(VAR, STMT); STMT; (VAR)++;
#define LOOP4(VAR, STMT) LOOP3(VAR, STMT); STMT; (VAR)++;
#define LOOP5(VAR, STMT) LOOP4(VAR, STMT); STMT; (VAR)++;
#define LOOP6(VAR, STMT) LOOP5(VAR, STMT); STMT; (VAR)++;
#define LOOP7(VAR, STMT) LOOP6(VAR, STMT); STMT; (VAR)++;
#define LOOP8(VAR, STMT) LOOP7(VAR, STMT); STMT; (VAR)++;
#define LOOP9(VAR, STMT) LOOP8(VAR, STMT); STMT; (VAR)++;
#define LOOP10(VAR, STMT) LOOP9(VAR, STMT); STMT; (VAR)++;
#define LOOP11(VAR, STMT) LOOP10(VAR, STMT); STMT; (VAR)++;
#define LOOP12(VAR, STMT) LOOP11(VAR, STMT); STMT; (VAR)++;
#define LOOP13(VAR, STMT) LOOP12(VAR, STMT); STMT; (VAR)++;
#define LOOP14(VAR, STMT) LOOP13(VAR, STMT); STMT; (VAR)++;
#define LOOP15(VAR, STMT) LOOP14(VAR, STMT); STMT; (VAR)++;
#define LOOP16(VAR, STMT) LOOP15(VAR, STMT); STMT; (VAR)++;
#define LOOP(N, VAR, STMT) CAT(LOOP, N)(VAR, STMT)
#if defined(convolve_simd) || defined(Conv_Interleaved)
#if TYPE == TYPE_HALF

28
modules/imgproc/src/opencl/filter2DSmall.cl
View File

@ -256,21 +256,21 @@ inline PX_LOAD_FLOAT_VEC_TYPE readSrcPixelGroup(int2 pos, __global const uchar*
}
// Macros to ensure unrolled loops
#define LOOP1(VAR, STMT) (STMT); (VAR)++;
#define LOOP2(VAR, STMT) LOOP1(VAR, STMT); (STMT); (VAR)++;
#define LOOP3(VAR, STMT) LOOP2(VAR, STMT); (STMT); (VAR)++;
#define LOOP4(VAR, STMT) LOOP3(VAR, STMT); (STMT); (VAR)++;
#define LOOP5(VAR, STMT) LOOP4(VAR, STMT); (STMT); (VAR)++;
#define LOOP6(VAR, STMT) LOOP5(VAR, STMT); (STMT); (VAR)++;
#define LOOP7(VAR, STMT) LOOP6(VAR, STMT); (STMT); (VAR)++;
#define LOOP8(VAR, STMT) LOOP7(VAR, STMT); (STMT); (VAR)++;
#define LOOP9(VAR, STMT) LOOP8(VAR, STMT); (STMT); (VAR)++;
#define LOOP10(VAR, STMT) LOOP9(VAR, STMT); (STMT); (VAR)++;
#define LOOP11(VAR, STMT) LOOP10(VAR, STMT); (STMT); (VAR)++;
#define LOOP12(VAR, STMT) LOOP11(VAR, STMT); (STMT); (VAR)++;
#define LOOP13(VAR, STMT) LOOP12(VAR, STMT); (STMT); (VAR)++;
#define LOOP1(VAR, STMT) STMT; (VAR)++;
#define LOOP2(VAR, STMT) LOOP1(VAR, STMT); STMT; (VAR)++;
#define LOOP3(VAR, STMT) LOOP2(VAR, STMT); STMT; (VAR)++;
#define LOOP4(VAR, STMT) LOOP3(VAR, STMT); STMT; (VAR)++;
#define LOOP5(VAR, STMT) LOOP4(VAR, STMT); STMT; (VAR)++;
#define LOOP6(VAR, STMT) LOOP5(VAR, STMT); STMT; (VAR)++;
#define LOOP7(VAR, STMT) LOOP6(VAR, STMT); STMT; (VAR)++;
#define LOOP8(VAR, STMT) LOOP7(VAR, STMT); STMT; (VAR)++;
#define LOOP9(VAR, STMT) LOOP8(VAR, STMT); STMT; (VAR)++;
#define LOOP10(VAR, STMT) LOOP9(VAR, STMT); STMT; (VAR)++;
#define LOOP11(VAR, STMT) LOOP10(VAR, STMT); STMT; (VAR)++;
#define LOOP12(VAR, STMT) LOOP11(VAR, STMT); STMT; (VAR)++;
#define LOOP13(VAR, STMT) LOOP12(VAR, STMT); STMT; (VAR)++;
#define LOOP(N, VAR, STMT) CAT(LOOP, N)((VAR), (STMT))
#define LOOP(N, VAR, STMT) CAT(LOOP, N)(VAR, STMT)
#define DIG(a) a,
__constant WT1 kernelData[] = { COEFF };

28
modules/imgproc/src/opencl/filterSmall.cl
View File

@ -177,21 +177,21 @@ inline PX_LOAD_FLOAT_VEC_TYPE readSrcPixelGroup(int2 pos, __global const uchar *
}
// Macros to ensure unrolled loops
#define LOOP1(VAR, STMT) (STMT); (VAR)++;
#define LOOP2(VAR, STMT) LOOP1(VAR, STMT); (STMT); (VAR)++;
#define LOOP3(VAR, STMT) LOOP2(VAR, STMT); (STMT); (VAR)++;
#define LOOP4(VAR, STMT) LOOP3(VAR, STMT); (STMT); (VAR)++;
#define LOOP5(VAR, STMT) LOOP4(VAR, STMT); (STMT); (VAR)++;
#define LOOP6(VAR, STMT) LOOP5(VAR, STMT); (STMT); (VAR)++;
#define LOOP7(VAR, STMT) LOOP6(VAR, STMT); (STMT); (VAR)++;
#define LOOP8(VAR, STMT) LOOP7(VAR, STMT); (STMT); (VAR)++;
#define LOOP9(VAR, STMT) LOOP8(VAR, STMT); (STMT); (VAR)++;
#define LOOP10(VAR, STMT) LOOP9(VAR, STMT); (STMT); (VAR)++;
#define LOOP11(VAR, STMT) LOOP10(VAR, STMT); (STMT); (VAR)++;
#define LOOP12(VAR, STMT) LOOP11(VAR, STMT); (STMT); (VAR)++;
#define LOOP13(VAR, STMT) LOOP12(VAR, STMT); (STMT); (VAR)++;
#define LOOP1(VAR, STMT) STMT; (VAR)++;
#define LOOP2(VAR, STMT) LOOP1(VAR, STMT); STMT; (VAR)++;
#define LOOP3(VAR, STMT) LOOP2(VAR, STMT); STMT; (VAR)++;
#define LOOP4(VAR, STMT) LOOP3(VAR, STMT); STMT; (VAR)++;
#define LOOP5(VAR, STMT) LOOP4(VAR, STMT); STMT; (VAR)++;
#define LOOP6(VAR, STMT) LOOP5(VAR, STMT); STMT; (VAR)++;
#define LOOP7(VAR, STMT) LOOP6(VAR, STMT); STMT; (VAR)++;
#define LOOP8(VAR, STMT) LOOP7(VAR, STMT); STMT; (VAR)++;
#define LOOP9(VAR, STMT) LOOP8(VAR, STMT); STMT; (VAR)++;
#define LOOP10(VAR, STMT) LOOP9(VAR, STMT); STMT; (VAR)++;
#define LOOP11(VAR, STMT) LOOP10(VAR, STMT); STMT; (VAR)++;
#define LOOP12(VAR, STMT) LOOP11(VAR, STMT); STMT; (VAR)++;
#define LOOP13(VAR, STMT) LOOP12(VAR, STMT); STMT; (VAR)++;
#define LOOP(N, VAR, STMT) CAT(LOOP, N)((VAR), (STMT))
#define LOOP(N, VAR, STMT) CAT(LOOP, N)(VAR, STMT)
#ifdef OP_BOX_FILTER
#define PROCESS_ELEM \