From bee970ab94384664c005142083e806bcf4a870f0 Mon Sep 17 00:00:00 2001 From: yao Date: Fri, 5 Apr 2013 21:29:29 +0800 Subject: [PATCH 1/2] remove the C3 kernels in arithm, as the oclMat will never store 3 channels data --- modules/ocl/src/opencl/arithm_add.cl | 315 -------------- modules/ocl/src/opencl/arithm_add_scalar.cl | 267 ------------ .../ocl/src/opencl/arithm_add_scalar_mask.cl | 305 -------------- .../ocl/src/opencl/arithm_bitwise_and_mask.cl | 391 ------------------ .../src/opencl/arithm_bitwise_and_scalar.cl | 333 --------------- .../ocl/src/opencl/arithm_bitwise_or_mask.cl | 390 ----------------- .../src/opencl/arithm_bitwise_or_scalar.cl | 337 --------------- .../opencl/arithm_bitwise_or_scalar_mask.cl | 380 ----------------- .../ocl/src/opencl/arithm_bitwise_xor_mask.cl | 391 ------------------ .../src/opencl/arithm_bitwise_xor_scalar.cl | 333 --------------- .../opencl/arithm_bitwise_xor_scalar_mask.cl | 373 ----------------- modules/ocl/src/opencl/arithm_sub.cl | 310 +------------- modules/ocl/src/opencl/arithm_sub_scalar.cl | 303 +------------- .../ocl/src/opencl/arithm_sub_scalar_mask.cl | 338 +-------------- modules/ocl/test/test_arithm.cpp | 12 +- 15 files changed, 20 insertions(+), 4758 deletions(-) diff --git a/modules/ocl/src/opencl/arithm_add.cl b/modules/ocl/src/opencl/arithm_add.cl index 647171578d..f8f32cde6b 100644 --- a/modules/ocl/src/opencl/arithm_add.cl +++ b/modules/ocl/src/opencl/arithm_add.cl @@ -675,322 +675,7 @@ __kernel void arithm_add_with_mask_C2_D6 (__global double *src1, int src1_step, } } #endif -__kernel void arithm_add_with_mask_C3_D0 (__global uchar *src1, int src1_step, int src1_offset, - __global uchar *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global uchar *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 3 ) & 3) - int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3)); - int src2_index = mad24(y, src2_step, (x * 3) + src2_offset - (dst_align * 3)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3)); - uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0); - uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4); - uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8); - - uchar4 src2_data_0 = vload4(0, src2 + src2_index + 0); - uchar4 src2_data_1 = vload4(0, src2 + src2_index + 4); - uchar4 src2_data_2 = vload4(0, src2 + src2_index + 8); - - uchar4 mask_data = vload4(0, mask + mask_index); - - uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0)); - uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4)); - uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8)); - - uchar4 tmp_data_0 = convert_uchar4_sat(convert_short4_sat(src1_data_0) + convert_short4_sat(src2_data_0)); - uchar4 tmp_data_1 = convert_uchar4_sat(convert_short4_sat(src1_data_1) + convert_short4_sat(src2_data_1)); - uchar4 tmp_data_2 = convert_uchar4_sat(convert_short4_sat(src1_data_2) + convert_short4_sat(src2_data_2)); - - data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz; - data_0.w = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_0.w : data_0.w; - - data_1.xy = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_1.xy : data_1.xy; - data_1.zw = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.zw : data_1.zw; - - data_2.x = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.x : data_2.x; - data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end)) - ? tmp_data_2.yzw : data_2.yzw; - - *((__global uchar4 *)(dst + dst_index + 0)) = data_0; - *((__global uchar4 *)(dst + dst_index + 4)) = data_1; - *((__global uchar4 *)(dst + dst_index + 8)) = data_2; - } -} -__kernel void arithm_add_with_mask_C3_D2 (__global ushort *src1, int src1_step, int src1_offset, - __global ushort *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global ushort *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - int src2_index = mad24(y, src2_step, (x * 6) + src2_offset - (dst_align * 6)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0)); - ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4)); - ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8)); - - ushort2 src2_data_0 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 0)); - ushort2 src2_data_1 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 4)); - ushort2 src2_data_2 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 8)); - - uchar2 mask_data = vload2(0, mask + mask_index); - - ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0)); - ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4)); - ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8)); - - ushort2 tmp_data_0 = convert_ushort2_sat(convert_int2_sat(src1_data_0) + convert_int2_sat(src2_data_0)); - ushort2 tmp_data_1 = convert_ushort2_sat(convert_int2_sat(src1_data_1) + convert_int2_sat(src2_data_1)); - ushort2 tmp_data_2 = convert_ushort2_sat(convert_int2_sat(src1_data_2) + convert_int2_sat(src2_data_2)); - - data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_add_with_mask_C3_D3 (__global short *src1, int src1_step, int src1_offset, - __global short *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global short *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - int src2_index = mad24(y, src2_step, (x * 6) + src2_offset - (dst_align * 6)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0)); - short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4)); - short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8)); - - short2 src2_data_0 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 0)); - short2 src2_data_1 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 4)); - short2 src2_data_2 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 8)); - - uchar2 mask_data = vload2(0, mask + mask_index); - - short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0)); - short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4)); - short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8)); - - short2 tmp_data_0 = convert_short2_sat(convert_int2_sat(src1_data_0) + convert_int2_sat(src2_data_0)); - short2 tmp_data_1 = convert_short2_sat(convert_int2_sat(src1_data_1) + convert_int2_sat(src2_data_1)); - short2 tmp_data_2 = convert_short2_sat(convert_int2_sat(src1_data_2) + convert_int2_sat(src2_data_2)); - - data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_add_with_mask_C3_D4 (__global int *src1, int src1_step, int src1_offset, - __global int *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global int *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int src2_index = mad24(y, src2_step, (x * 12) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0)); - int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4)); - int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8)); - - int src2_data_0 = *((__global int *)((__global char *)src2 + src2_index + 0)); - int src2_data_1 = *((__global int *)((__global char *)src2 + src2_index + 4)); - int src2_data_2 = *((__global int *)((__global char *)src2 + src2_index + 8)); - - uchar mask_data = * (mask + mask_index); - - int data_0 = *((__global int *)((__global char *)dst + dst_index + 0)); - int data_1 = *((__global int *)((__global char *)dst + dst_index + 4)); - int data_2 = *((__global int *)((__global char *)dst + dst_index + 8)); - - int tmp_data_0 = convert_int_sat((long)src1_data_0 + (long)src2_data_0); - int tmp_data_1 = convert_int_sat((long)src1_data_1 + (long)src2_data_1); - int tmp_data_2 = convert_int_sat((long)src1_data_2 + (long)src2_data_2); - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global int *)((__global char *)dst + dst_index + 0))= data_0; - *((__global int *)((__global char *)dst + dst_index + 4))= data_1; - *((__global int *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_add_with_mask_C3_D5 (__global float *src1, int src1_step, int src1_offset, - __global float *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global float *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int src2_index = mad24(y, src2_step, (x * 12) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - float src1_data_0 = *((__global float *)((__global char *)src1 + src1_index + 0)); - float src1_data_1 = *((__global float *)((__global char *)src1 + src1_index + 4)); - float src1_data_2 = *((__global float *)((__global char *)src1 + src1_index + 8)); - - float src2_data_0 = *((__global float *)((__global char *)src2 + src2_index + 0)); - float src2_data_1 = *((__global float *)((__global char *)src2 + src2_index + 4)); - float src2_data_2 = *((__global float *)((__global char *)src2 + src2_index + 8)); - - uchar mask_data = * (mask + mask_index); - - float data_0 = *((__global float *)((__global char *)dst + dst_index + 0)); - float data_1 = *((__global float *)((__global char *)dst + dst_index + 4)); - float data_2 = *((__global float *)((__global char *)dst + dst_index + 8)); - - float tmp_data_0 = src1_data_0 + src2_data_0; - float tmp_data_1 = src1_data_1 + src2_data_1; - float tmp_data_2 = src1_data_2 + src2_data_2; - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global float *)((__global char *)dst + dst_index + 0))= data_0; - *((__global float *)((__global char *)dst + dst_index + 4))= data_1; - *((__global float *)((__global char *)dst + dst_index + 8))= data_2; - } -} - -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_add_with_mask_C3_D6 (__global double *src1, int src1_step, int src1_offset, - __global double *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global double *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 24) + src1_offset); - int src2_index = mad24(y, src2_step, (x * 24) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 24)); - - double src1_data_0 = *((__global double *)((__global char *)src1 + src1_index + 0 )); - double src1_data_1 = *((__global double *)((__global char *)src1 + src1_index + 8 )); - double src1_data_2 = *((__global double *)((__global char *)src1 + src1_index + 16)); - - double src2_data_0 = *((__global double *)((__global char *)src2 + src2_index + 0 )); - double src2_data_1 = *((__global double *)((__global char *)src2 + src2_index + 8 )); - double src2_data_2 = *((__global double *)((__global char *)src2 + src2_index + 16)); - - uchar mask_data = * (mask + mask_index); - - double data_0 = *((__global double *)((__global char *)dst + dst_index + 0 )); - double data_1 = *((__global double *)((__global char *)dst + dst_index + 8 )); - double data_2 = *((__global double *)((__global char *)dst + dst_index + 16)); - - double tmp_data_0 = src1_data_0 + src2_data_0; - double tmp_data_1 = src1_data_1 + src2_data_1; - double tmp_data_2 = src1_data_2 + src2_data_2; - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global double *)((__global char *)dst + dst_index + 0 ))= data_0; - *((__global double *)((__global char *)dst + dst_index + 8 ))= data_1; - *((__global double *)((__global char *)dst + dst_index + 16))= data_2; - } -} -#endif __kernel void arithm_add_with_mask_C4_D0 (__global uchar *src1, int src1_step, int src1_offset, __global uchar *src2, int src2_step, int src2_offset, __global uchar *mask, int mask_step, int mask_offset, diff --git a/modules/ocl/src/opencl/arithm_add_scalar.cl b/modules/ocl/src/opencl/arithm_add_scalar.cl index 15ae95df25..152b5a19fb 100644 --- a/modules/ocl/src/opencl/arithm_add_scalar.cl +++ b/modules/ocl/src/opencl/arithm_add_scalar.cl @@ -382,274 +382,7 @@ __kernel void arithm_s_add_C2_D6 (__global double *src1, int src1_step, int sr } } #endif -__kernel void arithm_s_add_C3_D0 (__global uchar *src1, int src1_step, int src1_offset, - __global uchar *dst, int dst_step, int dst_offset, - int4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 3 ) & 3) - int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3)); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3)); - - uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0); - uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4); - uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8); - - int4 src2_data_0 = (int4)(src2.x, src2.y, src2.z, src2.x); - int4 src2_data_1 = (int4)(src2.y, src2.z, src2.x, src2.y); - int4 src2_data_2 = (int4)(src2.z, src2.x, src2.y, src2.z); - - uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0)); - uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4)); - uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8)); - - uchar4 tmp_data_0 = convert_uchar4_sat(convert_int4_sat(src1_data_0) + src2_data_0); - uchar4 tmp_data_1 = convert_uchar4_sat(convert_int4_sat(src1_data_1) + src2_data_1); - uchar4 tmp_data_2 = convert_uchar4_sat(convert_int4_sat(src1_data_2) + src2_data_2); - - data_0.xyz = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz; - data_0.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_0.w : data_0.w; - - data_1.xy = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_1.xy : data_1.xy; - data_1.zw = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.zw : data_1.zw; - - data_2.x = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.x : data_2.x; - data_2.yzw = ((dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end)) - ? tmp_data_2.yzw : data_2.yzw; - - *((__global uchar4 *)(dst + dst_index + 0)) = data_0; - *((__global uchar4 *)(dst + dst_index + 4)) = data_1; - *((__global uchar4 *)(dst + dst_index + 8)) = data_2; - } -} -__kernel void arithm_s_add_C3_D2 (__global ushort *src1, int src1_step, int src1_offset, - __global ushort *dst, int dst_step, int dst_offset, - int4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - if (x < cols && y < rows) - { - x = x << 1; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0)); - ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4)); - ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8)); - - int2 src2_data_0 = (int2)(src2.x, src2.y); - int2 src2_data_1 = (int2)(src2.z, src2.x); - int2 src2_data_2 = (int2)(src2.y, src2.z); - - ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0)); - ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4)); - ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8)); - - ushort2 tmp_data_0 = convert_ushort2_sat(convert_int2_sat(src1_data_0) + src2_data_0); - ushort2 tmp_data_1 = convert_ushort2_sat(convert_int2_sat(src1_data_1) + src2_data_1); - ushort2 tmp_data_2 = convert_ushort2_sat(convert_int2_sat(src1_data_2) + src2_data_2); - - data_0.xy = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_s_add_C3_D3 (__global short *src1, int src1_step, int src1_offset, - __global short *dst, int dst_step, int dst_offset, - int4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0)); - short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4)); - short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8)); - - int2 src2_data_0 = (int2)(src2.x, src2.y); - int2 src2_data_1 = (int2)(src2.z, src2.x); - int2 src2_data_2 = (int2)(src2.y, src2.z); - - short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0)); - short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4)); - short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8)); - - short2 tmp_data_0 = convert_short2_sat(convert_int2_sat(src1_data_0) + src2_data_0); - short2 tmp_data_1 = convert_short2_sat(convert_int2_sat(src1_data_1) + src2_data_1); - short2 tmp_data_2 = convert_short2_sat(convert_int2_sat(src1_data_2) + src2_data_2); - - data_0.xy = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_s_add_C3_D4 (__global int *src1, int src1_step, int src1_offset, - __global int *dst, int dst_step, int dst_offset, - int4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0)); - int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4)); - int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8)); - - int src2_data_0 = src2.x; - int src2_data_1 = src2.y; - int src2_data_2 = src2.z; - - int data_0 = *((__global int *)((__global char *)dst + dst_index + 0)); - int data_1 = *((__global int *)((__global char *)dst + dst_index + 4)); - int data_2 = *((__global int *)((__global char *)dst + dst_index + 8)); - - int tmp_data_0 = convert_int_sat((long)src1_data_0 + (long)src2_data_0); - int tmp_data_1 = convert_int_sat((long)src1_data_1 + (long)src2_data_1); - int tmp_data_2 = convert_int_sat((long)src1_data_2 + (long)src2_data_2); - - *((__global int *)((__global char *)dst + dst_index + 0))= tmp_data_0; - *((__global int *)((__global char *)dst + dst_index + 4))= tmp_data_1; - *((__global int *)((__global char *)dst + dst_index + 8))= tmp_data_2; - } -} -__kernel void arithm_s_add_C3_D5 (__global float *src1, int src1_step, int src1_offset, - __global float *dst, int dst_step, int dst_offset, - float4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - float src1_data_0 = *((__global float *)((__global char *)src1 + src1_index + 0)); - float src1_data_1 = *((__global float *)((__global char *)src1 + src1_index + 4)); - float src1_data_2 = *((__global float *)((__global char *)src1 + src1_index + 8)); - - float src2_data_0 = src2.x; - float src2_data_1 = src2.y; - float src2_data_2 = src2.z; - - float data_0 = *((__global float *)((__global char *)dst + dst_index + 0)); - float data_1 = *((__global float *)((__global char *)dst + dst_index + 4)); - float data_2 = *((__global float *)((__global char *)dst + dst_index + 8)); - - float tmp_data_0 = src1_data_0 + src2_data_0; - float tmp_data_1 = src1_data_1 + src2_data_1; - float tmp_data_2 = src1_data_2 + src2_data_2; - - *((__global float *)((__global char *)dst + dst_index + 0))= tmp_data_0; - *((__global float *)((__global char *)dst + dst_index + 4))= tmp_data_1; - *((__global float *)((__global char *)dst + dst_index + 8))= tmp_data_2; - } -} - -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_s_add_C3_D6 (__global double *src1, int src1_step, int src1_offset, - __global double *dst, int dst_step, int dst_offset, - double4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 24) + src1_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 24)); - - double src1_data_0 = *((__global double *)((__global char *)src1 + src1_index + 0 )); - double src1_data_1 = *((__global double *)((__global char *)src1 + src1_index + 8 )); - double src1_data_2 = *((__global double *)((__global char *)src1 + src1_index + 16)); - - double src2_data_0 = src2.x; - double src2_data_1 = src2.y; - double src2_data_2 = src2.z; - - double data_0 = *((__global double *)((__global char *)dst + dst_index + 0 )); - double data_1 = *((__global double *)((__global char *)dst + dst_index + 8 )); - double data_2 = *((__global double *)((__global char *)dst + dst_index + 16)); - - double tmp_data_0 = src1_data_0 + src2_data_0; - double tmp_data_1 = src1_data_1 + src2_data_1; - double tmp_data_2 = src1_data_2 + src2_data_2; - - *((__global double *)((__global char *)dst + dst_index + 0 ))= tmp_data_0; - *((__global double *)((__global char *)dst + dst_index + 8 ))= tmp_data_1; - *((__global double *)((__global char *)dst + dst_index + 16))= tmp_data_2; - } -} -#endif __kernel void arithm_s_add_C4_D0 (__global uchar *src1, int src1_step, int src1_offset, __global uchar *dst, int dst_step, int dst_offset, int4 src2, int rows, int cols, int dst_step1) diff --git a/modules/ocl/src/opencl/arithm_add_scalar_mask.cl b/modules/ocl/src/opencl/arithm_add_scalar_mask.cl index 1e2ae71af6..673e323ff6 100644 --- a/modules/ocl/src/opencl/arithm_add_scalar_mask.cl +++ b/modules/ocl/src/opencl/arithm_add_scalar_mask.cl @@ -446,311 +446,6 @@ __kernel void arithm_s_add_with_mask_C2_D6 (__global double *src1, int src1_st } #endif -__kernel void arithm_s_add_with_mask_C3_D0 (__global uchar *src1, int src1_step, int src1_offset, - __global uchar *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - int4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 3 ) & 3) - int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3)); - - uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0); - uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4); - uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8); - - int4 src2_data_0 = (int4)(src2.x, src2.y, src2.z, src2.x); - int4 src2_data_1 = (int4)(src2.y, src2.z, src2.x, src2.y); - int4 src2_data_2 = (int4)(src2.z, src2.x, src2.y, src2.z); - - uchar4 mask_data = vload4(0, mask + mask_index); - - uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0)); - uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4)); - uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8)); - - uchar4 tmp_data_0 = convert_uchar4_sat(convert_int4_sat(src1_data_0) + src2_data_0); - uchar4 tmp_data_1 = convert_uchar4_sat(convert_int4_sat(src1_data_1) + src2_data_1); - uchar4 tmp_data_2 = convert_uchar4_sat(convert_int4_sat(src1_data_2) + src2_data_2); - - data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz; - data_0.w = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_0.w : data_0.w; - - data_1.xy = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_1.xy : data_1.xy; - data_1.zw = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.zw : data_1.zw; - - data_2.x = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.x : data_2.x; - data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end)) - ? tmp_data_2.yzw : data_2.yzw; - - *((__global uchar4 *)(dst + dst_index + 0)) = data_0; - *((__global uchar4 *)(dst + dst_index + 4)) = data_1; - *((__global uchar4 *)(dst + dst_index + 8)) = data_2; - } -} -__kernel void arithm_s_add_with_mask_C3_D2 (__global ushort *src1, int src1_step, int src1_offset, - __global ushort *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - int4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0)); - ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4)); - ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8)); - - int2 src2_data_0 = (int2)(src2.x, src2.y); - int2 src2_data_1 = (int2)(src2.z, src2.x); - int2 src2_data_2 = (int2)(src2.y, src2.z); - - uchar2 mask_data = vload2(0, mask + mask_index); - - ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0)); - ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4)); - ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8)); - - ushort2 tmp_data_0 = convert_ushort2_sat(convert_int2_sat(src1_data_0) + src2_data_0); - ushort2 tmp_data_1 = convert_ushort2_sat(convert_int2_sat(src1_data_1) + src2_data_1); - ushort2 tmp_data_2 = convert_ushort2_sat(convert_int2_sat(src1_data_2) + src2_data_2); - - data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_s_add_with_mask_C3_D3 (__global short *src1, int src1_step, int src1_offset, - __global short *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - int4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0)); - short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4)); - short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8)); - - int2 src2_data_0 = (int2)(src2.x, src2.y); - int2 src2_data_1 = (int2)(src2.z, src2.x); - int2 src2_data_2 = (int2)(src2.y, src2.z); - - uchar2 mask_data = vload2(0, mask + mask_index); - - short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0)); - short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4)); - short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8)); - - short2 tmp_data_0 = convert_short2_sat(convert_int2_sat(src1_data_0) + src2_data_0); - short2 tmp_data_1 = convert_short2_sat(convert_int2_sat(src1_data_1) + src2_data_1); - short2 tmp_data_2 = convert_short2_sat(convert_int2_sat(src1_data_2) + src2_data_2); - - data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_s_add_with_mask_C3_D4 (__global int *src1, int src1_step, int src1_offset, - __global int *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - int4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0)); - int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4)); - int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8)); - - int src2_data_0 = src2.x; - int src2_data_1 = src2.y; - int src2_data_2 = src2.z; - - uchar mask_data = * (mask + mask_index); - - int data_0 = *((__global int *)((__global char *)dst + dst_index + 0)); - int data_1 = *((__global int *)((__global char *)dst + dst_index + 4)); - int data_2 = *((__global int *)((__global char *)dst + dst_index + 8)); - - int tmp_data_0 = convert_int_sat((long)src1_data_0 + (long)src2_data_0); - int tmp_data_1 = convert_int_sat((long)src1_data_1 + (long)src2_data_1); - int tmp_data_2 = convert_int_sat((long)src1_data_2 + (long)src2_data_2); - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global int *)((__global char *)dst + dst_index + 0))= data_0; - *((__global int *)((__global char *)dst + dst_index + 4))= data_1; - *((__global int *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_s_add_with_mask_C3_D5 (__global float *src1, int src1_step, int src1_offset, - __global float *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - float4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - float src1_data_0 = *((__global float *)((__global char *)src1 + src1_index + 0)); - float src1_data_1 = *((__global float *)((__global char *)src1 + src1_index + 4)); - float src1_data_2 = *((__global float *)((__global char *)src1 + src1_index + 8)); - - float src2_data_0 = src2.x; - float src2_data_1 = src2.y; - float src2_data_2 = src2.z; - - uchar mask_data = * (mask + mask_index); - - float data_0 = *((__global float *)((__global char *)dst + dst_index + 0)); - float data_1 = *((__global float *)((__global char *)dst + dst_index + 4)); - float data_2 = *((__global float *)((__global char *)dst + dst_index + 8)); - - float tmp_data_0 = src1_data_0 + src2_data_0; - float tmp_data_1 = src1_data_1 + src2_data_1; - float tmp_data_2 = src1_data_2 + src2_data_2; - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global float *)((__global char *)dst + dst_index + 0))= data_0; - *((__global float *)((__global char *)dst + dst_index + 4))= data_1; - *((__global float *)((__global char *)dst + dst_index + 8))= data_2; - } -} - -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_s_add_with_mask_C3_D6 (__global double *src1, int src1_step, int src1_offset, - __global double *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - double4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 24) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 24)); - - double src1_data_0 = *((__global double *)((__global char *)src1 + src1_index + 0 )); - double src1_data_1 = *((__global double *)((__global char *)src1 + src1_index + 8 )); - double src1_data_2 = *((__global double *)((__global char *)src1 + src1_index + 16)); - - double src2_data_0 = src2.x; - double src2_data_1 = src2.y; - double src2_data_2 = src2.z; - - uchar mask_data = * (mask + mask_index); - - double data_0 = *((__global double *)((__global char *)dst + dst_index + 0 )); - double data_1 = *((__global double *)((__global char *)dst + dst_index + 8 )); - double data_2 = *((__global double *)((__global char *)dst + dst_index + 16)); - - double tmp_data_0 = src1_data_0 + src2_data_0; - double tmp_data_1 = src1_data_1 + src2_data_1; - double tmp_data_2 = src1_data_2 + src2_data_2; - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global double *)((__global char *)dst + dst_index + 0 ))= data_0; - *((__global double *)((__global char *)dst + dst_index + 8 ))= data_1; - *((__global double *)((__global char *)dst + dst_index + 16))= data_2; - } -} -#endif - __kernel void arithm_s_add_with_mask_C4_D0 (__global uchar *src1, int src1_step, int src1_offset, __global uchar *dst, int dst_step, int dst_offset, __global uchar *mask, int mask_step, int mask_offset, diff --git a/modules/ocl/src/opencl/arithm_bitwise_and_mask.cl b/modules/ocl/src/opencl/arithm_bitwise_and_mask.cl index fbc42364ac..5e0428f34b 100644 --- a/modules/ocl/src/opencl/arithm_bitwise_and_mask.cl +++ b/modules/ocl/src/opencl/arithm_bitwise_and_mask.cl @@ -565,397 +565,6 @@ __kernel void arithm_bitwise_and_with_mask_C2_D6 ( } - -__kernel void arithm_bitwise_and_with_mask_C3_D0 ( - __global uchar *src1, int src1_step, int src1_offset, - __global uchar *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global uchar *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 3 ) & 3) - int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3)); - int src2_index = mad24(y, src2_step, (x * 3) + src2_offset - (dst_align * 3)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3)); - - uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0); - uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4); - uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8); - - uchar4 src2_data_0 = vload4(0, src2 + src2_index + 0); - uchar4 src2_data_1 = vload4(0, src2 + src2_index + 4); - uchar4 src2_data_2 = vload4(0, src2 + src2_index + 8); - - uchar4 mask_data = vload4(0, mask + mask_index); - - uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0)); - uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4)); - uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8)); - - uchar4 tmp_data_0 = src1_data_0 & src2_data_0; - uchar4 tmp_data_1 = src1_data_1 & src2_data_1; - uchar4 tmp_data_2 = src1_data_2 & src2_data_2; - - data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz; - data_0.w = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_0.w : data_0.w; - - data_1.xy = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_1.xy : data_1.xy; - data_1.zw = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.zw : data_1.zw; - - data_2.x = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.x : data_2.x; - data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end)) - ? tmp_data_2.yzw : data_2.yzw; - - *((__global uchar4 *)(dst + dst_index + 0)) = data_0; - *((__global uchar4 *)(dst + dst_index + 4)) = data_1; - *((__global uchar4 *)(dst + dst_index + 8)) = data_2; - } -} - - -__kernel void arithm_bitwise_and_with_mask_C3_D1 ( - __global char *src1, int src1_step, int src1_offset, - __global char *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global char *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 3 ) & 3) - int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3)); - int src2_index = mad24(y, src2_step, (x * 3) + src2_offset - (dst_align * 3)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3)); - - char4 src1_data_0 = vload4(0, src1 + src1_index + 0); - char4 src1_data_1 = vload4(0, src1 + src1_index + 4); - char4 src1_data_2 = vload4(0, src1 + src1_index + 8); - - char4 src2_data_0 = vload4(0, src2 + src2_index + 0); - char4 src2_data_1 = vload4(0, src2 + src2_index + 4); - char4 src2_data_2 = vload4(0, src2 + src2_index + 8); - - uchar4 mask_data = vload4(0, mask + mask_index); - - char4 data_0 = *((__global char4 *)(dst + dst_index + 0)); - char4 data_1 = *((__global char4 *)(dst + dst_index + 4)); - char4 data_2 = *((__global char4 *)(dst + dst_index + 8)); - - char4 tmp_data_0 = src1_data_0 & src2_data_0; - char4 tmp_data_1 = src1_data_1 & src2_data_1; - char4 tmp_data_2 = src1_data_2 & src2_data_2; - - data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz; - data_0.w = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_0.w : data_0.w; - - data_1.xy = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_1.xy : data_1.xy; - data_1.zw = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.zw : data_1.zw; - - data_2.x = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.x : data_2.x; - data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end)) - ? tmp_data_2.yzw : data_2.yzw; - - *((__global char4 *)(dst + dst_index + 0)) = data_0; - *((__global char4 *)(dst + dst_index + 4)) = data_1; - *((__global char4 *)(dst + dst_index + 8)) = data_2; - } -} - -__kernel void arithm_bitwise_and_with_mask_C3_D2 ( - __global ushort *src1, int src1_step, int src1_offset, - __global ushort *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global ushort *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - int src2_index = mad24(y, src2_step, (x * 6) + src2_offset - (dst_align * 6)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0)); - ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4)); - ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8)); - - ushort2 src2_data_0 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 0)); - ushort2 src2_data_1 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 4)); - ushort2 src2_data_2 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 8)); - - uchar2 mask_data = vload2(0, mask + mask_index); - - ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0)); - ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4)); - ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8)); - - ushort2 tmp_data_0 = src1_data_0 & src2_data_0 ; - ushort2 tmp_data_1 = src1_data_1 & src2_data_1 ; - ushort2 tmp_data_2 = src1_data_2 & src2_data_2 ; - - data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_bitwise_and_with_mask_C3_D3 ( - __global short *src1, int src1_step, int src1_offset, - __global short *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global short *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - int src2_index = mad24(y, src2_step, (x * 6) + src2_offset - (dst_align * 6)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0)); - short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4)); - short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8)); - - short2 src2_data_0 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 0)); - short2 src2_data_1 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 4)); - short2 src2_data_2 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 8)); - - uchar2 mask_data = vload2(0, mask + mask_index); - - short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0)); - short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4)); - short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8)); - - short2 tmp_data_0 = src1_data_0 & src2_data_0 ; - short2 tmp_data_1 = src1_data_1 & src2_data_1 ; - short2 tmp_data_2 = src1_data_2 & src2_data_2 ; - - data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_bitwise_and_with_mask_C3_D4 ( - __global int *src1, int src1_step, int src1_offset, - __global int *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global int *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int src2_index = mad24(y, src2_step, (x * 12) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0)); - int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4)); - int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8)); - - int src2_data_0 = *((__global int *)((__global char *)src2 + src2_index + 0)); - int src2_data_1 = *((__global int *)((__global char *)src2 + src2_index + 4)); - int src2_data_2 = *((__global int *)((__global char *)src2 + src2_index + 8)); - - uchar mask_data = * (mask + mask_index); - - int data_0 = *((__global int *)((__global char *)dst + dst_index + 0)); - int data_1 = *((__global int *)((__global char *)dst + dst_index + 4)); - int data_2 = *((__global int *)((__global char *)dst + dst_index + 8)); - - int tmp_data_0 = src1_data_0 & src2_data_0 ; - int tmp_data_1 = src1_data_1 & src2_data_1 ; - int tmp_data_2 = src1_data_2 & src2_data_2 ; - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global int *)((__global char *)dst + dst_index + 0))= data_0; - *((__global int *)((__global char *)dst + dst_index + 4))= data_1; - *((__global int *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_bitwise_and_with_mask_C3_D5 ( - __global char *src1, int src1_step, int src1_offset, - __global char *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global char *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int src2_index = mad24(y, src2_step, (x * 12) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - char4 src1_data_0 = *((__global char4 *)((__global char *)src1 + src1_index + 0)); - char4 src1_data_1 = *((__global char4 *)((__global char *)src1 + src1_index + 4)); - char4 src1_data_2 = *((__global char4 *)((__global char *)src1 + src1_index + 8)); - - char4 src2_data_0 = *((__global char4 *)((__global char *)src2 + src2_index + 0)); - char4 src2_data_1 = *((__global char4 *)((__global char *)src2 + src2_index + 4)); - char4 src2_data_2 = *((__global char4 *)((__global char *)src2 + src2_index + 8)); - - uchar mask_data = * (mask + mask_index); - - char4 data_0 = *((__global char4 *)((__global char *)dst + dst_index + 0)); - char4 data_1 = *((__global char4 *)((__global char *)dst + dst_index + 4)); - char4 data_2 = *((__global char4 *)((__global char *)dst + dst_index + 8)); - - char4 tmp_data_0 = src1_data_0 & src2_data_0; - char4 tmp_data_1 = src1_data_1 & src2_data_1; - char4 tmp_data_2 = src1_data_2 & src2_data_2; - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global char4 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global char4 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global char4 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_bitwise_and_with_mask_C3_D6 ( - __global char *src1, int src1_step, int src1_offset, - __global char *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global char *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 24) + src1_offset); - int src2_index = mad24(y, src2_step, (x * 24) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 24)); - - char8 src1_data_0 = *((__global char8 *)((__global char *)src1 + src1_index + 0 )); - char8 src1_data_1 = *((__global char8 *)((__global char *)src1 + src1_index + 8 )); - char8 src1_data_2 = *((__global char8 *)((__global char *)src1 + src1_index + 16)); - - char8 src2_data_0 = *((__global char8 *)((__global char *)src2 + src2_index + 0 )); - char8 src2_data_1 = *((__global char8 *)((__global char *)src2 + src2_index + 8 )); - char8 src2_data_2 = *((__global char8 *)((__global char *)src2 + src2_index + 16)); - - uchar mask_data = * (mask + mask_index); - - char8 data_0 = *((__global char8 *)((__global char *)dst + dst_index + 0 )); - char8 data_1 = *((__global char8 *)((__global char *)dst + dst_index + 8 )); - char8 data_2 = *((__global char8 *)((__global char *)dst + dst_index + 16)); - - char8 tmp_data_0 = src1_data_0 & src2_data_0; - char8 tmp_data_1 = src1_data_1 & src2_data_1; - char8 tmp_data_2 = src1_data_2 & src2_data_2; - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global char8 *)((__global char *)dst + dst_index + 0 ))= data_0; - *((__global char8 *)((__global char *)dst + dst_index + 8 ))= data_1; - *((__global char8 *)((__global char *)dst + dst_index + 16))= data_2; - } -} -#endif - - __kernel void arithm_bitwise_and_with_mask_C4_D0 ( __global uchar *src1, int src1_step, int src1_offset, __global uchar *src2, int src2_step, int src2_offset, diff --git a/modules/ocl/src/opencl/arithm_bitwise_and_scalar.cl b/modules/ocl/src/opencl/arithm_bitwise_and_scalar.cl index 5058d318e0..9605476581 100644 --- a/modules/ocl/src/opencl/arithm_bitwise_and_scalar.cl +++ b/modules/ocl/src/opencl/arithm_bitwise_and_scalar.cl @@ -461,340 +461,7 @@ __kernel void arithm_s_bitwise_and_C2_D6 ( } } #endif -__kernel void arithm_s_bitwise_and_C3_D0 ( - __global uchar *src1, int src1_step, int src1_offset, - __global uchar *dst, int dst_step, int dst_offset, - uchar4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 3 ) & 3) - int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3)); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3)); - - uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0); - uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4); - uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8); - - uchar4 src2_data_0 = (uchar4)(src2.x, src2.y, src2.z, src2.x); - uchar4 src2_data_1 = (uchar4)(src2.y, src2.z, src2.x, src2.y); - uchar4 src2_data_2 = (uchar4)(src2.z, src2.x, src2.y, src2.z); - - uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0)); - uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4)); - uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8)); - - uchar4 tmp_data_0 = src1_data_0 & src2_data_0; - uchar4 tmp_data_1 = src1_data_1 & src2_data_1; - uchar4 tmp_data_2 = src1_data_2 & src2_data_2; - - data_0.xyz = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz; - data_0.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_0.w : data_0.w; - - data_1.xy = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_1.xy : data_1.xy; - data_1.zw = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.zw : data_1.zw; - - data_2.x = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.x : data_2.x; - data_2.yzw = ((dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end)) - ? tmp_data_2.yzw : data_2.yzw; - - *((__global uchar4 *)(dst + dst_index + 0)) = data_0; - *((__global uchar4 *)(dst + dst_index + 4)) = data_1; - *((__global uchar4 *)(dst + dst_index + 8)) = data_2; - } -} - - -__kernel void arithm_s_bitwise_and_C3_D1 ( - __global char *src1, int src1_step, int src1_offset, - __global char *dst, int dst_step, int dst_offset, - char4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 3 ) & 3) - int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3)); - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3)); - - char4 src1_data_0 = vload4(0, src1 + src1_index + 0); - char4 src1_data_1 = vload4(0, src1 + src1_index + 4); - char4 src1_data_2 = vload4(0, src1 + src1_index + 8); - - char4 src2_data_0 = (char4)(src2.x, src2.y, src2.z, src2.x); - char4 src2_data_1 = (char4)(src2.y, src2.z, src2.x, src2.y); - char4 src2_data_2 = (char4)(src2.z, src2.x, src2.y, src2.z); - - char4 data_0 = *((__global char4 *)(dst + dst_index + 0)); - char4 data_1 = *((__global char4 *)(dst + dst_index + 4)); - char4 data_2 = *((__global char4 *)(dst + dst_index + 8)); - - char4 tmp_data_0 = convert_char4_sat(convert_uchar4_sat(src1_data_0) & convert_uchar4_sat(src2_data_0)); - char4 tmp_data_1 = convert_char4_sat(convert_uchar4_sat(src1_data_1) & convert_uchar4_sat(src2_data_1)); - char4 tmp_data_2 = convert_char4_sat(convert_uchar4_sat(src1_data_2) & convert_uchar4_sat(src2_data_2)); - - data_0.xyz = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz; - data_0.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_0.w : data_0.w; - - data_1.xy = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_1.xy : data_1.xy; - data_1.zw = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.zw : data_1.zw; - - data_2.x = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.x : data_2.x; - data_2.yzw = ((dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end)) - ? tmp_data_2.yzw : data_2.yzw; - - *((__global char4 *)(dst + dst_index + 0)) = data_0; - *((__global char4 *)(dst + dst_index + 4)) = data_1; - *((__global char4 *)(dst + dst_index + 8)) = data_2; - } -} - -__kernel void arithm_s_bitwise_and_C3_D2 ( - __global ushort *src1, int src1_step, int src1_offset, - __global ushort *dst, int dst_step, int dst_offset, - ushort4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0)); - ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4)); - ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8)); - - ushort2 src2_data_0 = (ushort2)(src2.x, src2.y); - ushort2 src2_data_1 = (ushort2)(src2.z, src2.x); - ushort2 src2_data_2 = (ushort2)(src2.y, src2.z); - - ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0)); - ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4)); - ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8)); - - ushort2 tmp_data_0 = src1_data_0 & src2_data_0; - ushort2 tmp_data_1 = src1_data_1 & src2_data_1; - ushort2 tmp_data_2 = src1_data_2 & src2_data_2; - - data_0.xy = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_s_bitwise_and_C3_D3 ( - __global short *src1, int src1_step, int src1_offset, - __global short *dst, int dst_step, int dst_offset, - short4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0)); - short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4)); - short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8)); - - short2 src2_data_0 = (short2)(src2.x, src2.y); - short2 src2_data_1 = (short2)(src2.z, src2.x); - short2 src2_data_2 = (short2)(src2.y, src2.z); - - short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0)); - short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4)); - short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8)); - - short2 tmp_data_0 = src1_data_0 & src2_data_0; - short2 tmp_data_1 = src1_data_1 & src2_data_1; - short2 tmp_data_2 = src1_data_2 & src2_data_2; - - data_0.xy = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_s_bitwise_and_C3_D4 ( - __global int *src1, int src1_step, int src1_offset, - __global int *dst, int dst_step, int dst_offset, - int4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0)); - int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4)); - int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8)); - - int src2_data_0 = src2.x; - int src2_data_1 = src2.y; - int src2_data_2 = src2.z; - - int data_0 = *((__global int *)((__global char *)dst + dst_index + 0)); - int data_1 = *((__global int *)((__global char *)dst + dst_index + 4)); - int data_2 = *((__global int *)((__global char *)dst + dst_index + 8)); - - int tmp_data_0 = src1_data_0 & src2_data_0; - int tmp_data_1 = src1_data_1 & src2_data_1; - int tmp_data_2 = src1_data_2 & src2_data_2; - - *((__global int *)((__global char *)dst + dst_index + 0))= tmp_data_0; - *((__global int *)((__global char *)dst + dst_index + 4))= tmp_data_1; - *((__global int *)((__global char *)dst + dst_index + 8))= tmp_data_2; - } -} -__kernel void arithm_s_bitwise_and_C3_D5 ( - __global char *src1, int src1_step, int src1_offset, - __global char *dst, int dst_step, int dst_offset, - char16 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - char4 src1_data_0 = *((__global char4 *)((__global char *)src1 + src1_index + 0)); - char4 src1_data_1 = *((__global char4 *)((__global char *)src1 + src1_index + 4)); - char4 src1_data_2 = *((__global char4 *)((__global char *)src1 + src1_index + 8)); - - char4 src2_data_0 = (char4)(src2.s0, src2.s1, src2.s2, src2.s3); - char4 src2_data_1 = (char4)(src2.s4, src2.s5, src2.s6, src2.s7); - char4 src2_data_2 = (char4)(src2.s8, src2.s9, src2.sA, src2.sB); - - char4 data_0 = *((__global char4 *)((__global char *)dst + dst_index + 0)); - char4 data_1 = *((__global char4 *)((__global char *)dst + dst_index + 4)); - char4 data_2 = *((__global char4 *)((__global char *)dst + dst_index + 8)); - - char4 tmp_data_0 = src1_data_0 & src2_data_0; - char4 tmp_data_1 = src1_data_1 & src2_data_1; - char4 tmp_data_2 = src1_data_2 & src2_data_2; - - *((__global char4 *)((__global char *)dst + dst_index + 0))= tmp_data_0; - *((__global char4 *)((__global char *)dst + dst_index + 4))= tmp_data_1; - *((__global char4 *)((__global char *)dst + dst_index + 8))= tmp_data_2; - } -} -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_s_bitwise_and_C3_D6 ( - __global short *src1, int src1_step, int src1_offset, - __global short *dst, int dst_step, int dst_offset, - short16 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 24) + src1_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 24)); - - short4 src1_data_0 = *((__global short4 *)((__global char *)src1 + src1_index + 0 )); - short4 src1_data_1 = *((__global short4 *)((__global char *)src1 + src1_index + 8 )); - short4 src1_data_2 = *((__global short4 *)((__global char *)src1 + src1_index + 16)); - - short4 src2_data_0 = (short4)(src2.s0, src2.s1, src2.s2, src2.s3); - short4 src2_data_1 = (short4)(src2.s4, src2.s5, src2.s6, src2.s7); - short4 src2_data_2 = (short4)(src2.s8, src2.s9, src2.sa, src2.sb); - - short4 data_0 = *((__global short4 *)((__global char *)dst + dst_index + 0 )); - short4 data_1 = *((__global short4 *)((__global char *)dst + dst_index + 8 )); - short4 data_2 = *((__global short4 *)((__global char *)dst + dst_index + 16)); - - short4 tmp_data_0 = src1_data_0 & src2_data_0; - short4 tmp_data_1 = src1_data_1 & src2_data_1; - short4 tmp_data_2 = src1_data_2 & src2_data_2; - - *((__global short4 *)((__global char *)dst + dst_index + 0 ))= tmp_data_0; - *((__global short4 *)((__global char *)dst + dst_index + 8 ))= tmp_data_1; - *((__global short4 *)((__global char *)dst + dst_index + 16))= tmp_data_2; - } -} -#endif __kernel void arithm_s_bitwise_and_C4_D0 ( __global uchar *src1, int src1_step, int src1_offset, __global uchar *dst, int dst_step, int dst_offset, diff --git a/modules/ocl/src/opencl/arithm_bitwise_or_mask.cl b/modules/ocl/src/opencl/arithm_bitwise_or_mask.cl index 2523eddcd9..f2cc36e1e0 100644 --- a/modules/ocl/src/opencl/arithm_bitwise_or_mask.cl +++ b/modules/ocl/src/opencl/arithm_bitwise_or_mask.cl @@ -566,396 +566,6 @@ __kernel void arithm_bitwise_or_with_mask_C2_D6 ( #endif -__kernel void arithm_bitwise_or_with_mask_C3_D0 ( - __global uchar *src1, int src1_step, int src1_offset, - __global uchar *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global uchar *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 3 ) & 3) - int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3)); - int src2_index = mad24(y, src2_step, (x * 3) + src2_offset - (dst_align * 3)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3)); - - uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0); - uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4); - uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8); - - uchar4 src2_data_0 = vload4(0, src2 + src2_index + 0); - uchar4 src2_data_1 = vload4(0, src2 + src2_index + 4); - uchar4 src2_data_2 = vload4(0, src2 + src2_index + 8); - - uchar4 mask_data = vload4(0, mask + mask_index); - - uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0)); - uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4)); - uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8)); - - uchar4 tmp_data_0 = src1_data_0 | src2_data_0; - uchar4 tmp_data_1 = src1_data_1 | src2_data_1; - uchar4 tmp_data_2 = src1_data_2 | src2_data_2; - - data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz; - data_0.w = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_0.w : data_0.w; - - data_1.xy = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_1.xy : data_1.xy; - data_1.zw = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.zw : data_1.zw; - - data_2.x = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.x : data_2.x; - data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end)) - ? tmp_data_2.yzw : data_2.yzw; - - *((__global uchar4 *)(dst + dst_index + 0)) = data_0; - *((__global uchar4 *)(dst + dst_index + 4)) = data_1; - *((__global uchar4 *)(dst + dst_index + 8)) = data_2; - } -} - - -__kernel void arithm_bitwise_or_with_mask_C3_D1 ( - __global char *src1, int src1_step, int src1_offset, - __global char *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global char *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 3 ) & 3) - int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3)); - int src2_index = mad24(y, src2_step, (x * 3) + src2_offset - (dst_align * 3)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3)); - - char4 src1_data_0 = vload4(0, src1 + src1_index + 0); - char4 src1_data_1 = vload4(0, src1 + src1_index + 4); - char4 src1_data_2 = vload4(0, src1 + src1_index + 8); - - char4 src2_data_0 = vload4(0, src2 + src2_index + 0); - char4 src2_data_1 = vload4(0, src2 + src2_index + 4); - char4 src2_data_2 = vload4(0, src2 + src2_index + 8); - - uchar4 mask_data = vload4(0, mask + mask_index); - - char4 data_0 = *((__global char4 *)(dst + dst_index + 0)); - char4 data_1 = *((__global char4 *)(dst + dst_index + 4)); - char4 data_2 = *((__global char4 *)(dst + dst_index + 8)); - - char4 tmp_data_0 = src1_data_0 | src2_data_0; - char4 tmp_data_1 = src1_data_1 | src2_data_1; - char4 tmp_data_2 = src1_data_2 | src2_data_2; - - data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz; - data_0.w = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_0.w : data_0.w; - - data_1.xy = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_1.xy : data_1.xy; - data_1.zw = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.zw : data_1.zw; - - data_2.x = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.x : data_2.x; - data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end)) - ? tmp_data_2.yzw : data_2.yzw; - - *((__global char4 *)(dst + dst_index + 0)) = data_0; - *((__global char4 *)(dst + dst_index + 4)) = data_1; - *((__global char4 *)(dst + dst_index + 8)) = data_2; - } -} - -__kernel void arithm_bitwise_or_with_mask_C3_D2 ( - __global ushort *src1, int src1_step, int src1_offset, - __global ushort *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global ushort *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - int src2_index = mad24(y, src2_step, (x * 6) + src2_offset - (dst_align * 6)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0)); - ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4)); - ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8)); - - ushort2 src2_data_0 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 0)); - ushort2 src2_data_1 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 4)); - ushort2 src2_data_2 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 8)); - - uchar2 mask_data = vload2(0, mask + mask_index); - - ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0)); - ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4)); - ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8)); - - ushort2 tmp_data_0 = src1_data_0 | src2_data_0 ; - ushort2 tmp_data_1 = src1_data_1 | src2_data_1 ; - ushort2 tmp_data_2 = src1_data_2 | src2_data_2 ; - - data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_bitwise_or_with_mask_C3_D3 ( - __global short *src1, int src1_step, int src1_offset, - __global short *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global short *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - int src2_index = mad24(y, src2_step, (x * 6) + src2_offset - (dst_align * 6)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0)); - short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4)); - short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8)); - - short2 src2_data_0 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 0)); - short2 src2_data_1 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 4)); - short2 src2_data_2 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 8)); - - uchar2 mask_data = vload2(0, mask + mask_index); - - short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0)); - short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4)); - short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8)); - - short2 tmp_data_0 = src1_data_0 | src2_data_0 ; - short2 tmp_data_1 = src1_data_1 | src2_data_1 ; - short2 tmp_data_2 = src1_data_2 | src2_data_2 ; - - data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_bitwise_or_with_mask_C3_D4 ( - __global int *src1, int src1_step, int src1_offset, - __global int *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global int *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int src2_index = mad24(y, src2_step, (x * 12) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0)); - int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4)); - int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8)); - - int src2_data_0 = *((__global int *)((__global char *)src2 + src2_index + 0)); - int src2_data_1 = *((__global int *)((__global char *)src2 + src2_index + 4)); - int src2_data_2 = *((__global int *)((__global char *)src2 + src2_index + 8)); - - uchar mask_data = * (mask + mask_index); - - int data_0 = *((__global int *)((__global char *)dst + dst_index + 0)); - int data_1 = *((__global int *)((__global char *)dst + dst_index + 4)); - int data_2 = *((__global int *)((__global char *)dst + dst_index + 8)); - - int tmp_data_0 = src1_data_0 | src2_data_0 ; - int tmp_data_1 = src1_data_1 | src2_data_1 ; - int tmp_data_2 = src1_data_2 | src2_data_2 ; - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global int *)((__global char *)dst + dst_index + 0))= data_0; - *((__global int *)((__global char *)dst + dst_index + 4))= data_1; - *((__global int *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_bitwise_or_with_mask_C3_D5 ( - __global char *src1, int src1_step, int src1_offset, - __global char *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global char *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int src2_index = mad24(y, src2_step, (x * 12) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - char4 src1_data_0 = *((__global char4 *)((__global char *)src1 + src1_index + 0)); - char4 src1_data_1 = *((__global char4 *)((__global char *)src1 + src1_index + 4)); - char4 src1_data_2 = *((__global char4 *)((__global char *)src1 + src1_index + 8)); - - char4 src2_data_0 = *((__global char4 *)((__global char *)src2 + src2_index + 0)); - char4 src2_data_1 = *((__global char4 *)((__global char *)src2 + src2_index + 4)); - char4 src2_data_2 = *((__global char4 *)((__global char *)src2 + src2_index + 8)); - - uchar mask_data = * (mask + mask_index); - - char4 data_0 = *((__global char4 *)((__global char *)dst + dst_index + 0)); - char4 data_1 = *((__global char4 *)((__global char *)dst + dst_index + 4)); - char4 data_2 = *((__global char4 *)((__global char *)dst + dst_index + 8)); - - char4 tmp_data_0 = src1_data_0 | src2_data_0; - char4 tmp_data_1 = src1_data_1 | src2_data_1; - char4 tmp_data_2 = src1_data_2 | src2_data_2; - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global char4 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global char4 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global char4 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_bitwise_or_with_mask_C3_D6 ( - __global char *src1, int src1_step, int src1_offset, - __global char *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global char *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 24) + src1_offset); - int src2_index = mad24(y, src2_step, (x * 24) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 24)); - - char8 src1_data_0 = *((__global char8 *)((__global char *)src1 + src1_index + 0 )); - char8 src1_data_1 = *((__global char8 *)((__global char *)src1 + src1_index + 8 )); - char8 src1_data_2 = *((__global char8 *)((__global char *)src1 + src1_index + 16)); - - char8 src2_data_0 = *((__global char8 *)((__global char *)src2 + src2_index + 0 )); - char8 src2_data_1 = *((__global char8 *)((__global char *)src2 + src2_index + 8 )); - char8 src2_data_2 = *((__global char8 *)((__global char *)src2 + src2_index + 16)); - - uchar mask_data = * (mask + mask_index); - - char8 data_0 = *((__global char8 *)((__global char *)dst + dst_index + 0 )); - char8 data_1 = *((__global char8 *)((__global char *)dst + dst_index + 8 )); - char8 data_2 = *((__global char8 *)((__global char *)dst + dst_index + 16)); - - char8 tmp_data_0 = src1_data_0 | src2_data_0; - char8 tmp_data_1 = src1_data_1 | src2_data_1; - char8 tmp_data_2 = src1_data_2 | src2_data_2; - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global char8 *)((__global char *)dst + dst_index + 0 ))= data_0; - *((__global char8 *)((__global char *)dst + dst_index + 8 ))= data_1; - *((__global char8 *)((__global char *)dst + dst_index + 16))= data_2; - } -} -#endif - - __kernel void arithm_bitwise_or_with_mask_C4_D0 ( __global uchar *src1, int src1_step, int src1_offset, __global uchar *src2, int src2_step, int src2_offset, diff --git a/modules/ocl/src/opencl/arithm_bitwise_or_scalar.cl b/modules/ocl/src/opencl/arithm_bitwise_or_scalar.cl index fdcc00c4ef..7ade345f9b 100644 --- a/modules/ocl/src/opencl/arithm_bitwise_or_scalar.cl +++ b/modules/ocl/src/opencl/arithm_bitwise_or_scalar.cl @@ -457,344 +457,7 @@ __kernel void arithm_s_bitwise_or_C2_D6 ( } } #endif -__kernel void arithm_s_bitwise_or_C3_D0 ( - __global uchar *src1, int src1_step, int src1_offset, - __global uchar *dst, int dst_step, int dst_offset, - uchar4 src2, int rows, int cols, int dst_step1) - -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 3 ) & 3) - int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3)); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3)); - - uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0); - uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4); - uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8); - - uchar4 src2_data_0 = (uchar4)(src2.x, src2.y, src2.z, src2.x); - uchar4 src2_data_1 = (uchar4)(src2.y, src2.z, src2.x, src2.y); - uchar4 src2_data_2 = (uchar4)(src2.z, src2.x, src2.y, src2.z); - - uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0)); - uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4)); - uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8)); - - uchar4 tmp_data_0 = src1_data_0 | src2_data_0 ; - uchar4 tmp_data_1 = src1_data_1 | src2_data_1 ; - uchar4 tmp_data_2 = src1_data_2 | src2_data_2 ; - - data_0.xyz = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz; - data_0.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_0.w : data_0.w; - - data_1.xy = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_1.xy : data_1.xy; - data_1.zw = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.zw : data_1.zw; - - data_2.x = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.x : data_2.x; - data_2.yzw = ((dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end)) - ? tmp_data_2.yzw : data_2.yzw; - - *((__global uchar4 *)(dst + dst_index + 0)) = data_0; - *((__global uchar4 *)(dst + dst_index + 4)) = data_1; - *((__global uchar4 *)(dst + dst_index + 8)) = data_2; - } -} - - -__kernel void arithm_s_bitwise_or_C3_D1 ( - __global char *src1, int src1_step, int src1_offset, - __global char *dst, int dst_step, int dst_offset, - char4 src2, int rows, int cols, int dst_step1) - -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 3 ) & 3) - int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3)); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3)); - - char4 src1_data_0 = vload4(0, src1 + src1_index + 0); - char4 src1_data_1 = vload4(0, src1 + src1_index + 4); - char4 src1_data_2 = vload4(0, src1 + src1_index + 8); - - char4 src2_data_0 = (char4)(src2.x, src2.y, src2.z, src2.x); - char4 src2_data_1 = (char4)(src2.y, src2.z, src2.x, src2.y); - char4 src2_data_2 = (char4)(src2.z, src2.x, src2.y, src2.z); - - char4 data_0 = *((__global char4 *)(dst + dst_index + 0)); - char4 data_1 = *((__global char4 *)(dst + dst_index + 4)); - char4 data_2 = *((__global char4 *)(dst + dst_index + 8)); - - char4 tmp_data_0 = src1_data_0 | src2_data_0; - char4 tmp_data_1 = src1_data_1 | src2_data_1; - char4 tmp_data_2 = src1_data_2 | src2_data_2; - - data_0.xyz = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz; - data_0.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_0.w : data_0.w; - - data_1.xy = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_1.xy : data_1.xy; - data_1.zw = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.zw : data_1.zw; - - data_2.x = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.x : data_2.x; - data_2.yzw = ((dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end)) - ? tmp_data_2.yzw : data_2.yzw; - - *((__global char4 *)(dst + dst_index + 0)) = data_0; - *((__global char4 *)(dst + dst_index + 4)) = data_1; - *((__global char4 *)(dst + dst_index + 8)) = data_2; - } -} - -__kernel void arithm_s_bitwise_or_C3_D2 ( - __global ushort *src1, int src1_step, int src1_offset, - __global ushort *dst, int dst_step, int dst_offset, - ushort4 src2, int rows, int cols, int dst_step1) - -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0)); - ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4)); - ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8)); - - ushort2 src2_data_0 = (ushort2)(src2.x, src2.y); - ushort2 src2_data_1 = (ushort2)(src2.z, src2.x); - ushort2 src2_data_2 = (ushort2)(src2.y, src2.z); - - ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0)); - ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4)); - ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8)); - - ushort2 tmp_data_0 = src1_data_0 | src2_data_0 ; - ushort2 tmp_data_1 = src1_data_1 | src2_data_1 ; - ushort2 tmp_data_2 = src1_data_2 | src2_data_2 ; - - data_0.xy = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_s_bitwise_or_C3_D3 ( - __global short *src1, int src1_step, int src1_offset, - __global short *dst, int dst_step, int dst_offset, - short4 src2, int rows, int cols, int dst_step1) - -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0)); - short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4)); - short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8)); - - short2 src2_data_0 = (short2)(src2.x, src2.y); - short2 src2_data_1 = (short2)(src2.z, src2.x); - short2 src2_data_2 = (short2)(src2.y, src2.z); - - short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0)); - short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4)); - short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8)); - - short2 tmp_data_0 = src1_data_0 | src2_data_0 ; - short2 tmp_data_1 = src1_data_1 | src2_data_1 ; - short2 tmp_data_2 = src1_data_2 | src2_data_2 ; - - data_0.xy = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_s_bitwise_or_C3_D4 ( - __global int *src1, int src1_step, int src1_offset, - __global int *dst, int dst_step, int dst_offset, - int4 src2, int rows, int cols, int dst_step1) - -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0)); - int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4)); - int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8)); - - int src2_data_0 = src2.x; - int src2_data_1 = src2.y; - int src2_data_2 = src2.z; - - int data_0 = *((__global int *)((__global char *)dst + dst_index + 0)); - int data_1 = *((__global int *)((__global char *)dst + dst_index + 4)); - int data_2 = *((__global int *)((__global char *)dst + dst_index + 8)); - - int tmp_data_0 = src1_data_0 | src2_data_0; - int tmp_data_1 = src1_data_1 | src2_data_1; - int tmp_data_2 = src1_data_2 | src2_data_2; - - *((__global int *)((__global char *)dst + dst_index + 0))= tmp_data_0; - *((__global int *)((__global char *)dst + dst_index + 4))= tmp_data_1; - *((__global int *)((__global char *)dst + dst_index + 8))= tmp_data_2; - } -} -__kernel void arithm_s_bitwise_or_C3_D5 ( - __global char *src1, int src1_step, int src1_offset, - __global char *dst, int dst_step, int dst_offset, - char16 src2, int rows, int cols, int dst_step1) - -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - char4 src1_data_0 = *((__global char4 *)((__global char *)src1 + src1_index + 0)); - char4 src1_data_1 = *((__global char4 *)((__global char *)src1 + src1_index + 4)); - char4 src1_data_2 = *((__global char4 *)((__global char *)src1 + src1_index + 8)); - - char4 src2_data_0 = (char4)(src2.s0, src2.s1, src2.s2, src2.s3); - char4 src2_data_1 = (char4)(src2.s4, src2.s5, src2.s6, src2.s7); - char4 src2_data_2 = (char4)(src2.s8, src2.s9, src2.sA, src2.sB); - - char4 tmp_data_0 = src1_data_0 | src2_data_0; - char4 tmp_data_1 = src1_data_1 | src2_data_1; - char4 tmp_data_2 = src1_data_2 | src2_data_2; - - *((__global char4 *)((__global char *)dst + dst_index + 0))= tmp_data_0; - *((__global char4 *)((__global char *)dst + dst_index + 4))= tmp_data_1; - *((__global char4 *)((__global char *)dst + dst_index + 8))= tmp_data_2; - } -} -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_s_bitwise_or_C3_D6 ( - __global short *src1, int src1_step, int src1_offset, - __global short *dst, int dst_step, int dst_offset, - short16 src2, int rows, int cols, int dst_step1) - -{ - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 24) + src1_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 24)); - - short4 src1_data_0 = *((__global short4 *)((__global char *)src1 + src1_index + 0 )); - short4 src1_data_1 = *((__global short4 *)((__global char *)src1 + src1_index + 8 )); - short4 src1_data_2 = *((__global short4 *)((__global char *)src1 + src1_index + 16)); - - short4 src2_data_0 = (short4)(src2.s0, src2.s1, src2.s2, src2.s3); - short4 src2_data_1 = (short4)(src2.s4, src2.s5, src2.s6, src2.s7); - short4 src2_data_2 = (short4)(src2.s8, src2.s9, src2.sa, src2.sb); - - short4 data_0 = *((__global short4 *)((__global char *)dst + dst_index + 0 )); - short4 data_1 = *((__global short4 *)((__global char *)dst + dst_index + 8 )); - short4 data_2 = *((__global short4 *)((__global char *)dst + dst_index + 16)); - - short4 tmp_data_0 = src1_data_0 | src2_data_0; - short4 tmp_data_1 = src1_data_1 | src2_data_1; - short4 tmp_data_2 = src1_data_2 | src2_data_2; - - *((__global short4 *)((__global char *)dst + dst_index + 0 ))= tmp_data_0; - *((__global short4 *)((__global char *)dst + dst_index + 8 ))= tmp_data_1; - *((__global short4 *)((__global char *)dst + dst_index + 16))= tmp_data_2; - } -} -#endif __kernel void arithm_s_bitwise_or_C4_D0 ( __global uchar *src1, int src1_step, int src1_offset, __global uchar *dst, int dst_step, int dst_offset, diff --git a/modules/ocl/src/opencl/arithm_bitwise_or_scalar_mask.cl b/modules/ocl/src/opencl/arithm_bitwise_or_scalar_mask.cl index 8baa9a2ca2..b8f07a8724 100644 --- a/modules/ocl/src/opencl/arithm_bitwise_or_scalar_mask.cl +++ b/modules/ocl/src/opencl/arithm_bitwise_or_scalar_mask.cl @@ -533,387 +533,7 @@ __kernel void arithm_s_bitwise_or_with_mask_C2_D6 ( } } #endif -__kernel void arithm_s_bitwise_or_with_mask_C3_D0 ( - __global uchar *src1, int src1_step, int src1_offset, - __global uchar *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - uchar4 src2, int rows, int cols, int dst_step1) - -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 3 ) & 3) - int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3)); - - uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0); - uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4); - uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8); - - uchar4 src2_data_0 = (uchar4)(src2.x, src2.y, src2.z, src2.x); - uchar4 src2_data_1 = (uchar4)(src2.y, src2.z, src2.x, src2.y); - uchar4 src2_data_2 = (uchar4)(src2.z, src2.x, src2.y, src2.z); - - uchar4 mask_data = vload4(0, mask + mask_index); - - uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0)); - uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4)); - uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8)); - - uchar4 tmp_data_0 = src1_data_0 | src2_data_0; - uchar4 tmp_data_1 = src1_data_1 | src2_data_1; - uchar4 tmp_data_2 = src1_data_2 | src2_data_2; - - data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz; - data_0.w = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_0.w : data_0.w; - - data_1.xy = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_1.xy : data_1.xy; - data_1.zw = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.zw : data_1.zw; - - data_2.x = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.x : data_2.x; - data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end)) - ? tmp_data_2.yzw : data_2.yzw; - - *((__global uchar4 *)(dst + dst_index + 0)) = data_0; - *((__global uchar4 *)(dst + dst_index + 4)) = data_1; - *((__global uchar4 *)(dst + dst_index + 8)) = data_2; - } -} - - -__kernel void arithm_s_bitwise_or_with_mask_C3_D1 ( - __global char *src1, int src1_step, int src1_offset, - __global char *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - char4 src2, int rows, int cols, int dst_step1) - -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 3 ) & 3) - int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3)); - - char4 src1_data_0 = vload4(0, src1 + src1_index + 0); - char4 src1_data_1 = vload4(0, src1 + src1_index + 4); - char4 src1_data_2 = vload4(0, src1 + src1_index + 8); - - char4 src2_data_0 = (char4)(src2.x, src2.y, src2.z, src2.x); - char4 src2_data_1 = (char4)(src2.y, src2.z, src2.x, src2.y); - char4 src2_data_2 = (char4)(src2.z, src2.x, src2.y, src2.z); - - uchar4 mask_data = vload4(0, mask + mask_index); - - char4 data_0 = *((__global char4 *)(dst + dst_index + 0)); - char4 data_1 = *((__global char4 *)(dst + dst_index + 4)); - char4 data_2 = *((__global char4 *)(dst + dst_index + 8)); - - char4 tmp_data_0 = src1_data_0 | src2_data_0; - char4 tmp_data_1 = src1_data_1 | src2_data_1; - char4 tmp_data_2 = src1_data_2 | src2_data_2; - - data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz; - data_0.w = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_0.w : data_0.w; - - data_1.xy = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_1.xy : data_1.xy; - data_1.zw = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.zw : data_1.zw; - - data_2.x = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.x : data_2.x; - data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end)) - ? tmp_data_2.yzw : data_2.yzw; - - *((__global char4 *)(dst + dst_index + 0)) = data_0; - *((__global char4 *)(dst + dst_index + 4)) = data_1; - *((__global char4 *)(dst + dst_index + 8)) = data_2; - } -} - -__kernel void arithm_s_bitwise_or_with_mask_C3_D2 ( - __global ushort *src1, int src1_step, int src1_offset, - __global ushort *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - ushort4 src2, int rows, int cols, int dst_step1) - -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0)); - ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4)); - ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8)); - - ushort2 src2_data_0 = (ushort2)(src2.x, src2.y); - ushort2 src2_data_1 = (ushort2)(src2.z, src2.x); - ushort2 src2_data_2 = (ushort2)(src2.y, src2.z); - - uchar2 mask_data = vload2(0, mask + mask_index); - - ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0)); - ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4)); - ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8)); - - ushort2 tmp_data_0 = src1_data_0 | src2_data_0; - ushort2 tmp_data_1 = src1_data_1 | src2_data_1; - ushort2 tmp_data_2 = src1_data_2 | src2_data_2; - - data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_s_bitwise_or_with_mask_C3_D3 ( - __global short *src1, int src1_step, int src1_offset, - __global short *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - short4 src2, int rows, int cols, int dst_step1) - -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0)); - short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4)); - short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8)); - - short2 src2_data_0 = (short2)(src2.x, src2.y); - short2 src2_data_1 = (short2)(src2.z, src2.x); - short2 src2_data_2 = (short2)(src2.y, src2.z); - - uchar2 mask_data = vload2(0, mask + mask_index); - - short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0)); - short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4)); - short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8)); - - short2 tmp_data_0 = src1_data_0 | src2_data_0; - short2 tmp_data_1 = src1_data_1 | src2_data_1; - short2 tmp_data_2 = src1_data_2 | src2_data_2; - - data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_s_bitwise_or_with_mask_C3_D4 ( - __global int *src1, int src1_step, int src1_offset, - __global int *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - int4 src2, int rows, int cols, int dst_step1) - -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0)); - int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4)); - int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8)); - - int src2_data_0 = src2.x; - int src2_data_1 = src2.y; - int src2_data_2 = src2.z; - - uchar mask_data = * (mask + mask_index); - - int data_0 = *((__global int *)((__global char *)dst + dst_index + 0)); - int data_1 = *((__global int *)((__global char *)dst + dst_index + 4)); - int data_2 = *((__global int *)((__global char *)dst + dst_index + 8)); - - int tmp_data_0 = src1_data_0 | src2_data_0; - int tmp_data_1 = src1_data_1 | src2_data_1; - int tmp_data_2 = src1_data_2 | src2_data_2; - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global int *)((__global char *)dst + dst_index + 0))= data_0; - *((__global int *)((__global char *)dst + dst_index + 4))= data_1; - *((__global int *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_s_bitwise_or_with_mask_C3_D5 ( - __global char *src1, int src1_step, int src1_offset, - __global char *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - char16 src2, int rows, int cols, int dst_step1) - -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - char4 src1_data_0 = *((__global char4 *)((__global char *)src1 + src1_index + 0)); - char4 src1_data_1 = *((__global char4 *)((__global char *)src1 + src1_index + 4)); - char4 src1_data_2 = *((__global char4 *)((__global char *)src1 + src1_index + 8)); - - char4 src2_data_0 = (char4)(src2.s0, src2.s1, src2.s2, src2.s3); - char4 src2_data_1 = (char4)(src2.s4, src2.s5, src2.s6, src2.s7); - char4 src2_data_2 = (char4)(src2.s8, src2.s9, src2.sA, src2.sB); - - uchar mask_data = * (mask + mask_index); - - char4 data_0 = *((__global char4 *)((__global char *)dst + dst_index + 0)); - char4 data_1 = *((__global char4 *)((__global char *)dst + dst_index + 4)); - char4 data_2 = *((__global char4 *)((__global char *)dst + dst_index + 8)); - - char4 tmp_data_0 = src1_data_0 | src2_data_0; - char4 tmp_data_1 = src1_data_1 | src2_data_1; - char4 tmp_data_2 = src1_data_2 | src2_data_2; - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global char4 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global char4 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global char4 *)((__global char *)dst + dst_index + 8))= data_2; - - } -} -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_s_bitwise_or_with_mask_C3_D6 ( - __global short *src1, int src1_step, int src1_offset, - __global short *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - short16 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 24) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 24)); - - short4 src1_data_0 = *((__global short4 *)((__global char *)src1 + src1_index + 0 )); - short4 src1_data_1 = *((__global short4 *)((__global char *)src1 + src1_index + 8 )); - short4 src1_data_2 = *((__global short4 *)((__global char *)src1 + src1_index + 16)); - - short4 src2_data_0 = (short4)(src2.s0, src2.s1, src2.s2, src2.s3); - short4 src2_data_1 = (short4)(src2.s4, src2.s5, src2.s6, src2.s7); - short4 src2_data_2 = (short4)(src2.s8, src2.s9, src2.sa, src2.sb); - - uchar mask_data = * (mask + mask_index); - - short4 data_0 = *((__global short4 *)((__global char *)dst + dst_index + 0 )); - short4 data_1 = *((__global short4 *)((__global char *)dst + dst_index + 8 )); - short4 data_2 = *((__global short4 *)((__global char *)dst + dst_index + 16)); - - short4 tmp_data_0 = src1_data_0 | src2_data_0; - short4 tmp_data_1 = src1_data_1 | src2_data_1; - short4 tmp_data_2 = src1_data_2 | src2_data_2; - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global short4 *)((__global char *)dst + dst_index + 0 ))= data_0; - *((__global short4 *)((__global char *)dst + dst_index + 8 ))= data_1; - *((__global short4 *)((__global char *)dst + dst_index + 16))= data_2; - } -} -#endif __kernel void arithm_s_bitwise_or_with_mask_C4_D0 ( __global uchar *src1, int src1_step, int src1_offset, __global uchar *dst, int dst_step, int dst_offset, diff --git a/modules/ocl/src/opencl/arithm_bitwise_xor_mask.cl b/modules/ocl/src/opencl/arithm_bitwise_xor_mask.cl index 48bd3e444a..7655be3a80 100644 --- a/modules/ocl/src/opencl/arithm_bitwise_xor_mask.cl +++ b/modules/ocl/src/opencl/arithm_bitwise_xor_mask.cl @@ -565,397 +565,6 @@ __kernel void arithm_bitwise_xor_with_mask_C2_D6 ( } #endif - -__kernel void arithm_bitwise_xor_with_mask_C3_D0 ( - __global uchar *src1, int src1_step, int src1_offset, - __global uchar *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global uchar *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 3 ) & 3) - int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3)); - int src2_index = mad24(y, src2_step, (x * 3) + src2_offset - (dst_align * 3)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3)); - - uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0); - uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4); - uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8); - - uchar4 src2_data_0 = vload4(0, src2 + src2_index + 0); - uchar4 src2_data_1 = vload4(0, src2 + src2_index + 4); - uchar4 src2_data_2 = vload4(0, src2 + src2_index + 8); - - uchar4 mask_data = vload4(0, mask + mask_index); - - uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0)); - uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4)); - uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8)); - - uchar4 tmp_data_0 = src1_data_0 ^ src2_data_0; - uchar4 tmp_data_1 = src1_data_1 ^ src2_data_1; - uchar4 tmp_data_2 = src1_data_2 ^ src2_data_2; - - data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz; - data_0.w = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_0.w : data_0.w; - - data_1.xy = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_1.xy : data_1.xy; - data_1.zw = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.zw : data_1.zw; - - data_2.x = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.x : data_2.x; - data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end)) - ? tmp_data_2.yzw : data_2.yzw; - - *((__global uchar4 *)(dst + dst_index + 0)) = data_0; - *((__global uchar4 *)(dst + dst_index + 4)) = data_1; - *((__global uchar4 *)(dst + dst_index + 8)) = data_2; - } -} - - -__kernel void arithm_bitwise_xor_with_mask_C3_D1 ( - __global char *src1, int src1_step, int src1_offset, - __global char *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global char *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 3 ) & 3) - int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3)); - int src2_index = mad24(y, src2_step, (x * 3) + src2_offset - (dst_align * 3)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3)); - - char4 src1_data_0 = vload4(0, src1 + src1_index + 0); - char4 src1_data_1 = vload4(0, src1 + src1_index + 4); - char4 src1_data_2 = vload4(0, src1 + src1_index + 8); - - char4 src2_data_0 = vload4(0, src2 + src2_index + 0); - char4 src2_data_1 = vload4(0, src2 + src2_index + 4); - char4 src2_data_2 = vload4(0, src2 + src2_index + 8); - - uchar4 mask_data = vload4(0, mask + mask_index); - - char4 data_0 = *((__global char4 *)(dst + dst_index + 0)); - char4 data_1 = *((__global char4 *)(dst + dst_index + 4)); - char4 data_2 = *((__global char4 *)(dst + dst_index + 8)); - - char4 tmp_data_0 = src1_data_0 ^ src2_data_0; - char4 tmp_data_1 = src1_data_1 ^ src2_data_1; - char4 tmp_data_2 = src1_data_2 ^ src2_data_2; - - data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz; - data_0.w = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_0.w : data_0.w; - - data_1.xy = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_1.xy : data_1.xy; - data_1.zw = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.zw : data_1.zw; - - data_2.x = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.x : data_2.x; - data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end)) - ? tmp_data_2.yzw : data_2.yzw; - - *((__global char4 *)(dst + dst_index + 0)) = data_0; - *((__global char4 *)(dst + dst_index + 4)) = data_1; - *((__global char4 *)(dst + dst_index + 8)) = data_2; - } -} - -__kernel void arithm_bitwise_xor_with_mask_C3_D2 ( - __global ushort *src1, int src1_step, int src1_offset, - __global ushort *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global ushort *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - int src2_index = mad24(y, src2_step, (x * 6) + src2_offset - (dst_align * 6)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0)); - ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4)); - ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8)); - - ushort2 src2_data_0 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 0)); - ushort2 src2_data_1 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 4)); - ushort2 src2_data_2 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 8)); - - uchar2 mask_data = vload2(0, mask + mask_index); - - ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0)); - ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4)); - ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8)); - - ushort2 tmp_data_0 = src1_data_0 ^ src2_data_0 ; - ushort2 tmp_data_1 = src1_data_1 ^ src2_data_1 ; - ushort2 tmp_data_2 = src1_data_2 ^ src2_data_2 ; - - data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_bitwise_xor_with_mask_C3_D3 ( - __global short *src1, int src1_step, int src1_offset, - __global short *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global short *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - int src2_index = mad24(y, src2_step, (x * 6) + src2_offset - (dst_align * 6)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0)); - short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4)); - short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8)); - - short2 src2_data_0 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 0)); - short2 src2_data_1 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 4)); - short2 src2_data_2 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 8)); - - uchar2 mask_data = vload2(0, mask + mask_index); - - short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0)); - short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4)); - short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8)); - - short2 tmp_data_0 = src1_data_0 ^ src2_data_0 ; - short2 tmp_data_1 = src1_data_1 ^ src2_data_1 ; - short2 tmp_data_2 = src1_data_2 ^ src2_data_2 ; - - data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_bitwise_xor_with_mask_C3_D4 ( - __global int *src1, int src1_step, int src1_offset, - __global int *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global int *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int src2_index = mad24(y, src2_step, (x * 12) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0)); - int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4)); - int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8)); - - int src2_data_0 = *((__global int *)((__global char *)src2 + src2_index + 0)); - int src2_data_1 = *((__global int *)((__global char *)src2 + src2_index + 4)); - int src2_data_2 = *((__global int *)((__global char *)src2 + src2_index + 8)); - - uchar mask_data = * (mask + mask_index); - - int data_0 = *((__global int *)((__global char *)dst + dst_index + 0)); - int data_1 = *((__global int *)((__global char *)dst + dst_index + 4)); - int data_2 = *((__global int *)((__global char *)dst + dst_index + 8)); - - int tmp_data_0 = src1_data_0 ^ src2_data_0 ; - int tmp_data_1 = src1_data_1 ^ src2_data_1 ; - int tmp_data_2 = src1_data_2 ^ src2_data_2 ; - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global int *)((__global char *)dst + dst_index + 0))= data_0; - *((__global int *)((__global char *)dst + dst_index + 4))= data_1; - *((__global int *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_bitwise_xor_with_mask_C3_D5 ( - __global char *src1, int src1_step, int src1_offset, - __global char *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global char *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int src2_index = mad24(y, src2_step, (x * 12) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - char4 src1_data_0 = *((__global char4 *)((__global char *)src1 + src1_index + 0)); - char4 src1_data_1 = *((__global char4 *)((__global char *)src1 + src1_index + 4)); - char4 src1_data_2 = *((__global char4 *)((__global char *)src1 + src1_index + 8)); - - char4 src2_data_0 = *((__global char4 *)((__global char *)src2 + src2_index + 0)); - char4 src2_data_1 = *((__global char4 *)((__global char *)src2 + src2_index + 4)); - char4 src2_data_2 = *((__global char4 *)((__global char *)src2 + src2_index + 8)); - - uchar mask_data = * (mask + mask_index); - - char4 data_0 = *((__global char4 *)((__global char *)dst + dst_index + 0)); - char4 data_1 = *((__global char4 *)((__global char *)dst + dst_index + 4)); - char4 data_2 = *((__global char4 *)((__global char *)dst + dst_index + 8)); - - char4 tmp_data_0 = src1_data_0 ^ src2_data_0; - char4 tmp_data_1 = src1_data_1 ^ src2_data_1; - char4 tmp_data_2 = src1_data_2 ^ src2_data_2; - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global char4 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global char4 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global char4 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_bitwise_xor_with_mask_C3_D6 ( - __global char *src1, int src1_step, int src1_offset, - __global char *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global char *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 24) + src1_offset); - int src2_index = mad24(y, src2_step, (x * 24) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 24)); - - char8 src1_data_0 = *((__global char8 *)((__global char *)src1 + src1_index + 0 )); - char8 src1_data_1 = *((__global char8 *)((__global char *)src1 + src1_index + 8 )); - char8 src1_data_2 = *((__global char8 *)((__global char *)src1 + src1_index + 16)); - - char8 src2_data_0 = *((__global char8 *)((__global char *)src2 + src2_index + 0 )); - char8 src2_data_1 = *((__global char8 *)((__global char *)src2 + src2_index + 8 )); - char8 src2_data_2 = *((__global char8 *)((__global char *)src2 + src2_index + 16)); - - uchar mask_data = * (mask + mask_index); - - char8 data_0 = *((__global char8 *)((__global char *)dst + dst_index + 0 )); - char8 data_1 = *((__global char8 *)((__global char *)dst + dst_index + 8 )); - char8 data_2 = *((__global char8 *)((__global char *)dst + dst_index + 16)); - - char8 tmp_data_0 = src1_data_0 ^ src2_data_0; - char8 tmp_data_1 = src1_data_1 ^ src2_data_1; - char8 tmp_data_2 = src1_data_2 ^ src2_data_2; - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global char8 *)((__global char *)dst + dst_index + 0 ))= data_0; - *((__global char8 *)((__global char *)dst + dst_index + 8 ))= data_1; - *((__global char8 *)((__global char *)dst + dst_index + 16))= data_2; - } -} -#endif - - __kernel void arithm_bitwise_xor_with_mask_C4_D0 ( __global uchar *src1, int src1_step, int src1_offset, __global uchar *src2, int src2_step, int src2_offset, diff --git a/modules/ocl/src/opencl/arithm_bitwise_xor_scalar.cl b/modules/ocl/src/opencl/arithm_bitwise_xor_scalar.cl index 2c6dd50cd4..73b5687c11 100644 --- a/modules/ocl/src/opencl/arithm_bitwise_xor_scalar.cl +++ b/modules/ocl/src/opencl/arithm_bitwise_xor_scalar.cl @@ -461,340 +461,7 @@ __kernel void arithm_s_bitwise_xor_C2_D6 ( } } #endif -__kernel void arithm_s_bitwise_xor_C3_D0 ( - __global uchar *src1, int src1_step, int src1_offset, - __global uchar *dst, int dst_step, int dst_offset, - uchar4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 3 ) & 3) - int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3)); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3)); - - uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0); - uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4); - uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8); - - uchar4 src2_data_0 = (uchar4)(src2.x, src2.y, src2.z, src2.x); - uchar4 src2_data_1 = (uchar4)(src2.y, src2.z, src2.x, src2.y); - uchar4 src2_data_2 = (uchar4)(src2.z, src2.x, src2.y, src2.z); - - uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0)); - uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4)); - uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8)); - - uchar4 tmp_data_0 = src1_data_0 ^ src2_data_0; - uchar4 tmp_data_1 = src1_data_1 ^ src2_data_1; - uchar4 tmp_data_2 = src1_data_2 ^ src2_data_2; - - data_0.xyz = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz; - data_0.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_0.w : data_0.w; - - data_1.xy = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_1.xy : data_1.xy; - data_1.zw = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.zw : data_1.zw; - - data_2.x = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.x : data_2.x; - data_2.yzw = ((dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end)) - ? tmp_data_2.yzw : data_2.yzw; - - *((__global uchar4 *)(dst + dst_index + 0)) = data_0; - *((__global uchar4 *)(dst + dst_index + 4)) = data_1; - *((__global uchar4 *)(dst + dst_index + 8)) = data_2; - } -} - - -__kernel void arithm_s_bitwise_xor_C3_D1 ( - __global char *src1, int src1_step, int src1_offset, - __global char *dst, int dst_step, int dst_offset, - char4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 3 ) & 3) - int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3)); - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3)); - - char4 src1_data_0 = vload4(0, src1 + src1_index + 0); - char4 src1_data_1 = vload4(0, src1 + src1_index + 4); - char4 src1_data_2 = vload4(0, src1 + src1_index + 8); - - char4 src2_data_0 = (char4)(src2.x, src2.y, src2.z, src2.x); - char4 src2_data_1 = (char4)(src2.y, src2.z, src2.x, src2.y); - char4 src2_data_2 = (char4)(src2.z, src2.x, src2.y, src2.z); - - char4 data_0 = *((__global char4 *)(dst + dst_index + 0)); - char4 data_1 = *((__global char4 *)(dst + dst_index + 4)); - char4 data_2 = *((__global char4 *)(dst + dst_index + 8)); - - char4 tmp_data_0 = src1_data_0 ^ src2_data_0; - char4 tmp_data_1 = src1_data_1 ^ src2_data_1; - char4 tmp_data_2 = src1_data_2 ^ src2_data_2; - - data_0.xyz = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz; - data_0.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_0.w : data_0.w; - - data_1.xy = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_1.xy : data_1.xy; - data_1.zw = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.zw : data_1.zw; - - data_2.x = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.x : data_2.x; - data_2.yzw = ((dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end)) - ? tmp_data_2.yzw : data_2.yzw; - - *((__global char4 *)(dst + dst_index + 0)) = data_0; - *((__global char4 *)(dst + dst_index + 4)) = data_1; - *((__global char4 *)(dst + dst_index + 8)) = data_2; - } -} - -__kernel void arithm_s_bitwise_xor_C3_D2 ( - __global ushort *src1, int src1_step, int src1_offset, - __global ushort *dst, int dst_step, int dst_offset, - ushort4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0)); - ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4)); - ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8)); - - ushort2 src2_data_0 = (ushort2)(src2.x, src2.y); - ushort2 src2_data_1 = (ushort2)(src2.z, src2.x); - ushort2 src2_data_2 = (ushort2)(src2.y, src2.z); - - ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0)); - ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4)); - ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8)); - - ushort2 tmp_data_0 = src1_data_0 ^ src2_data_0; - ushort2 tmp_data_1 = src1_data_1 ^ src2_data_1; - ushort2 tmp_data_2 = src1_data_2 ^ src2_data_2; - - data_0.xy = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_s_bitwise_xor_C3_D3 ( - __global short *src1, int src1_step, int src1_offset, - __global short *dst, int dst_step, int dst_offset, - short4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0)); - short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4)); - short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8)); - - short2 src2_data_0 = (short2)(src2.x, src2.y); - short2 src2_data_1 = (short2)(src2.z, src2.x); - short2 src2_data_2 = (short2)(src2.y, src2.z); - - short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0)); - short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4)); - short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8)); - - short2 tmp_data_0 = src1_data_0 ^ src2_data_0; - short2 tmp_data_1 = src1_data_1 ^ src2_data_1; - short2 tmp_data_2 = src1_data_2 ^ src2_data_2; - - data_0.xy = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_s_bitwise_xor_C3_D4 ( - __global int *src1, int src1_step, int src1_offset, - __global int *dst, int dst_step, int dst_offset, - int4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0)); - int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4)); - int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8)); - - int src2_data_0 = src2.x; - int src2_data_1 = src2.y; - int src2_data_2 = src2.z; - - int data_0 = *((__global int *)((__global char *)dst + dst_index + 0)); - int data_1 = *((__global int *)((__global char *)dst + dst_index + 4)); - int data_2 = *((__global int *)((__global char *)dst + dst_index + 8)); - - int tmp_data_0 = src1_data_0 ^ src2_data_0; - int tmp_data_1 = src1_data_1 ^ src2_data_1; - int tmp_data_2 = src1_data_2 ^ src2_data_2; - - *((__global int *)((__global char *)dst + dst_index + 0))= tmp_data_0; - *((__global int *)((__global char *)dst + dst_index + 4))= tmp_data_1; - *((__global int *)((__global char *)dst + dst_index + 8))= tmp_data_2; - } -} -__kernel void arithm_s_bitwise_xor_C3_D5 ( - __global char *src1, int src1_step, int src1_offset, - __global char *dst, int dst_step, int dst_offset, - char16 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - char4 src1_data_0 = *((__global char4 *)((__global char *)src1 + src1_index + 0)); - char4 src1_data_1 = *((__global char4 *)((__global char *)src1 + src1_index + 4)); - char4 src1_data_2 = *((__global char4 *)((__global char *)src1 + src1_index + 8)); - - char4 src2_data_0 = (char4)(src2.s0, src2.s1, src2.s2, src2.s3); - char4 src2_data_1 = (char4)(src2.s4, src2.s5, src2.s6, src2.s7); - char4 src2_data_2 = (char4)(src2.s8, src2.s9, src2.sA, src2.sB); - - char4 data_0 = *((__global char4 *)((__global char *)dst + dst_index + 0)); - char4 data_1 = *((__global char4 *)((__global char *)dst + dst_index + 4)); - char4 data_2 = *((__global char4 *)((__global char *)dst + dst_index + 8)); - - char4 tmp_data_0 = src1_data_0 ^ src2_data_0; - char4 tmp_data_1 = src1_data_1 ^ src2_data_1; - char4 tmp_data_2 = src1_data_2 ^ src2_data_2; - - *((__global char4 *)((__global char *)dst + dst_index + 0))= tmp_data_0; - *((__global char4 *)((__global char *)dst + dst_index + 4))= tmp_data_1; - *((__global char4 *)((__global char *)dst + dst_index + 8))= tmp_data_2; - } -} -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_s_bitwise_xor_C3_D6 ( - __global short *src1, int src1_step, int src1_offset, - __global short *dst, int dst_step, int dst_offset, - short16 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 24) + src1_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 24)); - - short4 src1_data_0 = *((__global short4 *)((__global char *)src1 + src1_index + 0 )); - short4 src1_data_1 = *((__global short4 *)((__global char *)src1 + src1_index + 8 )); - short4 src1_data_2 = *((__global short4 *)((__global char *)src1 + src1_index + 16)); - - short4 src2_data_0 = (short4)(src2.s0, src2.s1, src2.s2, src2.s3); - short4 src2_data_1 = (short4)(src2.s4, src2.s5, src2.s6, src2.s7); - short4 src2_data_2 = (short4)(src2.s8, src2.s9, src2.sa, src2.sb); - - short4 data_0 = *((__global short4 *)((__global char *)dst + dst_index + 0 )); - short4 data_1 = *((__global short4 *)((__global char *)dst + dst_index + 8 )); - short4 data_2 = *((__global short4 *)((__global char *)dst + dst_index + 16)); - - short4 tmp_data_0 = src1_data_0 ^ src2_data_0; - short4 tmp_data_1 = src1_data_1 ^ src2_data_1; - short4 tmp_data_2 = src1_data_2 ^ src2_data_2; - - *((__global short4 *)((__global char *)dst + dst_index + 0 ))= tmp_data_0; - *((__global short4 *)((__global char *)dst + dst_index + 8 ))= tmp_data_1; - *((__global short4 *)((__global char *)dst + dst_index + 16))= tmp_data_2; - } -} -#endif __kernel void arithm_s_bitwise_xor_C4_D0 ( __global uchar *src1, int src1_step, int src1_offset, __global uchar *dst, int dst_step, int dst_offset, diff --git a/modules/ocl/src/opencl/arithm_bitwise_xor_scalar_mask.cl b/modules/ocl/src/opencl/arithm_bitwise_xor_scalar_mask.cl index 26ca59c3a3..ad481aad80 100644 --- a/modules/ocl/src/opencl/arithm_bitwise_xor_scalar_mask.cl +++ b/modules/ocl/src/opencl/arithm_bitwise_xor_scalar_mask.cl @@ -523,380 +523,7 @@ __kernel void arithm_s_bitwise_xor_with_mask_C2_D6 ( } } #endif -__kernel void arithm_s_bitwise_xor_with_mask_C3_D0 ( - __global uchar *src1, int src1_step, int src1_offset, - __global uchar *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - uchar4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 3 ) & 3) - int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3)); - - uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0); - uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4); - uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8); - - uchar4 src2_data_0 = (uchar4)(src2.x, src2.y, src2.z, src2.x); - uchar4 src2_data_1 = (uchar4)(src2.y, src2.z, src2.x, src2.y); - uchar4 src2_data_2 = (uchar4)(src2.z, src2.x, src2.y, src2.z); - - uchar4 mask_data = vload4(0, mask + mask_index); - - uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0)); - uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4)); - uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8)); - - uchar4 tmp_data_0 = src1_data_0 ^ src2_data_0; - uchar4 tmp_data_1 = src1_data_1 ^ src2_data_1; - uchar4 tmp_data_2 = src1_data_2 ^ src2_data_2; - - data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz; - data_0.w = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_0.w : data_0.w; - - data_1.xy = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_1.xy : data_1.xy; - data_1.zw = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.zw : data_1.zw; - - data_2.x = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.x : data_2.x; - data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end)) - ? tmp_data_2.yzw : data_2.yzw; - - *((__global uchar4 *)(dst + dst_index + 0)) = data_0; - *((__global uchar4 *)(dst + dst_index + 4)) = data_1; - *((__global uchar4 *)(dst + dst_index + 8)) = data_2; - } -} - - -__kernel void arithm_s_bitwise_xor_with_mask_C3_D1 ( - __global char *src1, int src1_step, int src1_offset, - __global char *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - char4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 3 ) & 3) - int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3)); - - char4 src1_data_0 = vload4(0, src1 + src1_index + 0); - char4 src1_data_1 = vload4(0, src1 + src1_index + 4); - char4 src1_data_2 = vload4(0, src1 + src1_index + 8); - - char4 src2_data_0 = (char4)(src2.x, src2.y, src2.z, src2.x); - char4 src2_data_1 = (char4)(src2.y, src2.z, src2.x, src2.y); - char4 src2_data_2 = (char4)(src2.z, src2.x, src2.y, src2.z); - - uchar4 mask_data = vload4(0, mask + mask_index); - - char4 data_0 = *((__global char4 *)(dst + dst_index + 0)); - char4 data_1 = *((__global char4 *)(dst + dst_index + 4)); - char4 data_2 = *((__global char4 *)(dst + dst_index + 8)); - - char4 tmp_data_0 = src1_data_0 ^ src2_data_0; - char4 tmp_data_1 = src1_data_1 ^ src2_data_1; - char4 tmp_data_2 = src1_data_2 ^ src2_data_2; - - data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz; - data_0.w = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_0.w : data_0.w; - - data_1.xy = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_1.xy : data_1.xy; - data_1.zw = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.zw : data_1.zw; - data_2.x = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.x : data_2.x; - data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end)) - ? tmp_data_2.yzw : data_2.yzw; - - *((__global char4 *)(dst + dst_index + 0)) = data_0; - *((__global char4 *)(dst + dst_index + 4)) = data_1; - *((__global char4 *)(dst + dst_index + 8)) = data_2; - } -} - -__kernel void arithm_s_bitwise_xor_with_mask_C3_D2 ( - __global ushort *src1, int src1_step, int src1_offset, - __global ushort *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - ushort4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0)); - ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4)); - ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8)); - - ushort2 src2_data_0 = (ushort2)(src2.x, src2.y); - ushort2 src2_data_1 = (ushort2)(src2.z, src2.x); - ushort2 src2_data_2 = (ushort2)(src2.y, src2.z); - - uchar2 mask_data = vload2(0, mask + mask_index); - - ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0)); - ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4)); - ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8)); - - ushort2 tmp_data_0 = src1_data_0 ^ src2_data_0; - ushort2 tmp_data_1 = src1_data_1 ^ src2_data_1; - ushort2 tmp_data_2 = src1_data_2 ^ src2_data_2; - - data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_s_bitwise_xor_with_mask_C3_D3 ( - __global short *src1, int src1_step, int src1_offset, - __global short *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - short4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - -#ifdef dst_align -#undef dst_align -#endif -#define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0)); - short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4)); - short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8)); - - short2 src2_data_0 = (short2)(src2.x, src2.y); - short2 src2_data_1 = (short2)(src2.z, src2.x); - short2 src2_data_2 = (short2)(src2.y, src2.z); - - uchar2 mask_data = vload2(0, mask + mask_index); - - short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0)); - short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4)); - short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8)); - - short2 tmp_data_0 = src1_data_0 ^ src2_data_0; - short2 tmp_data_1 = src1_data_1 ^ src2_data_1; - short2 tmp_data_2 = src1_data_2 ^ src2_data_2; - - data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_s_bitwise_xor_with_mask_C3_D4 ( - __global int *src1, int src1_step, int src1_offset, - __global int *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - int4 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0)); - int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4)); - int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8)); - - int src2_data_0 = src2.x; - int src2_data_1 = src2.y; - int src2_data_2 = src2.z; - - uchar mask_data = * (mask + mask_index); - - int data_0 = *((__global int *)((__global char *)dst + dst_index + 0)); - int data_1 = *((__global int *)((__global char *)dst + dst_index + 4)); - int data_2 = *((__global int *)((__global char *)dst + dst_index + 8)); - - int tmp_data_0 = src1_data_0 ^ src2_data_0; - int tmp_data_1 = src1_data_1 ^ src2_data_1; - int tmp_data_2 = src1_data_2 ^ src2_data_2; - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global int *)((__global char *)dst + dst_index + 0))= data_0; - *((__global int *)((__global char *)dst + dst_index + 4))= data_1; - *((__global int *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_s_bitwise_xor_with_mask_C3_D5 ( - __global char *src1, int src1_step, int src1_offset, - __global char *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - char16 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - char4 src1_data_0 = *((__global char4 *)((__global char *)src1 + src1_index + 0)); - char4 src1_data_1 = *((__global char4 *)((__global char *)src1 + src1_index + 4)); - char4 src1_data_2 = *((__global char4 *)((__global char *)src1 + src1_index + 8)); - - char4 src2_data_0 = (char4)(src2.s0, src2.s1, src2.s2, src2.s3); - char4 src2_data_1 = (char4)(src2.s4, src2.s5, src2.s6, src2.s7); - char4 src2_data_2 = (char4)(src2.s8, src2.s9, src2.sA, src2.sB); - - uchar mask_data = * (mask + mask_index); - - char4 data_0 = *((__global char4 *)((__global char *)dst + dst_index + 0)); - char4 data_1 = *((__global char4 *)((__global char *)dst + dst_index + 4)); - char4 data_2 = *((__global char4 *)((__global char *)dst + dst_index + 8)); - - char4 tmp_data_0 = src1_data_0 ^ src2_data_0; - char4 tmp_data_1 = src1_data_1 ^ src2_data_1; - char4 tmp_data_2 = src1_data_2 ^ src2_data_2; - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global char4 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global char4 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global char4 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_s_bitwise_xor_with_mask_C3_D6 ( - __global short *src1, int src1_step, int src1_offset, - __global short *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - short16 src2, int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 24) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 24)); - - short4 src1_data_0 = *((__global short4 *)((__global char *)src1 + src1_index + 0 )); - short4 src1_data_1 = *((__global short4 *)((__global char *)src1 + src1_index + 8 )); - short4 src1_data_2 = *((__global short4 *)((__global char *)src1 + src1_index + 16)); - - short4 src2_data_0 = (short4)(src2.s0, src2.s1, src2.s2, src2.s3); - short4 src2_data_1 = (short4)(src2.s4, src2.s5, src2.s6, src2.s7); - short4 src2_data_2 = (short4)(src2.s8, src2.s9, src2.sa, src2.sb); - - uchar mask_data = * (mask + mask_index); - - short4 data_0 = *((__global short4 *)((__global char *)dst + dst_index + 0 )); - short4 data_1 = *((__global short4 *)((__global char *)dst + dst_index + 8 )); - short4 data_2 = *((__global short4 *)((__global char *)dst + dst_index + 16)); - - short4 tmp_data_0 = src1_data_0 ^ src2_data_0; - short4 tmp_data_1 = src1_data_1 ^ src2_data_1; - short4 tmp_data_2 = src1_data_2 ^ src2_data_2; - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global short4 *)((__global char *)dst + dst_index + 0 ))= data_0; - *((__global short4 *)((__global char *)dst + dst_index + 8 ))= data_1; - *((__global short4 *)((__global char *)dst + dst_index + 16))= data_2; - } -} -#endif __kernel void arithm_s_bitwise_xor_with_mask_C4_D0 ( __global uchar *src1, int src1_step, int src1_offset, __global uchar *dst, int dst_step, int dst_offset, diff --git a/modules/ocl/src/opencl/arithm_sub.cl b/modules/ocl/src/opencl/arithm_sub.cl index 9cf37970b2..d461d3aaea 100644 --- a/modules/ocl/src/opencl/arithm_sub.cl +++ b/modules/ocl/src/opencl/arithm_sub.cl @@ -44,7 +44,11 @@ //M*/ #if defined (DOUBLE_SUPPORT) +#ifdef cl_khr_fp64 #pragma OPENCL EXTENSION cl_khr_fp64:enable +#elif defined (cl_amd_fp64) +#pragma OPENCL EXTENSION cl_amd_fp64:enable +#endif #endif ////////////////////////////////////////////////////////////////////////////////////////////////////// @@ -618,313 +622,7 @@ __kernel void arithm_sub_with_mask_C2_D6 (__global double *src1, int src1_step, } } #endif -__kernel void arithm_sub_with_mask_C3_D0 (__global uchar *src1, int src1_step, int src1_offset, - __global uchar *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global uchar *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - - #define dst_align (((dst_offset % dst_step) / 3 ) & 3) - int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3)); - int src2_index = mad24(y, src2_step, (x * 3) + src2_offset - (dst_align * 3)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3)); - - uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0); - uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4); - uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8); - - uchar4 src2_data_0 = vload4(0, src2 + src2_index + 0); - uchar4 src2_data_1 = vload4(0, src2 + src2_index + 4); - uchar4 src2_data_2 = vload4(0, src2 + src2_index + 8); - - uchar4 mask_data = vload4(0, mask + mask_index); - - uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0)); - uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4)); - uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8)); - - uchar4 tmp_data_0 = convert_uchar4_sat(convert_short4_sat(src1_data_0) - convert_short4_sat(src2_data_0)); - uchar4 tmp_data_1 = convert_uchar4_sat(convert_short4_sat(src1_data_1) - convert_short4_sat(src2_data_1)); - uchar4 tmp_data_2 = convert_uchar4_sat(convert_short4_sat(src1_data_2) - convert_short4_sat(src2_data_2)); - - data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz; - data_0.w = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_0.w : data_0.w; - - data_1.xy = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_1.xy : data_1.xy; - data_1.zw = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.zw : data_1.zw; - - data_2.x = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.x : data_2.x; - data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end)) - ? tmp_data_2.yzw : data_2.yzw; - - *((__global uchar4 *)(dst + dst_index + 0)) = data_0; - *((__global uchar4 *)(dst + dst_index + 4)) = data_1; - *((__global uchar4 *)(dst + dst_index + 8)) = data_2; - } -} -__kernel void arithm_sub_with_mask_C3_D2 (__global ushort *src1, int src1_step, int src1_offset, - __global ushort *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global ushort *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - - #define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - int src2_index = mad24(y, src2_step, (x * 6) + src2_offset - (dst_align * 6)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0)); - ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4)); - ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8)); - - ushort2 src2_data_0 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 0)); - ushort2 src2_data_1 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 4)); - ushort2 src2_data_2 = vload2(0, (__global ushort *)((__global char *)src2 + src2_index + 8)); - - uchar2 mask_data = vload2(0, mask + mask_index); - - ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0)); - ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4)); - ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8)); - - ushort2 tmp_data_0 = convert_ushort2_sat(convert_int2_sat(src1_data_0) - convert_int2_sat(src2_data_0)); - ushort2 tmp_data_1 = convert_ushort2_sat(convert_int2_sat(src1_data_1) - convert_int2_sat(src2_data_1)); - ushort2 tmp_data_2 = convert_ushort2_sat(convert_int2_sat(src1_data_2) - convert_int2_sat(src2_data_2)); - - data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_sub_with_mask_C3_D3 (__global short *src1, int src1_step, int src1_offset, - __global short *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global short *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - - #define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - int src2_index = mad24(y, src2_step, (x * 6) + src2_offset - (dst_align * 6)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0)); - short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4)); - short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8)); - - short2 src2_data_0 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 0)); - short2 src2_data_1 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 4)); - short2 src2_data_2 = vload2(0, (__global short *)((__global char *)src2 + src2_index + 8)); - - uchar2 mask_data = vload2(0, mask + mask_index); - - short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0)); - short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4)); - short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8)); - - short2 tmp_data_0 = convert_short2_sat(convert_int2_sat(src1_data_0) - convert_int2_sat(src2_data_0)); - short2 tmp_data_1 = convert_short2_sat(convert_int2_sat(src1_data_1) - convert_int2_sat(src2_data_1)); - short2 tmp_data_2 = convert_short2_sat(convert_int2_sat(src1_data_2) - convert_int2_sat(src2_data_2)); - - data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_sub_with_mask_C3_D4 (__global int *src1, int src1_step, int src1_offset, - __global int *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global int *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int src2_index = mad24(y, src2_step, (x * 12) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0)); - int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4)); - int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8)); - - int src2_data_0 = *((__global int *)((__global char *)src2 + src2_index + 0)); - int src2_data_1 = *((__global int *)((__global char *)src2 + src2_index + 4)); - int src2_data_2 = *((__global int *)((__global char *)src2 + src2_index + 8)); - - uchar mask_data = * (mask + mask_index); - - int data_0 = *((__global int *)((__global char *)dst + dst_index + 0)); - int data_1 = *((__global int *)((__global char *)dst + dst_index + 4)); - int data_2 = *((__global int *)((__global char *)dst + dst_index + 8)); - - int tmp_data_0 = convert_int_sat((long)src1_data_0 - (long)src2_data_0); - int tmp_data_1 = convert_int_sat((long)src1_data_1 - (long)src2_data_1); - int tmp_data_2 = convert_int_sat((long)src1_data_2 - (long)src2_data_2); - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global int *)((__global char *)dst + dst_index + 0))= data_0; - *((__global int *)((__global char *)dst + dst_index + 4))= data_1; - *((__global int *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_sub_with_mask_C3_D5 (__global float *src1, int src1_step, int src1_offset, - __global float *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global float *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int src2_index = mad24(y, src2_step, (x * 12) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - float src1_data_0 = *((__global float *)((__global char *)src1 + src1_index + 0)); - float src1_data_1 = *((__global float *)((__global char *)src1 + src1_index + 4)); - float src1_data_2 = *((__global float *)((__global char *)src1 + src1_index + 8)); - - float src2_data_0 = *((__global float *)((__global char *)src2 + src2_index + 0)); - float src2_data_1 = *((__global float *)((__global char *)src2 + src2_index + 4)); - float src2_data_2 = *((__global float *)((__global char *)src2 + src2_index + 8)); - - uchar mask_data = * (mask + mask_index); - - float data_0 = *((__global float *)((__global char *)dst + dst_index + 0)); - float data_1 = *((__global float *)((__global char *)dst + dst_index + 4)); - float data_2 = *((__global float *)((__global char *)dst + dst_index + 8)); - - float tmp_data_0 = src1_data_0 - src2_data_0; - float tmp_data_1 = src1_data_1 - src2_data_1; - float tmp_data_2 = src1_data_2 - src2_data_2; - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global float *)((__global char *)dst + dst_index + 0))= data_0; - *((__global float *)((__global char *)dst + dst_index + 4))= data_1; - *((__global float *)((__global char *)dst + dst_index + 8))= data_2; - } -} - -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_sub_with_mask_C3_D6 (__global double *src1, int src1_step, int src1_offset, - __global double *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global double *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 24) + src1_offset); - int src2_index = mad24(y, src2_step, (x * 24) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 24)); - - double src1_data_0 = *((__global double *)((__global char *)src1 + src1_index + 0 )); - double src1_data_1 = *((__global double *)((__global char *)src1 + src1_index + 8 )); - double src1_data_2 = *((__global double *)((__global char *)src1 + src1_index + 16)); - - double src2_data_0 = *((__global double *)((__global char *)src2 + src2_index + 0 )); - double src2_data_1 = *((__global double *)((__global char *)src2 + src2_index + 8 )); - double src2_data_2 = *((__global double *)((__global char *)src2 + src2_index + 16)); - - uchar mask_data = * (mask + mask_index); - - double data_0 = *((__global double *)((__global char *)dst + dst_index + 0 )); - double data_1 = *((__global double *)((__global char *)dst + dst_index + 8 )); - double data_2 = *((__global double *)((__global char *)dst + dst_index + 16)); - - double tmp_data_0 = src1_data_0 - src2_data_0; - double tmp_data_1 = src1_data_1 - src2_data_1; - double tmp_data_2 = src1_data_2 - src2_data_2; - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global double *)((__global char *)dst + dst_index + 0 ))= data_0; - *((__global double *)((__global char *)dst + dst_index + 8 ))= data_1; - *((__global double *)((__global char *)dst + dst_index + 16))= data_2; - } -} -#endif __kernel void arithm_sub_with_mask_C4_D0 (__global uchar *src1, int src1_step, int src1_offset, __global uchar *src2, int src2_step, int src2_offset, __global uchar *mask, int mask_step, int mask_offset, diff --git a/modules/ocl/src/opencl/arithm_sub_scalar.cl b/modules/ocl/src/opencl/arithm_sub_scalar.cl index 782bcd0607..76bb29418c 100644 --- a/modules/ocl/src/opencl/arithm_sub_scalar.cl +++ b/modules/ocl/src/opencl/arithm_sub_scalar.cl @@ -42,9 +42,12 @@ // the use of this software, even if advised of the possibility of such damage. // //M*/ - #if defined (DOUBLE_SUPPORT) +#ifdef cl_khr_fp64 #pragma OPENCL EXTENSION cl_khr_fp64:enable +#elif defined (cl_amd_fp64) +#pragma OPENCL EXTENSION cl_amd_fp64:enable +#endif #endif /**************************************sub with scalar without mask**************************************/ __kernel void arithm_s_sub_C1_D0 (__global uchar *src1, int src1_step, int src1_offset, @@ -372,305 +375,7 @@ __kernel void arithm_s_sub_C2_D6 (__global double *src1, int src1_step, int sr } } #endif -__kernel void arithm_s_sub_C3_D0 (__global uchar *src1, int src1_step, int src1_offset, - __global uchar *dst, int dst_step, int dst_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - - #define dst_align (((dst_offset % dst_step) / 3 ) & 3) - int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3)); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3)); - - uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0); - uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4); - uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8); - - int4 src2_data_0 = (int4)(src2.x, src2.y, src2.z, src2.x); - int4 src2_data_1 = (int4)(src2.y, src2.z, src2.x, src2.y); - int4 src2_data_2 = (int4)(src2.z, src2.x, src2.y, src2.z); - - uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0)); - uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4)); - uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8)); - - int4 tmp_0 = convert_int4_sat(src1_data_0) - src2_data_0; - int4 tmp_1 = convert_int4_sat(src1_data_1) - src2_data_1; - int4 tmp_2 = convert_int4_sat(src1_data_2) - src2_data_2; - - tmp_0 = isMatSubScalar ? tmp_0 : -tmp_0; - tmp_1 = isMatSubScalar ? tmp_1 : -tmp_1; - tmp_2 = isMatSubScalar ? tmp_2 : -tmp_2; - - uchar4 tmp_data_0 = convert_uchar4_sat(tmp_0); - uchar4 tmp_data_1 = convert_uchar4_sat(tmp_1); - uchar4 tmp_data_2 = convert_uchar4_sat(tmp_2); - - data_0.xyz = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz; - data_0.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_0.w : data_0.w; - - data_1.xy = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_1.xy : data_1.xy; - data_1.zw = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.zw : data_1.zw; - - data_2.x = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.x : data_2.x; - data_2.yzw = ((dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end)) - ? tmp_data_2.yzw : data_2.yzw; - - *((__global uchar4 *)(dst + dst_index + 0)) = data_0; - *((__global uchar4 *)(dst + dst_index + 4)) = data_1; - *((__global uchar4 *)(dst + dst_index + 8)) = data_2; - } -} -__kernel void arithm_s_sub_C3_D2 (__global ushort *src1, int src1_step, int src1_offset, - __global ushort *dst, int dst_step, int dst_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - - #define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0)); - ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4)); - ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8)); - - int2 src2_data_0 = (int2)(src2.x, src2.y); - int2 src2_data_1 = (int2)(src2.z, src2.x); - int2 src2_data_2 = (int2)(src2.y, src2.z); - - ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0)); - ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4)); - ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8)); - - int2 tmp_0 = convert_int2_sat(src1_data_0) - src2_data_0; - int2 tmp_1 = convert_int2_sat(src1_data_1) - src2_data_1; - int2 tmp_2 = convert_int2_sat(src1_data_2) - src2_data_2; - - tmp_0 = isMatSubScalar ? tmp_0 : -tmp_0; - tmp_1 = isMatSubScalar ? tmp_1 : -tmp_1; - tmp_2 = isMatSubScalar ? tmp_2 : -tmp_2; - - ushort2 tmp_data_0 = convert_ushort2_sat(tmp_0); - ushort2 tmp_data_1 = convert_ushort2_sat(tmp_1); - ushort2 tmp_data_2 = convert_ushort2_sat(tmp_2); - - data_0.xy = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_s_sub_C3_D3 (__global short *src1, int src1_step, int src1_offset, - __global short *dst, int dst_step, int dst_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - - #define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0)); - short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4)); - short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8)); - - int2 src2_data_0 = (int2)(src2.x, src2.y); - int2 src2_data_1 = (int2)(src2.z, src2.x); - int2 src2_data_2 = (int2)(src2.y, src2.z); - - short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0)); - short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4)); - short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8)); - - int2 tmp_0 = convert_int2_sat(src1_data_0) - src2_data_0; - int2 tmp_1 = convert_int2_sat(src1_data_1) - src2_data_1; - int2 tmp_2 = convert_int2_sat(src1_data_2) - src2_data_2; - - tmp_0 = isMatSubScalar ? tmp_0 : -tmp_0; - tmp_1 = isMatSubScalar ? tmp_1 : -tmp_1; - tmp_2 = isMatSubScalar ? tmp_2 : -tmp_2; - - short2 tmp_data_0 = convert_short2_sat(tmp_0); - short2 tmp_data_1 = convert_short2_sat(tmp_1); - short2 tmp_data_2 = convert_short2_sat(tmp_2); - - data_0.xy = ((dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_s_sub_C3_D4 (__global int *src1, int src1_step, int src1_offset, - __global int *dst, int dst_step, int dst_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0)); - int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4)); - int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8)); - - int src2_data_0 = src2.x; - int src2_data_1 = src2.y; - int src2_data_2 = src2.z; - - int data_0 = *((__global int *)((__global char *)dst + dst_index + 0)); - int data_1 = *((__global int *)((__global char *)dst + dst_index + 4)); - int data_2 = *((__global int *)((__global char *)dst + dst_index + 8)); - - long tmp_0 = (long)src1_data_0 - (long)src2_data_0; - long tmp_1 = (long)src1_data_1 - (long)src2_data_1; - long tmp_2 = (long)src1_data_2 - (long)src2_data_2; - - tmp_0 = isMatSubScalar ? tmp_0 : -tmp_0; - tmp_1 = isMatSubScalar ? tmp_1 : -tmp_1; - tmp_2 = isMatSubScalar ? tmp_2 : -tmp_2; - - int tmp_data_0 = convert_int_sat(tmp_0); - int tmp_data_1 = convert_int_sat(tmp_1); - int tmp_data_2 = convert_int_sat(tmp_2); - - *((__global int *)((__global char *)dst + dst_index + 0))= tmp_data_0; - *((__global int *)((__global char *)dst + dst_index + 4))= tmp_data_1; - *((__global int *)((__global char *)dst + dst_index + 8))= tmp_data_2; - } -} -__kernel void arithm_s_sub_C3_D5 (__global float *src1, int src1_step, int src1_offset, - __global float *dst, int dst_step, int dst_offset, - float4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - float src1_data_0 = *((__global float *)((__global char *)src1 + src1_index + 0)); - float src1_data_1 = *((__global float *)((__global char *)src1 + src1_index + 4)); - float src1_data_2 = *((__global float *)((__global char *)src1 + src1_index + 8)); - - float src2_data_0 = src2.x; - float src2_data_1 = src2.y; - float src2_data_2 = src2.z; - - float data_0 = *((__global float *)((__global char *)dst + dst_index + 0)); - float data_1 = *((__global float *)((__global char *)dst + dst_index + 4)); - float data_2 = *((__global float *)((__global char *)dst + dst_index + 8)); - - float tmp_0 = src1_data_0 - src2_data_0; - float tmp_1 = src1_data_1 - src2_data_1; - float tmp_2 = src1_data_2 - src2_data_2; - - tmp_0 = isMatSubScalar ? tmp_0 : -tmp_0; - tmp_1 = isMatSubScalar ? tmp_1 : -tmp_1; - tmp_2 = isMatSubScalar ? tmp_2 : -tmp_2; - - *((__global float *)((__global char *)dst + dst_index + 0))= tmp_0; - *((__global float *)((__global char *)dst + dst_index + 4))= tmp_1; - *((__global float *)((__global char *)dst + dst_index + 8))= tmp_2; - } -} - -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_s_sub_C3_D6 (__global double *src1, int src1_step, int src1_offset, - __global double *dst, int dst_step, int dst_offset, - double4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 24) + src1_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 24)); - - double src1_data_0 = *((__global double *)((__global char *)src1 + src1_index + 0 )); - double src1_data_1 = *((__global double *)((__global char *)src1 + src1_index + 8 )); - double src1_data_2 = *((__global double *)((__global char *)src1 + src1_index + 16)); - - double src2_data_0 = src2.x; - double src2_data_1 = src2.y; - double src2_data_2 = src2.z; - - double data_0 = *((__global double *)((__global char *)dst + dst_index + 0 )); - double data_1 = *((__global double *)((__global char *)dst + dst_index + 8 )); - double data_2 = *((__global double *)((__global char *)dst + dst_index + 16)); - - double tmp_data_0 = src1_data_0 - src2_data_0; - double tmp_data_1 = src1_data_1 - src2_data_1; - double tmp_data_2 = src1_data_2 - src2_data_2; - - tmp_data_0 = isMatSubScalar ? tmp_data_0 : -tmp_data_0; - tmp_data_1 = isMatSubScalar ? tmp_data_1 : -tmp_data_1; - tmp_data_2 = isMatSubScalar ? tmp_data_2 : -tmp_data_2; - - *((__global double *)((__global char *)dst + dst_index + 0 ))= tmp_data_0; - *((__global double *)((__global char *)dst + dst_index + 8 ))= tmp_data_1; - *((__global double *)((__global char *)dst + dst_index + 16))= tmp_data_2; - } -} -#endif __kernel void arithm_s_sub_C4_D0 (__global uchar *src1, int src1_step, int src1_offset, __global uchar *dst, int dst_step, int dst_offset, int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) diff --git a/modules/ocl/src/opencl/arithm_sub_scalar_mask.cl b/modules/ocl/src/opencl/arithm_sub_scalar_mask.cl index 135354993c..9b758cf4c9 100644 --- a/modules/ocl/src/opencl/arithm_sub_scalar_mask.cl +++ b/modules/ocl/src/opencl/arithm_sub_scalar_mask.cl @@ -44,7 +44,11 @@ //M*/ #if defined (DOUBLE_SUPPORT) +#ifdef cl_khr_fp64 #pragma OPENCL EXTENSION cl_khr_fp64:enable +#elif defined (cl_amd_fp64) +#pragma OPENCL EXTENSION cl_amd_fp64:enable +#endif #endif /**************************************sub with scalar with mask**************************************/ @@ -430,341 +434,7 @@ __kernel void arithm_s_sub_with_mask_C2_D6 (__global double *src1, int src1_st } } #endif -__kernel void arithm_s_sub_with_mask_C3_D0 (__global uchar *src1, int src1_step, int src1_offset, - __global uchar *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - - #define dst_align (((dst_offset % dst_step) / 3 ) & 3) - int src1_index = mad24(y, src1_step, (x * 3) + src1_offset - (dst_align * 3)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 3) - (dst_align * 3)); - - uchar4 src1_data_0 = vload4(0, src1 + src1_index + 0); - uchar4 src1_data_1 = vload4(0, src1 + src1_index + 4); - uchar4 src1_data_2 = vload4(0, src1 + src1_index + 8); - - int4 src2_data_0 = (int4)(src2.x, src2.y, src2.z, src2.x); - int4 src2_data_1 = (int4)(src2.y, src2.z, src2.x, src2.y); - int4 src2_data_2 = (int4)(src2.z, src2.x, src2.y, src2.z); - - uchar4 mask_data = vload4(0, mask + mask_index); - - uchar4 data_0 = *((__global uchar4 *)(dst + dst_index + 0)); - uchar4 data_1 = *((__global uchar4 *)(dst + dst_index + 4)); - uchar4 data_2 = *((__global uchar4 *)(dst + dst_index + 8)); - - int4 tmp_0 = convert_int4_sat(src1_data_0) - src2_data_0; - int4 tmp_1 = convert_int4_sat(src1_data_1) - src2_data_1; - int4 tmp_2 = convert_int4_sat(src1_data_2) - src2_data_2; - - tmp_0 = isMatSubScalar ? tmp_0 : -tmp_0; - tmp_1 = isMatSubScalar ? tmp_1 : -tmp_1; - tmp_2 = isMatSubScalar ? tmp_2 : -tmp_2; - - uchar4 tmp_data_0 = convert_uchar4_sat(tmp_0); - uchar4 tmp_data_1 = convert_uchar4_sat(tmp_1); - uchar4 tmp_data_2 = convert_uchar4_sat(tmp_2); - - data_0.xyz = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xyz : data_0.xyz; - data_0.w = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_0.w : data_0.w; - - data_1.xy = ((mask_data.y) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) - ? tmp_data_1.xy : data_1.xy; - data_1.zw = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.zw : data_1.zw; - - data_2.x = ((mask_data.z) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.x : data_2.x; - data_2.yzw = ((mask_data.w) && (dst_index + 9 >= dst_start) && (dst_index + 9 < dst_end)) - ? tmp_data_2.yzw : data_2.yzw; - - *((__global uchar4 *)(dst + dst_index + 0)) = data_0; - *((__global uchar4 *)(dst + dst_index + 4)) = data_1; - *((__global uchar4 *)(dst + dst_index + 8)) = data_2; - } -} -__kernel void arithm_s_sub_with_mask_C3_D2 (__global ushort *src1, int src1_step, int src1_offset, - __global ushort *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - - #define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - - ushort2 src1_data_0 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 0)); - ushort2 src1_data_1 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 4)); - ushort2 src1_data_2 = vload2(0, (__global ushort *)((__global char *)src1 + src1_index + 8)); - - int2 src2_data_0 = (int2)(src2.x, src2.y); - int2 src2_data_1 = (int2)(src2.z, src2.x); - int2 src2_data_2 = (int2)(src2.y, src2.z); - - uchar2 mask_data = vload2(0, mask + mask_index); - - ushort2 data_0 = *((__global ushort2 *)((__global char *)dst + dst_index + 0)); - ushort2 data_1 = *((__global ushort2 *)((__global char *)dst + dst_index + 4)); - ushort2 data_2 = *((__global ushort2 *)((__global char *)dst + dst_index + 8)); - - int2 tmp_0 = convert_int2_sat(src1_data_0) - src2_data_0; - int2 tmp_1 = convert_int2_sat(src1_data_1) - src2_data_1; - int2 tmp_2 = convert_int2_sat(src1_data_2) - src2_data_2; - - tmp_0 = isMatSubScalar ? tmp_0 : -tmp_0; - tmp_1 = isMatSubScalar ? tmp_1 : -tmp_1; - tmp_2 = isMatSubScalar ? tmp_2 : -tmp_2; - - ushort2 tmp_data_0 = convert_ushort2_sat(tmp_0); - ushort2 tmp_data_1 = convert_ushort2_sat(tmp_1); - ushort2 tmp_data_2 = convert_ushort2_sat(tmp_2); - - data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global ushort2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global ushort2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global ushort2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_s_sub_with_mask_C3_D3 (__global short *src1, int src1_step, int src1_offset, - __global short *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - - #define dst_align (((dst_offset % dst_step) / 6 ) & 1) - int src1_index = mad24(y, src1_step, (x * 6) + src1_offset - (dst_align * 6)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x * 6) - (dst_align * 6)); - short2 src1_data_0 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 0)); - short2 src1_data_1 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 4)); - short2 src1_data_2 = vload2(0, (__global short *)((__global char *)src1 + src1_index + 8)); - - int2 src2_data_0 = (int2)(src2.x, src2.y); - int2 src2_data_1 = (int2)(src2.z, src2.x); - int2 src2_data_2 = (int2)(src2.y, src2.z); - - uchar2 mask_data = vload2(0, mask + mask_index); - - short2 data_0 = *((__global short2 *)((__global char *)dst + dst_index + 0)); - short2 data_1 = *((__global short2 *)((__global char *)dst + dst_index + 4)); - short2 data_2 = *((__global short2 *)((__global char *)dst + dst_index + 8)); - - int2 tmp_0 = convert_int2_sat(src1_data_0) - src2_data_0; - int2 tmp_1 = convert_int2_sat(src1_data_1) - src2_data_1; - int2 tmp_2 = convert_int2_sat(src1_data_2) - src2_data_2; - - tmp_0 = isMatSubScalar ? tmp_0 : -tmp_0; - tmp_1 = isMatSubScalar ? tmp_1 : -tmp_1; - tmp_2 = isMatSubScalar ? tmp_2 : -tmp_2; - - short2 tmp_data_0 = convert_short2_sat(tmp_0); - short2 tmp_data_1 = convert_short2_sat(tmp_1); - short2 tmp_data_2 = convert_short2_sat(tmp_2); - - data_0.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data_0.xy : data_0.xy; - - data_1.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) - ? tmp_data_1.x : data_1.x; - data_1.y = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_1.y : data_1.y; - - data_2.xy = ((mask_data.y) && (dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) - ? tmp_data_2.xy : data_2.xy; - - *((__global short2 *)((__global char *)dst + dst_index + 0))= data_0; - *((__global short2 *)((__global char *)dst + dst_index + 4))= data_1; - *((__global short2 *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_s_sub_with_mask_C3_D4 (__global int *src1, int src1_step, int src1_offset, - __global int *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - int src1_data_0 = *((__global int *)((__global char *)src1 + src1_index + 0)); - int src1_data_1 = *((__global int *)((__global char *)src1 + src1_index + 4)); - int src1_data_2 = *((__global int *)((__global char *)src1 + src1_index + 8)); - - int src2_data_0 = src2.x; - int src2_data_1 = src2.y; - int src2_data_2 = src2.z; - - uchar mask_data = * (mask + mask_index); - - int data_0 = *((__global int *)((__global char *)dst + dst_index + 0)); - int data_1 = *((__global int *)((__global char *)dst + dst_index + 4)); - int data_2 = *((__global int *)((__global char *)dst + dst_index + 8)); - - long tmp_0 = (long)src1_data_0 - (long)src2_data_0; - long tmp_1 = (long)src1_data_1 - (long)src2_data_1; - long tmp_2 = (long)src1_data_2 - (long)src2_data_2; - - tmp_0 = isMatSubScalar ? tmp_0 : -tmp_0; - tmp_1 = isMatSubScalar ? tmp_1 : -tmp_1; - tmp_2 = isMatSubScalar ? tmp_2 : -tmp_2; - - int tmp_data_0 = convert_int_sat(tmp_0); - int tmp_data_1 = convert_int_sat(tmp_1); - int tmp_data_2 = convert_int_sat(tmp_2); - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global int *)((__global char *)dst + dst_index + 0))= data_0; - *((__global int *)((__global char *)dst + dst_index + 4))= data_1; - *((__global int *)((__global char *)dst + dst_index + 8))= data_2; - } -} -__kernel void arithm_s_sub_with_mask_C3_D5 (__global float *src1, int src1_step, int src1_offset, - __global float *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - float4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 12) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 12)); - - float src1_data_0 = *((__global float *)((__global char *)src1 + src1_index + 0)); - float src1_data_1 = *((__global float *)((__global char *)src1 + src1_index + 4)); - float src1_data_2 = *((__global float *)((__global char *)src1 + src1_index + 8)); - - float src2_data_0 = src2.x; - float src2_data_1 = src2.y; - float src2_data_2 = src2.z; - - uchar mask_data = * (mask + mask_index); - - float data_0 = *((__global float *)((__global char *)dst + dst_index + 0)); - float data_1 = *((__global float *)((__global char *)dst + dst_index + 4)); - float data_2 = *((__global float *)((__global char *)dst + dst_index + 8)); - - float tmp_data_0 = src1_data_0 - src2_data_0; - float tmp_data_1 = src1_data_1 - src2_data_1; - float tmp_data_2 = src1_data_2 - src2_data_2; - - tmp_data_0 = isMatSubScalar ? tmp_data_0 : -tmp_data_0; - tmp_data_1 = isMatSubScalar ? tmp_data_1 : -tmp_data_1; - tmp_data_2 = isMatSubScalar ? tmp_data_2 : -tmp_data_2; - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global float *)((__global char *)dst + dst_index + 0))= data_0; - *((__global float *)((__global char *)dst + dst_index + 4))= data_1; - *((__global float *)((__global char *)dst + dst_index + 8))= data_2; - } -} - -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_s_sub_with_mask_C3_D6 (__global double *src1, int src1_step, int src1_offset, - __global double *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - double4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x * 24) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, dst_offset + (x * 24)); - - double src1_data_0 = *((__global double *)((__global char *)src1 + src1_index + 0 )); - double src1_data_1 = *((__global double *)((__global char *)src1 + src1_index + 8 )); - double src1_data_2 = *((__global double *)((__global char *)src1 + src1_index + 16)); - - double src2_data_0 = src2.x; - double src2_data_1 = src2.y; - double src2_data_2 = src2.z; - - uchar mask_data = * (mask + mask_index); - - double data_0 = *((__global double *)((__global char *)dst + dst_index + 0 )); - double data_1 = *((__global double *)((__global char *)dst + dst_index + 8 )); - double data_2 = *((__global double *)((__global char *)dst + dst_index + 16)); - - double tmp_data_0 = src1_data_0 - src2_data_0; - double tmp_data_1 = src1_data_1 - src2_data_1; - double tmp_data_2 = src1_data_2 - src2_data_2; - - tmp_data_0 = isMatSubScalar ? tmp_data_0 : -tmp_data_0; - tmp_data_1 = isMatSubScalar ? tmp_data_1 : -tmp_data_1; - tmp_data_2 = isMatSubScalar ? tmp_data_2 : -tmp_data_2; - - data_0 = mask_data ? tmp_data_0 : data_0; - data_1 = mask_data ? tmp_data_1 : data_1; - data_2 = mask_data ? tmp_data_2 : data_2; - - *((__global double *)((__global char *)dst + dst_index + 0 ))= data_0; - *((__global double *)((__global char *)dst + dst_index + 8 ))= data_1; - *((__global double *)((__global char *)dst + dst_index + 16))= data_2; - } -} -#endif __kernel void arithm_s_sub_with_mask_C4_D0 (__global uchar *src1, int src1_step, int src1_offset, __global uchar *dst, int dst_step, int dst_offset, __global uchar *mask, int mask_step, int mask_offset, diff --git a/modules/ocl/test/test_arithm.cpp b/modules/ocl/test/test_arithm.cpp index f643864a86..e46fdbddd1 100644 --- a/modules/ocl/test/test_arithm.cpp +++ b/modules/ocl/test/test_arithm.cpp @@ -1531,6 +1531,10 @@ INSTANTIATE_TEST_CASE_P(Arithm, Add, Combine( Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32SC1, CV_32SC3, CV_32SC4, CV_32FC1, CV_32FC3, CV_32FC4), Values(false))); +INSTANTIATE_TEST_CASE_P(Arithm, Sub, Combine( + Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32SC1, CV_32SC3, CV_32SC4, CV_32FC1, CV_32FC3, CV_32FC4), + Values(false))); + INSTANTIATE_TEST_CASE_P(Arithm, Mul, Combine( Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32SC1, CV_32SC3, CV_32SC4, CV_32FC1, CV_32FC3, CV_32FC4), Values(false))); // Values(false) is the reserved parameter @@ -1586,19 +1590,19 @@ INSTANTIATE_TEST_CASE_P(Arithm, Phase, Combine(Values(CV_32FC1, CV_32FC3, CV_32F INSTANTIATE_TEST_CASE_P(Arithm, Bitwise_and, Combine( - Values(CV_8UC1, CV_32SC1, CV_32SC3, CV_32SC4, CV_32FC1, CV_32FC3, CV_32FC4), Values(false))); + Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32SC1, CV_32SC3, CV_32SC4, CV_32FC1, CV_32FC3, CV_32FC4), Values(false))); //Values(false) is the reserved parameter INSTANTIATE_TEST_CASE_P(Arithm, Bitwise_or, Combine( - Values(CV_8UC1, CV_8UC3, CV_32SC1, CV_32FC1, CV_32FC3, CV_32FC4), Values(false))); + Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32SC1, CV_32FC1, CV_32FC3, CV_32FC4), Values(false))); //Values(false) is the reserved parameter INSTANTIATE_TEST_CASE_P(Arithm, Bitwise_xor, Combine( - Values(CV_8UC1, CV_8UC3, CV_32SC1, CV_32FC1, CV_32FC3, CV_32FC4), Values(false))); + Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32SC1, CV_32FC1, CV_32FC3, CV_32FC4), Values(false))); //Values(false) is the reserved parameter INSTANTIATE_TEST_CASE_P(Arithm, Bitwise_not, Combine( - Values(CV_8UC1, CV_8UC3, CV_32SC1, CV_32FC1, CV_32FC3, CV_32FC4), Values(false))); + Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_32SC1, CV_32FC1, CV_32FC3, CV_32FC4), Values(false))); //Values(false) is the reserved parameter INSTANTIATE_TEST_CASE_P(Arithm, Compare, Combine(Values(CV_8UC1, CV_32SC1, CV_32FC1), Values(false))); From 7726e273a9b160635aefa895798673509979010f Mon Sep 17 00:00:00 2001 From: yao Date: Sat, 6 Apr 2013 13:37:36 +0800 Subject: [PATCH 2/2] merge add and sub into one set of kernels --- modules/ocl/src/arithm.cpp | 64 +- modules/ocl/src/opencl/arithm_add.cl | 53 +- modules/ocl/src/opencl/arithm_add_scalar.cl | 41 +- .../ocl/src/opencl/arithm_add_scalar_mask.cl | 42 +- modules/ocl/src/opencl/arithm_sub.cl | 802 ------------------ modules/ocl/src/opencl/arithm_sub_scalar.cl | 511 ----------- .../ocl/src/opencl/arithm_sub_scalar_mask.cl | 611 ------------- 7 files changed, 115 insertions(+), 2009 deletions(-) delete mode 100644 modules/ocl/src/opencl/arithm_sub.cl delete mode 100644 modules/ocl/src/opencl/arithm_sub_scalar.cl delete mode 100644 modules/ocl/src/opencl/arithm_sub_scalar_mask.cl diff --git a/modules/ocl/src/arithm.cpp b/modules/ocl/src/arithm.cpp index 410e460b6c..cba6ccf22f 100644 --- a/modules/ocl/src/arithm.cpp +++ b/modules/ocl/src/arithm.cpp @@ -92,9 +92,6 @@ namespace cv extern const char *arithm_bitwise_xor_scalar_mask; extern const char *arithm_compare_eq; extern const char *arithm_compare_ne; - extern const char *arithm_sub; - extern const char *arithm_sub_scalar; - extern const char *arithm_sub_scalar_mask; extern const char *arithm_mul; extern const char *arithm_div; extern const char *arithm_absdiff; @@ -130,7 +127,8 @@ inline int divUp(int total, int grain) /////////////////////// add subtract multiply divide ///////////////////////// ////////////////////////////////////////////////////////////////////////////// template -void arithmetic_run(const oclMat &src1, const oclMat &src2, oclMat &dst, string kernelName, const char **kernelString, void *_scalar) +void arithmetic_run(const oclMat &src1, const oclMat &src2, oclMat &dst, + string kernelName, const char **kernelString, void *_scalar, int op_type = 0) { if(!src1.clCxt->supportsFeature(Context::CL_DOUBLE) && src1.type() == CV_64F) { @@ -186,14 +184,25 @@ void arithmetic_run(const oclMat &src1, const oclMat &src2, oclMat &dst, string scalar = (T)scalar1; args.push_back( make_pair( sizeof(T), (void *)&scalar )); } - - openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, -1, depth); + switch(op_type) + { + case MAT_ADD: + openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, -1, depth, "-D ARITHM_ADD"); + break; + case MAT_SUB: + openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, -1, depth, "-D ARITHM_SUB"); + break; + default: + openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, -1, depth); + } } -static void arithmetic_run(const oclMat &src1, const oclMat &src2, oclMat &dst, string kernelName, const char **kernelString) +static void arithmetic_run(const oclMat &src1, const oclMat &src2, oclMat &dst, + string kernelName, const char **kernelString, int op_type = 0) { - arithmetic_run(src1, src2, dst, kernelName, kernelString, (void *)NULL); + arithmetic_run(src1, src2, dst, kernelName, kernelString, (void *)NULL, op_type); } -static void arithmetic_run(const oclMat &src1, const oclMat &src2, oclMat &dst, const oclMat &mask, string kernelName, const char **kernelString) +static void arithmetic_run(const oclMat &src1, const oclMat &src2, oclMat &dst, const oclMat &mask, + string kernelName, const char **kernelString, int op_type = 0) { if(!src1.clCxt->supportsFeature(Context::CL_DOUBLE) && src1.type() == CV_64F) { @@ -248,24 +257,34 @@ static void arithmetic_run(const oclMat &src1, const oclMat &src2, oclMat &dst, args.push_back( make_pair( sizeof(cl_int), (void *)&cols )); args.push_back( make_pair( sizeof(cl_int), (void *)&dst_step1 )); - openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, channels, depth); + switch (op_type) + { + case MAT_ADD: + openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, channels, depth, "-D ARITHM_ADD"); + break; + case MAT_SUB: + openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, channels, depth, "-D ARITHM_SUB"); + break; + default: + openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, channels, depth); + } } void cv::ocl::add(const oclMat &src1, const oclMat &src2, oclMat &dst) { - arithmetic_run(src1, src2, dst, "arithm_add", &arithm_add); + arithmetic_run(src1, src2, dst, "arithm_add", &arithm_add, MAT_ADD); } void cv::ocl::add(const oclMat &src1, const oclMat &src2, oclMat &dst, const oclMat &mask) { - arithmetic_run(src1, src2, dst, mask, "arithm_add_with_mask", &arithm_add); + arithmetic_run(src1, src2, dst, mask, "arithm_add_with_mask", &arithm_add, MAT_ADD); } void cv::ocl::subtract(const oclMat &src1, const oclMat &src2, oclMat &dst) { - arithmetic_run(src1, src2, dst, "arithm_sub", &arithm_sub); + arithmetic_run(src1, src2, dst, "arithm_add", &arithm_add, MAT_SUB); } void cv::ocl::subtract(const oclMat &src1, const oclMat &src2, oclMat &dst, const oclMat &mask) { - arithmetic_run(src1, src2, dst, mask, "arithm_sub_with_mask", &arithm_sub); + arithmetic_run(src1, src2, dst, mask, "arithm_add_with_mask", &arithm_add, MAT_SUB); } typedef void (*MulDivFunc)(const oclMat &src1, const oclMat &src2, oclMat &dst, string kernelName, const char **kernelString, void *scalar); @@ -351,12 +370,9 @@ void arithmetic_scalar_run(const oclMat &src1, const Scalar &src2, oclMat &dst, args.push_back( make_pair( sizeof(cl_int) , (void *)&cols )); args.push_back( make_pair( sizeof(cl_int) , (void *)&dst_step1 )); if(isMatSubScalar != 0) - { - isMatSubScalar = isMatSubScalar > 0 ? 1 : 0; - args.push_back( make_pair( sizeof(cl_int) , (void *)&isMatSubScalar)); - } - - openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, channels, depth); + openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, channels, depth, "-D ARITHM_SUB"); + else + openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, channels, depth, "-D ARITHM_ADD"); } static void arithmetic_scalar_run(const oclMat &src, oclMat &dst, string kernelName, const char **kernelString, double scalar) @@ -452,14 +468,14 @@ void cv::ocl::add(const oclMat &src1, const Scalar &src2, oclMat &dst, const ocl void cv::ocl::subtract(const oclMat &src1, const Scalar &src2, oclMat &dst, const oclMat &mask) { - string kernelName = mask.data ? "arithm_s_sub_with_mask" : "arithm_s_sub"; - const char **kernelString = mask.data ? &arithm_sub_scalar_mask : &arithm_sub_scalar; + string kernelName = mask.data ? "arithm_s_add_with_mask" : "arithm_s_add"; + const char **kernelString = mask.data ? &arithm_add_scalar_mask : &arithm_add_scalar; arithmetic_scalar( src1, src2, dst, mask, kernelName, kernelString, 1); } void cv::ocl::subtract(const Scalar &src2, const oclMat &src1, oclMat &dst, const oclMat &mask) { - string kernelName = mask.data ? "arithm_s_sub_with_mask" : "arithm_s_sub"; - const char **kernelString = mask.data ? &arithm_sub_scalar_mask : &arithm_sub_scalar; + string kernelName = mask.data ? "arithm_s_add_with_mask" : "arithm_s_add"; + const char **kernelString = mask.data ? &arithm_add_scalar_mask : &arithm_add_scalar; arithmetic_scalar( src1, src2, dst, mask, kernelName, kernelString, -1); } void cv::ocl::divide(double scalar, const oclMat &src, oclMat &dst) diff --git a/modules/ocl/src/opencl/arithm_add.cl b/modules/ocl/src/opencl/arithm_add.cl index f8f32cde6b..7d4b0a7653 100644 --- a/modules/ocl/src/opencl/arithm_add.cl +++ b/modules/ocl/src/opencl/arithm_add.cl @@ -52,6 +52,11 @@ #endif #endif +#ifdef ARITHM_ADD + #define ARITHM_OP(A,B) ((A)+(B)) +#elif defined ARITHM_SUB + #define ARITHM_OP(A,B) ((A)-(B)) +#endif ////////////////////////////////////////////////////////////////////////////////////////////////////// /////////////////////////////////////////////ADD//////////////////////////////////////////////////// /////////////////////////////////////////////////////////////////////////////////////////////////////// @@ -95,7 +100,7 @@ __kernel void arithm_add_D0 (__global uchar *src1, int src1_step, int src1_offse src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw; } uchar4 dst_data = *((__global uchar4 *)(dst + dst_index)); - short4 tmp = convert_short4_sat(src1_data) + convert_short4_sat(src2_data); + short4 tmp = ARITHM_OP(convert_short4_sat(src1_data), convert_short4_sat(src2_data)); uchar4 tmp_data = convert_uchar4_sat(tmp); dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x; @@ -134,7 +139,7 @@ __kernel void arithm_add_D2 (__global ushort *src1, int src1_step, int src1_offs ushort4 src2_data = vload4(0, (__global ushort *)((__global char *)src2 + src2_index)); ushort4 dst_data = *((__global ushort4 *)((__global char *)dst + dst_index)); - int4 tmp = convert_int4_sat(src1_data) + convert_int4_sat(src2_data); + int4 tmp = ARITHM_OP(convert_int4_sat(src1_data), convert_int4_sat(src2_data)); ushort4 tmp_data = convert_ushort4_sat(tmp); dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x; @@ -172,7 +177,7 @@ __kernel void arithm_add_D3 (__global short *src1, int src1_step, int src1_offse short4 src2_data = vload4(0, (__global short *)((__global char *)src2 + src2_index)); short4 dst_data = *((__global short4 *)((__global char *)dst + dst_index)); - int4 tmp = convert_int4_sat(src1_data) + convert_int4_sat(src2_data); + int4 tmp = ARITHM_OP(convert_int4_sat(src1_data), convert_int4_sat(src2_data)); short4 tmp_data = convert_short4_sat(tmp); dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x; @@ -200,7 +205,7 @@ __kernel void arithm_add_D4 (__global int *src1, int src1_step, int src1_offset, int data1 = *((__global int *)((__global char *)src1 + src1_index)); int data2 = *((__global int *)((__global char *)src2 + src2_index)); - long tmp = (long)(data1) + (long)(data2); + long tmp = ARITHM_OP((long)(data1), (long)(data2)); *((__global int *)((__global char *)dst + dst_index)) = convert_int_sat(tmp); } @@ -221,7 +226,7 @@ __kernel void arithm_add_D5 (__global float *src1, int src1_step, int src1_offse float data1 = *((__global float *)((__global char *)src1 + src1_index)); float data2 = *((__global float *)((__global char *)src2 + src2_index)); - float tmp = data1 + data2; + float tmp = ARITHM_OP(data1, data2); *((__global float *)((__global char *)dst + dst_index)) = tmp; } @@ -245,7 +250,7 @@ __kernel void arithm_add_D6 (__global double *src1, int src1_step, int src1_offs double data1 = *((__global double *)((__global char *)src1 + src1_index)); double data2 = *((__global double *)((__global char *)src2 + src2_index)); - *((__global double *)((__global char *)dst + dst_index)) = data1 + data2; + *((__global double *)((__global char *)dst + dst_index)) = ARITHM_OP(data1, data2); } } #endif @@ -302,7 +307,7 @@ __kernel void arithm_add_with_mask_C1_D0 (__global uchar *src1, int src1_step, i } uchar4 data = *((__global uchar4 *)(dst + dst_index)); - short4 tmp = convert_short4_sat(src1_data) + convert_short4_sat(src2_data); + short4 tmp = ARITHM_OP(convert_short4_sat(src1_data), convert_short4_sat(src2_data)); uchar4 tmp_data = convert_uchar4_sat(tmp); data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x; @@ -344,7 +349,7 @@ __kernel void arithm_add_with_mask_C1_D2 (__global ushort *src1, int src1_step, uchar2 mask_data = vload2(0, mask + mask_index); ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index)); - int2 tmp = convert_int2_sat(src1_data) + convert_int2_sat(src2_data); + int2 tmp = ARITHM_OP(convert_int2_sat(src1_data), convert_int2_sat(src2_data)); ushort2 tmp_data = convert_ushort2_sat(tmp); data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x; @@ -384,7 +389,7 @@ __kernel void arithm_add_with_mask_C1_D3 (__global short *src1, int src1_step, i uchar2 mask_data = vload2(0, mask + mask_index); short2 data = *((__global short2 *)((__global uchar *)dst + dst_index)); - int2 tmp = convert_int2_sat(src1_data) + convert_int2_sat(src2_data); + int2 tmp = ARITHM_OP(convert_int2_sat(src1_data), convert_int2_sat(src2_data)); short2 tmp_data = convert_short2_sat(tmp); data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x; @@ -416,7 +421,7 @@ __kernel void arithm_add_with_mask_C1_D4 (__global int *src1, int src1_step, i int src_data2 = *((__global int *)((__global char *)src2 + src2_index)); int dst_data = *((__global int *)((__global char *)dst + dst_index)); - int data = convert_int_sat((long)src_data1 + (long)src_data2); + int data = convert_int_sat(ARITHM_OP((long)src_data1, (long)src_data2)); data = mask_data ? data : dst_data; *((__global int *)((__global char *)dst + dst_index)) = data; @@ -446,7 +451,7 @@ __kernel void arithm_add_with_mask_C1_D5 (__global float *src1, int src1_step, i float src_data2 = *((__global float *)((__global char *)src2 + src2_index)); float dst_data = *((__global float *)((__global char *)dst + dst_index)); - float data = src_data1 + src_data2; + float data = ARITHM_OP(src_data1, src_data2); data = mask_data ? data : dst_data; *((__global float *)((__global char *)dst + dst_index)) = data; @@ -477,7 +482,7 @@ __kernel void arithm_add_with_mask_C1_D6 (__global double *src1, int src1_step, double src_data2 = *((__global double *)((__global char *)src2 + src2_index)); double dst_data = *((__global double *)((__global char *)dst + dst_index)); - double data = src_data1 + src_data2; + double data = ARITHM_OP(src_data1, src_data2); data = mask_data ? data : dst_data; *((__global double *)((__global char *)dst + dst_index)) = data; @@ -516,7 +521,7 @@ __kernel void arithm_add_with_mask_C2_D0 (__global uchar *src1, int src1_step, i uchar2 mask_data = vload2(0, mask + mask_index); uchar4 data = *((__global uchar4 *)(dst + dst_index)); - short4 tmp = convert_short4_sat(src1_data) + convert_short4_sat(src2_data); + short4 tmp = ARITHM_OP(convert_short4_sat(src1_data), convert_short4_sat(src2_data)); uchar4 tmp_data = convert_uchar4_sat(tmp); data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy; @@ -548,7 +553,7 @@ __kernel void arithm_add_with_mask_C2_D2 (__global ushort *src1, int src1_step, ushort2 src_data2 = *((__global ushort2 *)((__global char *)src2 + src2_index)); ushort2 dst_data = *((__global ushort2 *)((__global char *)dst + dst_index)); - int2 tmp = convert_int2_sat(src_data1) + convert_int2_sat(src_data2); + int2 tmp = ARITHM_OP(convert_int2_sat(src_data1), convert_int2_sat(src_data2)); ushort2 data = convert_ushort2_sat(tmp); data = mask_data ? data : dst_data; @@ -578,7 +583,7 @@ __kernel void arithm_add_with_mask_C2_D3 (__global short *src1, int src1_step, i short2 src_data2 = *((__global short2 *)((__global char *)src2 + src2_index)); short2 dst_data = *((__global short2 *)((__global char *)dst + dst_index)); - int2 tmp = convert_int2_sat(src_data1) + convert_int2_sat(src_data2); + int2 tmp = ARITHM_OP(convert_int2_sat(src_data1), convert_int2_sat(src_data2)); short2 data = convert_short2_sat(tmp); data = mask_data ? data : dst_data; @@ -608,7 +613,7 @@ __kernel void arithm_add_with_mask_C2_D4 (__global int *src1, int src1_step, i int2 src_data2 = *((__global int2 *)((__global char *)src2 + src2_index)); int2 dst_data = *((__global int2 *)((__global char *)dst + dst_index)); - int2 data = convert_int2_sat(convert_long2_sat(src_data1) + convert_long2_sat(src_data2)); + int2 data = convert_int2_sat(ARITHM_OP(convert_long2_sat(src_data1), convert_long2_sat(src_data2))); data = mask_data ? data : dst_data; *((__global int2 *)((__global char *)dst + dst_index)) = data; @@ -637,7 +642,7 @@ __kernel void arithm_add_with_mask_C2_D5 (__global float *src1, int src1_step, i float2 src_data2 = *((__global float2 *)((__global char *)src2 + src2_index)); float2 dst_data = *((__global float2 *)((__global char *)dst + dst_index)); - float2 data = src_data1 + src_data2; + float2 data = ARITHM_OP(src_data1, src_data2); data = mask_data ? data : dst_data; *((__global float2 *)((__global char *)dst + dst_index)) = data; @@ -668,7 +673,7 @@ __kernel void arithm_add_with_mask_C2_D6 (__global double *src1, int src1_step, double2 src_data2 = *((__global double2 *)((__global char *)src2 + src2_index)); double2 dst_data = *((__global double2 *)((__global char *)dst + dst_index)); - double2 data = src_data1 + src_data2; + double2 data = ARITHM_OP(src_data1, src_data2); data = mask_data ? data : dst_data; *((__global double2 *)((__global char *)dst + dst_index)) = data; @@ -699,7 +704,7 @@ __kernel void arithm_add_with_mask_C4_D0 (__global uchar *src1, int src1_step, i uchar4 src_data2 = *((__global uchar4 *)(src2 + src2_index)); uchar4 dst_data = *((__global uchar4 *)(dst + dst_index)); - uchar4 data = convert_uchar4_sat(convert_ushort4_sat(src_data1) + convert_ushort4_sat(src_data2)); + uchar4 data = convert_uchar4_sat(ARITHM_OP(convert_short4_sat(src_data1), convert_short4_sat(src_data2))); data = mask_data ? data : dst_data; *((__global uchar4 *)(dst + dst_index)) = data; @@ -728,7 +733,7 @@ __kernel void arithm_add_with_mask_C4_D2 (__global ushort *src1, int src1_step, ushort4 src_data2 = *((__global ushort4 *)((__global char *)src2 + src2_index)); ushort4 dst_data = *((__global ushort4 *)((__global char *)dst + dst_index)); - ushort4 data = convert_ushort4_sat(convert_int4_sat(src_data1) + convert_int4_sat(src_data2)); + ushort4 data = convert_ushort4_sat(ARITHM_OP(convert_int4_sat(src_data1), convert_int4_sat(src_data2))); data = mask_data ? data : dst_data; *((__global ushort4 *)((__global char *)dst + dst_index)) = data; @@ -757,7 +762,7 @@ __kernel void arithm_add_with_mask_C4_D3 (__global short *src1, int src1_step, i short4 src_data2 = *((__global short4 *)((__global char *)src2 + src2_index)); short4 dst_data = *((__global short4 *)((__global char *)dst + dst_index)); - short4 data = convert_short4_sat(convert_int4_sat(src_data1) + convert_int4_sat(src_data2)); + short4 data = convert_short4_sat(ARITHM_OP(convert_int4_sat(src_data1), convert_int4_sat(src_data2))); data = mask_data ? data : dst_data; *((__global short4 *)((__global char *)dst + dst_index)) = data; @@ -786,7 +791,7 @@ __kernel void arithm_add_with_mask_C4_D4 (__global int *src1, int src1_step, i int4 src_data2 = *((__global int4 *)((__global char *)src2 + src2_index)); int4 dst_data = *((__global int4 *)((__global char *)dst + dst_index)); - int4 data = convert_int4_sat(convert_long4_sat(src_data1) + convert_long4_sat(src_data2)); + int4 data = convert_int4_sat(ARITHM_OP(convert_long4_sat(src_data1), convert_long4_sat(src_data2))); data = mask_data ? data : dst_data; *((__global int4 *)((__global char *)dst + dst_index)) = data; @@ -815,7 +820,7 @@ __kernel void arithm_add_with_mask_C4_D5 (__global float *src1, int src1_step, i float4 src_data2 = *((__global float4 *)((__global char *)src2 + src2_index)); float4 dst_data = *((__global float4 *)((__global char *)dst + dst_index)); - float4 data = src_data1 + src_data2; + float4 data = ARITHM_OP(src_data1, src_data2); data = mask_data ? data : dst_data; *((__global float4 *)((__global char *)dst + dst_index)) = data; @@ -846,7 +851,7 @@ __kernel void arithm_add_with_mask_C4_D6 (__global double *src1, int src1_step, double4 src_data2 = *((__global double4 *)((__global char *)src2 + src2_index)); double4 dst_data = *((__global double4 *)((__global char *)dst + dst_index)); - double4 data = src_data1 + src_data2; + double4 data = ARITHM_OP(src_data1, src_data2); data = mask_data ? data : dst_data; *((__global double4 *)((__global char *)dst + dst_index)) = data; diff --git a/modules/ocl/src/opencl/arithm_add_scalar.cl b/modules/ocl/src/opencl/arithm_add_scalar.cl index 152b5a19fb..0552fc8a7b 100644 --- a/modules/ocl/src/opencl/arithm_add_scalar.cl +++ b/modules/ocl/src/opencl/arithm_add_scalar.cl @@ -49,7 +49,12 @@ #elif defined (cl_amd_fp64) #pragma OPENCL EXTENSION cl_amd_fp64:enable #endif +#endif +#ifdef ARITHM_ADD + #define ARITHM_OP(A,B) ((A)+(B)) +#elif defined ARITHM_SUB + #define ARITHM_OP(A,B) ((A)-(B)) #endif /**************************************add with scalar without mask**************************************/ __kernel void arithm_s_add_C1_D0 (__global uchar *src1, int src1_step, int src1_offset, @@ -83,7 +88,7 @@ __kernel void arithm_s_add_C1_D0 (__global uchar *src1, int src1_step, int src } uchar4 data = *((__global uchar4 *)(dst + dst_index)); - int4 tmp = convert_int4_sat(src1_data) + src2_data; + int4 tmp = ARITHM_OP(convert_int4_sat(src1_data), src2_data); uchar4 tmp_data = convert_uchar4_sat(tmp); data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x; @@ -120,7 +125,7 @@ __kernel void arithm_s_add_C1_D2 (__global ushort *src1, int src1_step, int sr int2 src2_data = (int2)(src2.x, src2.x); ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index)); - int2 tmp = convert_int2_sat(src1_data) + src2_data; + int2 tmp = ARITHM_OP(convert_int2_sat(src1_data), src2_data); ushort2 tmp_data = convert_ushort2_sat(tmp); data.x = (dst_index + 0 >= dst_start) ? tmp_data.x : data.x; @@ -155,7 +160,7 @@ __kernel void arithm_s_add_C1_D3 (__global short *src1, int src1_step, int src int2 src2_data = (int2)(src2.x, src2.x); short2 data = *((__global short2 *)((__global uchar *)dst + dst_index)); - int2 tmp = convert_int2_sat(src1_data) + src2_data; + int2 tmp = ARITHM_OP(convert_int2_sat(src1_data), src2_data); short2 tmp_data = convert_short2_sat(tmp); data.x = (dst_index + 0 >= dst_start) ? tmp_data.x : data.x; @@ -181,7 +186,7 @@ __kernel void arithm_s_add_C1_D4 (__global int *src1, int src1_step, int src1_ int src_data2 = src2.x; int dst_data = *((__global int *)((__global char *)dst + dst_index)); - int data = convert_int_sat((long)src_data1 + (long)src_data2); + int data = convert_int_sat(ARITHM_OP((long)src_data1, (long)src_data2)); *((__global int *)((__global char *)dst + dst_index)) = data; } @@ -203,7 +208,7 @@ __kernel void arithm_s_add_C1_D5 (__global float *src1, int src1_step, int src float src_data2 = src2.x; float dst_data = *((__global float *)((__global char *)dst + dst_index)); - float data = src_data1 + src_data2; + float data = ARITHM_OP(src_data1, src_data2); *((__global float *)((__global char *)dst + dst_index)) = data; } @@ -227,7 +232,7 @@ __kernel void arithm_s_add_C1_D6 (__global double *src1, int src1_step, int sr double src2_data = src2.x; double dst_data = *((__global double *)((__global char *)dst + dst_index)); - double data = src_data1 + src2_data; + double data = ARITHM_OP(src_data1, src2_data); *((__global double *)((__global char *)dst + dst_index)) = data; } @@ -260,7 +265,7 @@ __kernel void arithm_s_add_C2_D0 (__global uchar *src1, int src1_step, int src int4 src2_data = (int4)(src2.x, src2.y, src2.x, src2.y); uchar4 data = *((__global uchar4 *)(dst + dst_index)); - int4 tmp = convert_int4_sat(src1_data) + src2_data; + int4 tmp = ARITHM_OP(convert_int4_sat(src1_data), src2_data); uchar4 tmp_data = convert_uchar4_sat(tmp); data.xy = (dst_index + 0 >= dst_start) ? tmp_data.xy : data.xy; @@ -286,7 +291,7 @@ __kernel void arithm_s_add_C2_D2 (__global ushort *src1, int src1_step, int sr int2 src_data2 = (int2)(src2.x, src2.y); ushort2 dst_data = *((__global ushort2 *)((__global char *)dst + dst_index)); - int2 tmp = convert_int2_sat(src_data1) + src_data2; + int2 tmp = ARITHM_OP(convert_int2_sat(src_data1), src_data2); ushort2 data = convert_ushort2_sat(tmp); *((__global ushort2 *)((__global char *)dst + dst_index)) = data; @@ -309,7 +314,7 @@ __kernel void arithm_s_add_C2_D3 (__global short *src1, int src1_step, int src int2 src_data2 = (int2)(src2.x, src2.y); short2 dst_data = *((__global short2 *)((__global char *)dst + dst_index)); - int2 tmp = convert_int2_sat(src_data1) + src_data2; + int2 tmp = ARITHM_OP(convert_int2_sat(src_data1), src_data2); short2 data = convert_short2_sat(tmp); *((__global short2 *)((__global char *)dst + dst_index)) = data; @@ -332,7 +337,7 @@ __kernel void arithm_s_add_C2_D4 (__global int *src1, int src1_step, int src1_ int2 src_data2 = (int2)(src2.x, src2.y); int2 dst_data = *((__global int2 *)((__global char *)dst + dst_index)); - int2 data = convert_int2_sat(convert_long2_sat(src_data1) + convert_long2_sat(src_data2)); + int2 data = convert_int2_sat(ARITHM_OP(convert_long2_sat(src_data1), convert_long2_sat(src_data2))); *((__global int2 *)((__global char *)dst + dst_index)) = data; } } @@ -353,7 +358,7 @@ __kernel void arithm_s_add_C2_D5 (__global float *src1, int src1_step, int src float2 src_data2 = (float2)(src2.x, src2.y); float2 dst_data = *((__global float2 *)((__global char *)dst + dst_index)); - float2 data = src_data1 + src_data2; + float2 data = ARITHM_OP(src_data1, src_data2); *((__global float2 *)((__global char *)dst + dst_index)) = data; } } @@ -376,7 +381,7 @@ __kernel void arithm_s_add_C2_D6 (__global double *src1, int src1_step, int sr double2 src_data2 = (double2)(src2.x, src2.y); double2 dst_data = *((__global double2 *)((__global char *)dst + dst_index)); - double2 data = src_data1 + src_data2; + double2 data = ARITHM_OP(src_data1, src_data2); *((__global double2 *)((__global char *)dst + dst_index)) = data; } @@ -398,7 +403,7 @@ __kernel void arithm_s_add_C4_D0 (__global uchar *src1, int src1_step, int src uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index)); - uchar4 data = convert_uchar4_sat(convert_int4_sat(src_data1) + src2); + uchar4 data = convert_uchar4_sat(ARITHM_OP(convert_int4_sat(src_data1), src2)); *((__global uchar4 *)(dst + dst_index)) = data; } @@ -418,7 +423,7 @@ __kernel void arithm_s_add_C4_D2 (__global ushort *src1, int src1_step, int sr ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index)); - ushort4 data = convert_ushort4_sat(convert_int4_sat(src_data1) + src2); + ushort4 data = convert_ushort4_sat(ARITHM_OP(convert_int4_sat(src_data1), src2)); *((__global ushort4 *)((__global char *)dst + dst_index)) = data; } @@ -438,7 +443,7 @@ __kernel void arithm_s_add_C4_D3 (__global short *src1, int src1_step, int src short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index)); - short4 data = convert_short4_sat(convert_int4_sat(src_data1) + src2); + short4 data = convert_short4_sat(ARITHM_OP(convert_int4_sat(src_data1), src2)); *((__global short4 *)((__global char *)dst + dst_index)) = data; } @@ -458,7 +463,7 @@ __kernel void arithm_s_add_C4_D4 (__global int *src1, int src1_step, int src1_ int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index)); - int4 data = convert_int4_sat(convert_long4_sat(src_data1) + convert_long4_sat(src2)); + int4 data = convert_int4_sat(ARITHM_OP(convert_long4_sat(src_data1), convert_long4_sat(src2))); *((__global int4 *)((__global char *)dst + dst_index)) = data; } @@ -478,7 +483,7 @@ __kernel void arithm_s_add_C4_D5 (__global float *src1, int src1_step, int src float4 src_data1 = *((__global float4 *)((__global char *)src1 + src1_index)); - float4 data = src_data1 + src2; + float4 data = ARITHM_OP(src_data1, src2); *((__global float4 *)((__global char *)dst + dst_index)) = data; } @@ -500,7 +505,7 @@ __kernel void arithm_s_add_C4_D6 (__global double *src1, int src1_step, int sr double4 src_data1 = *((__global double4 *)((__global char *)src1 + src1_index)); - double4 data = src_data1 + src2; + double4 data = ARITHM_OP(src_data1, src2); *((__global double4 *)((__global char *)dst + dst_index)) = data; } diff --git a/modules/ocl/src/opencl/arithm_add_scalar_mask.cl b/modules/ocl/src/opencl/arithm_add_scalar_mask.cl index 673e323ff6..fdf65923cd 100644 --- a/modules/ocl/src/opencl/arithm_add_scalar_mask.cl +++ b/modules/ocl/src/opencl/arithm_add_scalar_mask.cl @@ -51,6 +51,11 @@ #endif #endif +#ifdef ARITHM_ADD + #define ARITHM_OP(A,B) ((A)+(B)) +#elif defined ARITHM_SUB + #define ARITHM_OP(A,B) ((A)-(B)) +#endif /**************************************add with scalar with mask**************************************/ __kernel void arithm_s_add_with_mask_C1_D0 (__global uchar *src1, int src1_step, int src1_offset, __global uchar *dst, int dst_step, int dst_offset, @@ -94,7 +99,7 @@ __kernel void arithm_s_add_with_mask_C1_D0 (__global uchar *src1, int src1_ste } uchar4 data = *((__global uchar4 *)(dst + dst_index)); - int4 tmp = convert_int4_sat(src1_data) + src2_data; + int4 tmp = ARITHM_OP(convert_int4_sat(src1_data), src2_data); uchar4 tmp_data = convert_uchar4_sat(tmp); data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x; @@ -134,7 +139,7 @@ __kernel void arithm_s_add_with_mask_C1_D2 (__global ushort *src1, int src1_st uchar2 mask_data = vload2(0, mask + mask_index); ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index)); - int2 tmp = convert_int2_sat(src1_data) + src2_data; + int2 tmp = ARITHM_OP(convert_int2_sat(src1_data), src2_data); ushort2 tmp_data = convert_ushort2_sat(tmp); data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x; @@ -172,7 +177,7 @@ __kernel void arithm_s_add_with_mask_C1_D3 (__global short *src1, int src1_ste uchar2 mask_data = vload2(0, mask + mask_index); short2 data = *((__global short2 *)((__global uchar *)dst + dst_index)); - int2 tmp = convert_int2_sat(src1_data) + src2_data; + int2 tmp = ARITHM_OP(convert_int2_sat(src1_data), src2_data); short2 tmp_data = convert_short2_sat(tmp); data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x; @@ -202,7 +207,7 @@ __kernel void arithm_s_add_with_mask_C1_D4 (__global int *src1, int src1_ste int src_data2 = src2.x; int dst_data = *((__global int *)((__global char *)dst + dst_index)); - int data = convert_int_sat((long)src_data1 + (long)src_data2); + int data = convert_int_sat(ARITHM_OP((long)src_data1, (long)src_data2)); data = mask_data ? data : dst_data; *((__global int *)((__global char *)dst + dst_index)) = data; @@ -230,7 +235,7 @@ __kernel void arithm_s_add_with_mask_C1_D5 (__global float *src1, int src1_s float src_data2 = src2.x; float dst_data = *((__global float *)((__global char *)dst + dst_index)); - float data = src_data1 + src_data2; + float data = ARITHM_OP(src_data1, src_data2); data = mask_data ? data : dst_data; *((__global float *)((__global char *)dst + dst_index)) = data; @@ -260,7 +265,7 @@ __kernel void arithm_s_add_with_mask_C1_D6 (__global double *src1, int src1_ double src_data2 = src2.x; double dst_data = *((__global double *)((__global char *)dst + dst_index)); - double data = src_data1 + src_data2; + double data = ARITHM_OP(src_data1, src_data2); data = mask_data ? data : dst_data; *((__global double *)((__global char *)dst + dst_index)) = data; @@ -296,7 +301,7 @@ __kernel void arithm_s_add_with_mask_C2_D0 (__global uchar *src1, int src1_ste uchar2 mask_data = vload2(0, mask + mask_index); uchar4 data = *((__global uchar4 *)(dst + dst_index)); - int4 tmp = convert_int4_sat(src1_data) + src2_data; + int4 tmp = ARITHM_OP(convert_int4_sat(src1_data), src2_data); uchar4 tmp_data = convert_uchar4_sat(tmp); data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy; @@ -326,7 +331,7 @@ __kernel void arithm_s_add_with_mask_C2_D2 (__global ushort *src1, int src1_st int2 src_data2 = (int2)(src2.x, src2.y); ushort2 dst_data = *((__global ushort2 *)((__global char *)dst + dst_index)); - int2 tmp = convert_int2_sat(src_data1) + src_data2; + int2 tmp = ARITHM_OP(convert_int2_sat(src_data1), src_data2); ushort2 data = convert_ushort2_sat(tmp); data = mask_data ? data : dst_data; @@ -354,7 +359,7 @@ __kernel void arithm_s_add_with_mask_C2_D3 (__global short *src1, int src1_ste int2 src_data2 = (int2)(src2.x, src2.y); short2 dst_data = *((__global short2 *)((__global char *)dst + dst_index)); - int2 tmp = convert_int2_sat(src_data1) + src_data2; + int2 tmp = ARITHM_OP(convert_int2_sat(src_data1), src_data2); short2 data = convert_short2_sat(tmp); data = mask_data ? data : dst_data; @@ -382,7 +387,7 @@ __kernel void arithm_s_add_with_mask_C2_D4 (__global int *src1, int src1_step, int2 src_data2 = (int2)(src2.x, src2.y); int2 dst_data = *((__global int2 *)((__global char *)dst + dst_index)); - int2 data = convert_int2_sat(convert_long2_sat(src_data1) + convert_long2_sat(src_data2)); + int2 data = convert_int2_sat(ARITHM_OP(convert_long2_sat(src_data1), convert_long2_sat(src_data2))); data = mask_data ? data : dst_data; *((__global int2 *)((__global char *)dst + dst_index)) = data; @@ -409,7 +414,7 @@ __kernel void arithm_s_add_with_mask_C2_D5 (__global float *src1, int src1_ste float2 src_data2 = (float2)(src2.x, src2.y); float2 dst_data = *((__global float2 *)((__global char *)dst + dst_index)); - float2 data = src_data1 + src_data2; + float2 data = ARITHM_OP(src_data1, src_data2); data = mask_data ? data : dst_data; *((__global float2 *)((__global char *)dst + dst_index)) = data; @@ -438,7 +443,7 @@ __kernel void arithm_s_add_with_mask_C2_D6 (__global double *src1, int src1_st double2 src_data2 = (double2)(src2.x, src2.y); double2 dst_data = *((__global double2 *)((__global char *)dst + dst_index)); - double2 data = src_data1 + src_data2; + double2 data = ARITHM_OP(src_data1, src_data2); data = mask_data ? data : dst_data; *((__global double2 *)((__global char *)dst + dst_index)) = data; @@ -451,7 +456,6 @@ __kernel void arithm_s_add_with_mask_C4_D0 (__global uchar *src1, int src1_ste __global uchar *mask, int mask_step, int mask_offset, int4 src2, int rows, int cols, int dst_step1) { - int x = get_global_id(0); int y = get_global_id(1); @@ -466,7 +470,7 @@ __kernel void arithm_s_add_with_mask_C4_D0 (__global uchar *src1, int src1_ste uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index)); uchar4 dst_data = *((__global uchar4 *)(dst + dst_index)); - uchar4 data = convert_uchar4_sat(convert_int4_sat(src_data1) + src2); + uchar4 data = convert_uchar4_sat(ARITHM_OP(convert_int4_sat(src_data1), src2)); data = mask_data ? data : dst_data; *((__global uchar4 *)(dst + dst_index)) = data; @@ -492,7 +496,7 @@ __kernel void arithm_s_add_with_mask_C4_D2 (__global ushort *src1, int src1_st ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index)); ushort4 dst_data = *((__global ushort4 *)((__global char *)dst + dst_index)); - ushort4 data = convert_ushort4_sat(convert_int4_sat(src_data1) + src2); + ushort4 data = convert_ushort4_sat(ARITHM_OP(convert_int4_sat(src_data1), src2)); data = mask_data ? data : dst_data; *((__global ushort4 *)((__global char *)dst + dst_index)) = data; @@ -518,7 +522,7 @@ __kernel void arithm_s_add_with_mask_C4_D3 (__global short *src1, int src1_ste short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index)); short4 dst_data = *((__global short4 *)((__global char *)dst + dst_index)); - short4 data = convert_short4_sat(convert_int4_sat(src_data1) + src2); + short4 data = convert_short4_sat(ARITHM_OP(convert_int4_sat(src_data1), src2)); data = mask_data ? data : dst_data; *((__global short4 *)((__global char *)dst + dst_index)) = data; @@ -544,7 +548,7 @@ __kernel void arithm_s_add_with_mask_C4_D4 (__global int *src1, int src1_step, int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index)); int4 dst_data = *((__global int4 *)((__global char *)dst + dst_index)); - int4 data = convert_int4_sat(convert_long4_sat(src_data1) + convert_long4_sat(src2)); + int4 data = convert_int4_sat(ARITHM_OP(convert_long4_sat(src_data1), convert_long4_sat(src2))); data = mask_data ? data : dst_data; *((__global int4 *)((__global char *)dst + dst_index)) = data; @@ -570,7 +574,7 @@ __kernel void arithm_s_add_with_mask_C4_D5 (__global float *src1, int src1_ste float4 src_data1 = *((__global float4 *)((__global char *)src1 + src1_index)); float4 dst_data = *((__global float4 *)((__global char *)dst + dst_index)); - float4 data = src_data1 + src2; + float4 data = ARITHM_OP(src_data1, src2); data = mask_data ? data : dst_data; *((__global float4 *)((__global char *)dst + dst_index)) = data; @@ -598,7 +602,7 @@ __kernel void arithm_s_add_with_mask_C4_D6 (__global double *src1, int src1_st double4 src_data1 = *((__global double4 *)((__global char *)src1 + src1_index)); double4 dst_data = *((__global double4 *)((__global char *)dst + dst_index)); - double4 data = src_data1 + src2; + double4 data = ARITHM_OP(src_data1, src2); data = mask_data ? data : dst_data; *((__global double4 *)((__global char *)dst + dst_index)) = data; diff --git a/modules/ocl/src/opencl/arithm_sub.cl b/modules/ocl/src/opencl/arithm_sub.cl deleted file mode 100644 index d461d3aaea..0000000000 --- a/modules/ocl/src/opencl/arithm_sub.cl +++ /dev/null @@ -1,802 +0,0 @@ -/*M/////////////////////////////////////////////////////////////////////////////////////// -// -// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING. -// -// By downloading, copying, installing or using the software you agree to this license. -// If you do not agree to this license, do not download, install, -// copy or use the software. -// -// -// License Agreement -// For Open Source Computer Vision Library -// -// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved. -// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved. -// Third party copyrights are property of their respective owners. -// -// @Authors -// Jia Haipeng, jiahaipeng95@gmail.com -// -// Redistribution and use in source and binary forms, with or without modification, -// are permitted provided that the following conditions are met: -// -// * Redistribution's of source code must retain the above copyright notice, -// this list of conditions and the following disclaimer. -// -// * Redistribution's in binary form must reproduce the above copyright notice, -// this list of conditions and the following disclaimer in the documentation -// and/or other GpuMaterials provided with the distribution. -// -// * The name of the copyright holders may not be used to endorse or promote products -// derived from this software without specific prior written permission. -// -// This software is provided by the copyright holders and contributors as is and -// any express or implied warranties, including, but not limited to, the implied -// warranties of merchantability and fitness for a particular purpose are disclaimed. -// In no event shall the Intel Corporation or contributors be liable for any direct, -// indirect, incidental, special, exemplary, or consequential damages -// (including, but not limited to, procurement of substitute goods or services; -// loss of use, data, or profits; or business interruption) however caused -// and on any theory of liability, whether in contract, strict liability, -// or tort (including negligence or otherwise) arising in any way out of -// the use of this software, even if advised of the possibility of such damage. -// -//M*/ - -#if defined (DOUBLE_SUPPORT) -#ifdef cl_khr_fp64 -#pragma OPENCL EXTENSION cl_khr_fp64:enable -#elif defined (cl_amd_fp64) -#pragma OPENCL EXTENSION cl_amd_fp64:enable -#endif -#endif - -////////////////////////////////////////////////////////////////////////////////////////////////////// -/////////////////////////////////////////////SUB//////////////////////////////////////////////////// -/////////////////////////////////////////////////////////////////////////////////////////////////////// -/**************************************sub without mask**************************************/ -__kernel void arithm_sub_D0 (__global uchar *src1, int src1_step, int src1_offset, - __global uchar *src2, int src2_step, int src2_offset, - __global uchar *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - - #define dst_align (dst_offset & 3) - int src1_index = mad24(y, src1_step, x + src1_offset - dst_align); - int src2_index = mad24(y, src2_step, x + src2_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc); - - uchar4 src1_data = vload4(0, src1 + src1_index); - uchar4 src2_data = vload4(0, src2 + src2_index); - - uchar4 dst_data = *((__global uchar4 *)(dst + dst_index)); - short4 tmp = convert_short4_sat(src1_data) - convert_short4_sat(src2_data); - uchar4 tmp_data = convert_uchar4_sat(tmp); - - dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x; - dst_data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : dst_data.y; - dst_data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : dst_data.z; - dst_data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : dst_data.w; - - *((__global uchar4 *)(dst + dst_index)) = dst_data; - } -} -__kernel void arithm_sub_D2 (__global ushort *src1, int src1_step, int src1_offset, - __global ushort *src2, int src2_step, int src2_offset, - __global ushort *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - - #define dst_align ((dst_offset >> 1) & 3) - int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); - int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1)); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffff8); - - ushort4 src1_data = vload4(0, (__global ushort *)((__global char *)src1 + src1_index)); - ushort4 src2_data = vload4(0, (__global ushort *)((__global char *)src2 + src2_index)); - - ushort4 dst_data = *((__global ushort4 *)((__global char *)dst + dst_index)); - int4 tmp = convert_int4_sat(src1_data) - convert_int4_sat(src2_data); - ushort4 tmp_data = convert_ushort4_sat(tmp); - - dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x; - dst_data.y = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : dst_data.y; - dst_data.z = ((dst_index + 4 >= dst_start) && (dst_index + 4 < dst_end)) ? tmp_data.z : dst_data.z; - dst_data.w = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) ? tmp_data.w : dst_data.w; - - *((__global ushort4 *)((__global char *)dst + dst_index)) = dst_data; - } -} -__kernel void arithm_sub_D3 (__global short *src1, int src1_step, int src1_offset, - __global short *src2, int src2_step, int src2_offset, - __global short *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - - #define dst_align ((dst_offset >> 1) & 3) - int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); - int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1)); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffff8); - - short4 src1_data = vload4(0, (__global short *)((__global char *)src1 + src1_index)); - short4 src2_data = vload4(0, (__global short *)((__global char *)src2 + src2_index)); - - short4 dst_data = *((__global short4 *)((__global char *)dst + dst_index)); - int4 tmp = convert_int4_sat(src1_data) - convert_int4_sat(src2_data); - short4 tmp_data = convert_short4_sat(tmp); - - dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x; - dst_data.y = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : dst_data.y; - dst_data.z = ((dst_index + 4 >= dst_start) && (dst_index + 4 < dst_end)) ? tmp_data.z : dst_data.z; - dst_data.w = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) ? tmp_data.w : dst_data.w; - - *((__global short4 *)((__global char *)dst + dst_index)) = dst_data; - } -} - -__kernel void arithm_sub_D4 (__global int *src1, int src1_step, int src1_offset, - __global int *src2, int src2_step, int src2_offset, - __global int *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 2) + src1_offset); - int src2_index = mad24(y, src2_step, (x << 2) + src2_offset); - int dst_index = mad24(y, dst_step, (x << 2) + dst_offset); - - int data1 = *((__global int *)((__global char *)src1 + src1_index)); - int data2 = *((__global int *)((__global char *)src2 + src2_index)); - long tmp = (long)(data1) - (long)(data2); - - *((__global int *)((__global char *)dst + dst_index)) = convert_int_sat(tmp); - } -} -__kernel void arithm_sub_D5 (__global float *src1, int src1_step, int src1_offset, - __global float *src2, int src2_step, int src2_offset, - __global float *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 2) + src1_offset); - int src2_index = mad24(y, src2_step, (x << 2) + src2_offset); - int dst_index = mad24(y, dst_step, (x << 2) + dst_offset); - - float data1 = *((__global float *)((__global char *)src1 + src1_index)); - float data2 = *((__global float *)((__global char *)src2 + src2_index)); - float tmp = data1 - data2; - - *((__global float *)((__global char *)dst + dst_index)) = tmp; - } -} - - -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_sub_D6 (__global double *src1, int src1_step, int src1_offset, - __global double *src2, int src2_step, int src2_offset, - __global double *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 3) + src1_offset); - int src2_index = mad24(y, src2_step, (x << 3) + src2_offset); - int dst_index = mad24(y, dst_step, (x << 3) + dst_offset); - - double data1 = *((__global double *)((__global char *)src1 + src1_index)); - double data2 = *((__global double *)((__global char *)src2 + src2_index)); - - *((__global double *)((__global char *)dst + dst_index)) = data1 - data2; - } -} -#endif - -/**************************************sub with mask**************************************/ -__kernel void arithm_sub_with_mask_C1_D0 (__global uchar *src1, int src1_step, int src1_offset, - __global uchar *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global uchar *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - - #define dst_align (dst_offset & 3) - int src1_index = mad24(y, src1_step, x + src1_offset - dst_align); - int src2_index = mad24(y, src2_step, x + src2_offset - dst_align); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc); - - uchar4 src1_data = vload4(0, src1 + src1_index); - uchar4 src2_data = vload4(0, src2 + src2_index); - uchar4 mask_data = vload4(0, mask + mask_index); - - uchar4 data = *((__global uchar4 *)(dst + dst_index)); - short4 tmp = convert_short4_sat(src1_data) - convert_short4_sat(src2_data); - uchar4 tmp_data = convert_uchar4_sat(tmp); - - data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x; - data.y = ((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y; - data.z = ((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z; - data.w = ((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w; - - *((__global uchar4 *)(dst + dst_index)) = data; - } -} -__kernel void arithm_sub_with_mask_C1_D2 (__global ushort *src1, int src1_step, int src1_offset, - __global ushort *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global ushort *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - - #define dst_align ((dst_offset >> 1) & 1) - int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); - int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc); - - ushort2 src1_data = vload2(0, (__global ushort *)((__global char *)src1 + src1_index)); - ushort2 src2_data = vload2(0, (__global ushort *)((__global char *)src2 + src2_index)); - uchar2 mask_data = vload2(0, mask + mask_index); - - ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index)); - int2 tmp = convert_int2_sat(src1_data) - convert_int2_sat(src2_data); - ushort2 tmp_data = convert_ushort2_sat(tmp); - - data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x; - data.y = ((mask_data.y) && (dst_index + 2 < dst_end )) ? tmp_data.y : data.y; - - *((__global ushort2 *)((__global uchar *)dst + dst_index)) = data; - } -} -__kernel void arithm_sub_with_mask_C1_D3 (__global short *src1, int src1_step, int src1_offset, - __global short *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global short *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - - #define dst_align ((dst_offset >> 1) & 1) - int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); - int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc); - - short2 src1_data = vload2(0, (__global short *)((__global char *)src1 + src1_index)); - short2 src2_data = vload2(0, (__global short *)((__global char *)src2 + src2_index)); - uchar2 mask_data = vload2(0, mask + mask_index); - - short2 data = *((__global short2 *)((__global uchar *)dst + dst_index)); - int2 tmp = convert_int2_sat(src1_data) - convert_int2_sat(src2_data); - short2 tmp_data = convert_short2_sat(tmp); - - data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x; - data.y = ((mask_data.y) && (dst_index + 2 < dst_end )) ? tmp_data.y : data.y; - - *((__global short2 *)((__global uchar *)dst + dst_index)) = data; - } -} -__kernel void arithm_sub_with_mask_C1_D4 (__global int *src1, int src1_step, int src1_offset, - __global int *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global int *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 2) + src1_offset); - int src2_index = mad24(y, src2_step, (x << 2) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 2) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - int src_data1 = *((__global int *)((__global char *)src1 + src1_index)); - int src_data2 = *((__global int *)((__global char *)src2 + src2_index)); - int dst_data = *((__global int *)((__global char *)dst + dst_index)); - - int data = convert_int_sat((long)src_data1 - (long)src_data2); - data = mask_data ? data : dst_data; - - *((__global int *)((__global char *)dst + dst_index)) = data; - } -} - -__kernel void arithm_sub_with_mask_C1_D5 (__global float *src1, int src1_step, int src1_offset, - __global float *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global float *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 2) + src1_offset); - int src2_index = mad24(y, src2_step, (x << 2) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 2) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - float src_data1 = *((__global float *)((__global char *)src1 + src1_index)); - float src_data2 = *((__global float *)((__global char *)src2 + src2_index)); - float dst_data = *((__global float *)((__global char *)dst + dst_index)); - - float data = src_data1 - src_data2; - data = mask_data ? data : dst_data; - - *((__global float *)((__global char *)dst + dst_index)) = data; - } -} - - -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_sub_with_mask_C1_D6 (__global double *src1, int src1_step, int src1_offset, - __global double *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global double *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 3) + src1_offset); - int src2_index = mad24(y, src2_step, (x << 3) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 3) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - double src_data1 = *((__global double *)((__global char *)src1 + src1_index)); - double src_data2 = *((__global double *)((__global char *)src2 + src2_index)); - double dst_data = *((__global double *)((__global char *)dst + dst_index)); - - double data = src_data1 - src_data2; - data = mask_data ? data : dst_data; - - *((__global double *)((__global char *)dst + dst_index)) = data; - } -} -#endif - -__kernel void arithm_sub_with_mask_C2_D0 (__global uchar *src1, int src1_step, int src1_offset, - __global uchar *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global uchar *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - - #define dst_align ((dst_offset >> 1) & 1) - int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); - int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc); - - uchar4 src1_data = vload4(0, src1 + src1_index); - uchar4 src2_data = vload4(0, src2 + src2_index); - uchar2 mask_data = vload2(0, mask + mask_index); - - uchar4 data = *((__global uchar4 *)(dst + dst_index)); - short4 tmp = convert_short4_sat(src1_data) - convert_short4_sat(src2_data); - uchar4 tmp_data = convert_uchar4_sat(tmp); - - data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy; - data.zw = ((mask_data.y) && (dst_index + 2 < dst_end )) ? tmp_data.zw : data.zw; - - *((__global uchar4 *)(dst + dst_index)) = data; - } -} -__kernel void arithm_sub_with_mask_C2_D2 (__global ushort *src1, int src1_step, int src1_offset, - __global ushort *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global ushort *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 2) + src1_offset); - int src2_index = mad24(y, src2_step, (x << 2) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 2) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index)); - ushort2 src_data2 = *((__global ushort2 *)((__global char *)src2 + src2_index)); - ushort2 dst_data = *((__global ushort2 *)((__global char *)dst + dst_index)); - - int2 tmp = convert_int2_sat(src_data1) - convert_int2_sat(src_data2); - ushort2 data = convert_ushort2_sat(tmp); - data = mask_data ? data : dst_data; - - *((__global ushort2 *)((__global char *)dst + dst_index)) = data; - } -} -__kernel void arithm_sub_with_mask_C2_D3 (__global short *src1, int src1_step, int src1_offset, - __global short *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global short *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 2) + src1_offset); - int src2_index = mad24(y, src2_step, (x << 2) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 2) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index)); - short2 src_data2 = *((__global short2 *)((__global char *)src2 + src2_index)); - short2 dst_data = *((__global short2 *)((__global char *)dst + dst_index)); - - int2 tmp = convert_int2_sat(src_data1) - convert_int2_sat(src_data2); - short2 data = convert_short2_sat(tmp); - data = mask_data ? data : dst_data; - - *((__global short2 *)((__global char *)dst + dst_index)) = data; - } -} -__kernel void arithm_sub_with_mask_C2_D4 (__global int *src1, int src1_step, int src1_offset, - __global int *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global int *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 3) + src1_offset); - int src2_index = mad24(y, src2_step, (x << 3) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 3) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index)); - int2 src_data2 = *((__global int2 *)((__global char *)src2 + src2_index)); - int2 dst_data = *((__global int2 *)((__global char *)dst + dst_index)); - - int2 data = convert_int2_sat(convert_long2_sat(src_data1) - convert_long2_sat(src_data2)); - data = mask_data ? data : dst_data; - - *((__global int2 *)((__global char *)dst + dst_index)) = data; - } -} -__kernel void arithm_sub_with_mask_C2_D5 (__global float *src1, int src1_step, int src1_offset, - __global float *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global float *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 3) + src1_offset); - int src2_index = mad24(y, src2_step, (x << 3) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 3) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - float2 src_data1 = *((__global float2 *)((__global char *)src1 + src1_index)); - float2 src_data2 = *((__global float2 *)((__global char *)src2 + src2_index)); - float2 dst_data = *((__global float2 *)((__global char *)dst + dst_index)); - - float2 data = src_data1 - src_data2; - data = mask_data ? data : dst_data; - - *((__global float2 *)((__global char *)dst + dst_index)) = data; - } -} - -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_sub_with_mask_C2_D6 (__global double *src1, int src1_step, int src1_offset, - __global double *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global double *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 4) + src1_offset); - int src2_index = mad24(y, src2_step, (x << 4) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 4) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - double2 src_data1 = *((__global double2 *)((__global char *)src1 + src1_index)); - double2 src_data2 = *((__global double2 *)((__global char *)src2 + src2_index)); - double2 dst_data = *((__global double2 *)((__global char *)dst + dst_index)); - - double2 data = src_data1 - src_data2; - data = mask_data ? data : dst_data; - - *((__global double2 *)((__global char *)dst + dst_index)) = data; - } -} -#endif - -__kernel void arithm_sub_with_mask_C4_D0 (__global uchar *src1, int src1_step, int src1_offset, - __global uchar *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global uchar *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 2) + src1_offset); - int src2_index = mad24(y, src2_step, (x << 2) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 2) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index)); - uchar4 src_data2 = *((__global uchar4 *)(src2 + src2_index)); - uchar4 dst_data = *((__global uchar4 *)(dst + dst_index)); - - uchar4 data = convert_uchar4_sat(convert_short4_sat(src_data1) - convert_short4_sat(src_data2)); - data = mask_data ? data : dst_data; - - *((__global uchar4 *)(dst + dst_index)) = data; - } -} -__kernel void arithm_sub_with_mask_C4_D2 (__global ushort *src1, int src1_step, int src1_offset, - __global ushort *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global ushort *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 3) + src1_offset); - int src2_index = mad24(y, src2_step, (x << 3) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 3) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index)); - ushort4 src_data2 = *((__global ushort4 *)((__global char *)src2 + src2_index)); - ushort4 dst_data = *((__global ushort4 *)((__global char *)dst + dst_index)); - - ushort4 data = convert_ushort4_sat(convert_int4_sat(src_data1) - convert_int4_sat(src_data2)); - data = mask_data ? data : dst_data; - - *((__global ushort4 *)((__global char *)dst + dst_index)) = data; - } -} -__kernel void arithm_sub_with_mask_C4_D3 (__global short *src1, int src1_step, int src1_offset, - __global short *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global short *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 3) + src1_offset); - int src2_index = mad24(y, src2_step, (x << 3) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 3) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index)); - short4 src_data2 = *((__global short4 *)((__global char *)src2 + src2_index)); - short4 dst_data = *((__global short4 *)((__global char *)dst + dst_index)); - - short4 data = convert_short4_sat(convert_int4_sat(src_data1) - convert_int4_sat(src_data2)); - data = mask_data ? data : dst_data; - - *((__global short4 *)((__global char *)dst + dst_index)) = data; - } -} -__kernel void arithm_sub_with_mask_C4_D4 (__global int *src1, int src1_step, int src1_offset, - __global int *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global int *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 4) + src1_offset); - int src2_index = mad24(y, src2_step, (x << 4) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 4) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index)); - int4 src_data2 = *((__global int4 *)((__global char *)src2 + src2_index)); - int4 dst_data = *((__global int4 *)((__global char *)dst + dst_index)); - - int4 data = convert_int4_sat(convert_long4_sat(src_data1) - convert_long4_sat(src_data2)); - data = mask_data ? data : dst_data; - - *((__global int4 *)((__global char *)dst + dst_index)) = data; - } -} -__kernel void arithm_sub_with_mask_C4_D5 (__global float *src1, int src1_step, int src1_offset, - __global float *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global float *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 4) + src1_offset); - int src2_index = mad24(y, src2_step, (x << 4) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 4) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - float4 src_data1 = *((__global float4 *)((__global char *)src1 + src1_index)); - float4 src_data2 = *((__global float4 *)((__global char *)src2 + src2_index)); - float4 dst_data = *((__global float4 *)((__global char *)dst + dst_index)); - - float4 data = src_data1 - src_data2; - data = mask_data ? data : dst_data; - - *((__global float4 *)((__global char *)dst + dst_index)) = data; - } -} - -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_sub_with_mask_C4_D6 (__global double *src1, int src1_step, int src1_offset, - __global double *src2, int src2_step, int src2_offset, - __global uchar *mask, int mask_step, int mask_offset, - __global double *dst, int dst_step, int dst_offset, - int rows, int cols, int dst_step1) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 5) + src1_offset); - int src2_index = mad24(y, src2_step, (x << 5) + src2_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 5) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - double4 src_data1 = *((__global double4 *)((__global char *)src1 + src1_index)); - double4 src_data2 = *((__global double4 *)((__global char *)src2 + src2_index)); - double4 dst_data = *((__global double4 *)((__global char *)dst + dst_index)); - - double4 data = src_data1 - src_data2; - data = mask_data ? data : dst_data; - - *((__global double4 *)((__global char *)dst + dst_index)) = data; - } -} -#endif diff --git a/modules/ocl/src/opencl/arithm_sub_scalar.cl b/modules/ocl/src/opencl/arithm_sub_scalar.cl deleted file mode 100644 index 76bb29418c..0000000000 --- a/modules/ocl/src/opencl/arithm_sub_scalar.cl +++ /dev/null @@ -1,511 +0,0 @@ -/*M/////////////////////////////////////////////////////////////////////////////////////// -// -// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING. -// -// By downloading, copying, installing or using the software you agree to this license. -// If you do not agree to this license, do not download, install, -// copy or use the software. -// -// -// License Agreement -// For Open Source Computer Vision Library -// -// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved. -// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved. -// Third party copyrights are property of their respective owners. -// -// @Authors -// Jia Haipeng, jiahaipeng95@gmail.com -// -// Redistribution and use in source and binary forms, with or without modification, -// are permitted provided that the following conditions are met: -// -// * Redistribution's of source code must retain the above copyright notice, -// this list of conditions and the following disclaimer. -// -// * Redistribution's in binary form must reproduce the above copyright notice, -// this list of conditions and the following disclaimer in the documentation -// and/or other oclMaterials provided with the distribution. -// -// * The name of the copyright holders may not be used to endorse or promote products -// derived from this software without specific prior written permission. -// -// This software is provided by the copyright holders and contributors as is and -// any express or implied warranties, including, but not limited to, the implied -// warranties of merchantability and fitness for a particular purpose are disclaimed. -// In no event shall the Intel Corporation or contributors be liable for any direct, -// indirect, incidental, special, exemplary, or consequential damages -// (including, but not limited to, procurement of substitute goods or services; -// loss of use, data, or profits; or business interruption) however caused -// and on any theory of liability, whether in contract, strict liability, -// or tort (including negligence or otherwise) arising in any way out of -// the use of this software, even if advised of the possibility of such damage. -// -//M*/ -#if defined (DOUBLE_SUPPORT) -#ifdef cl_khr_fp64 -#pragma OPENCL EXTENSION cl_khr_fp64:enable -#elif defined (cl_amd_fp64) -#pragma OPENCL EXTENSION cl_amd_fp64:enable -#endif -#endif -/**************************************sub with scalar without mask**************************************/ -__kernel void arithm_s_sub_C1_D0 (__global uchar *src1, int src1_step, int src1_offset, - __global uchar *dst, int dst_step, int dst_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - - #define dst_align (dst_offset & 3) - int src1_index = mad24(y, src1_step, x + src1_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc); - - uchar4 src1_data = vload4(0, src1 + src1_index); - int4 src2_data = (int4)(src2.x, src2.x, src2.x, src2.x); - - uchar4 data = *((__global uchar4 *)(dst + dst_index)); - int4 tmp = convert_int4_sat(src1_data) - src2_data; - tmp = isMatSubScalar ? tmp : -tmp; - uchar4 tmp_data = convert_uchar4_sat(tmp); - - data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x; - data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y; - data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z; - data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w; - - *((__global uchar4 *)(dst + dst_index)) = data; - } -} -__kernel void arithm_s_sub_C1_D2 (__global ushort *src1, int src1_step, int src1_offset, - __global ushort *dst, int dst_step, int dst_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - - #define dst_align ((dst_offset >> 1) & 1) - int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc); - - ushort2 src1_data = vload2(0, (__global ushort *)((__global char *)src1 + src1_index)); - int2 src2_data = (int2)(src2.x, src2.x); - - ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index)); - int2 tmp = convert_int2_sat(src1_data) - src2_data; - tmp = isMatSubScalar ? tmp : -tmp; - ushort2 tmp_data = convert_ushort2_sat(tmp); - - data.x = (dst_index + 0 >= dst_start) ? tmp_data.x : data.x; - data.y = (dst_index + 2 < dst_end ) ? tmp_data.y : data.y; - - *((__global ushort2 *)((__global uchar *)dst + dst_index)) = data; - } -} -__kernel void arithm_s_sub_C1_D3 (__global short *src1, int src1_step, int src1_offset, - __global short *dst, int dst_step, int dst_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - - #define dst_align ((dst_offset >> 1) & 1) - int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc); - - short2 src1_data = vload2(0, (__global short *)((__global char *)src1 + src1_index)); - int2 src2_data = (int2)(src2.x, src2.x); - short2 data = *((__global short2 *)((__global uchar *)dst + dst_index)); - - int2 tmp = convert_int2_sat(src1_data) - src2_data; - tmp = isMatSubScalar ? tmp : -tmp; - short2 tmp_data = convert_short2_sat(tmp); - - data.x = (dst_index + 0 >= dst_start) ? tmp_data.x : data.x; - data.y = (dst_index + 2 < dst_end ) ? tmp_data.y : data.y; - - *((__global short2 *)((__global uchar *)dst + dst_index)) = data; - } -} -__kernel void arithm_s_sub_C1_D4 (__global int *src1, int src1_step, int src1_offset, - __global int *dst, int dst_step, int dst_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 2) + src1_offset); - int dst_index = mad24(y, dst_step, (x << 2) + dst_offset); - - int src_data1 = *((__global int *)((__global char *)src1 + src1_index)); - int src_data2 = src2.x; - - long tmp = (long)src_data1 - (long)src_data2; - tmp = isMatSubScalar ? tmp : -tmp; - int data = convert_int_sat(tmp); - - *((__global int *)((__global char *)dst + dst_index)) = data; - } -} -__kernel void arithm_s_sub_C1_D5 (__global float *src1, int src1_step, int src1_offset, - __global float *dst, int dst_step, int dst_offset, - float4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 2) + src1_offset); - int dst_index = mad24(y, dst_step, (x << 2) + dst_offset); - - float src_data1 = *((__global float *)((__global char *)src1 + src1_index)); - float src_data2 = src2.x; - - float tmp = src_data1 - src_data2; - tmp = isMatSubScalar ? tmp : -tmp; - - *((__global float *)((__global char *)dst + dst_index)) = tmp; - } -} - -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_s_sub_C1_D6 (__global double *src1, int src1_step, int src1_offset, - __global double *dst, int dst_step, int dst_offset, - double4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 3) + src1_offset); - int dst_index = mad24(y, dst_step, (x << 3) + dst_offset); - - double src_data1 = *((__global double *)((__global char *)src1 + src1_index)); - double src2_data = src2.x; - - double data = src_data1 - src2_data; - data = isMatSubScalar ? data : -data; - - *((__global double *)((__global char *)dst + dst_index)) = data; - } -} -#endif - -__kernel void arithm_s_sub_C2_D0 (__global uchar *src1, int src1_step, int src1_offset, - __global uchar *dst, int dst_step, int dst_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - - #define dst_align ((dst_offset >> 1) & 1) - int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc); - - uchar4 src1_data = vload4(0, src1 + src1_index); - int4 src2_data = (int4)(src2.x, src2.y, src2.x, src2.y); - - uchar4 data = *((__global uchar4 *)(dst + dst_index)); - int4 tmp = convert_int4_sat(src1_data) - src2_data; - tmp = isMatSubScalar ? tmp : -tmp; - uchar4 tmp_data = convert_uchar4_sat(tmp); - - data.xy = (dst_index + 0 >= dst_start) ? tmp_data.xy : data.xy; - data.zw = (dst_index + 2 < dst_end ) ? tmp_data.zw : data.zw; - - *((__global uchar4 *)(dst + dst_index)) = data; - } -} -__kernel void arithm_s_sub_C2_D2 (__global ushort *src1, int src1_step, int src1_offset, - __global ushort *dst, int dst_step, int dst_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 2) + src1_offset); - int dst_index = mad24(y, dst_step, (x << 2) + dst_offset); - - ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index)); - int2 src_data2 = (int2)(src2.x, src2.y); - ushort2 dst_data = *((__global ushort2 *)((__global char *)dst + dst_index)); - - int2 tmp = convert_int2_sat(src_data1) - src_data2; - tmp = isMatSubScalar ? tmp : -tmp; - ushort2 data = convert_ushort2_sat(tmp); - - *((__global ushort2 *)((__global char *)dst + dst_index)) = data; - } -} -__kernel void arithm_s_sub_C2_D3 (__global short *src1, int src1_step, int src1_offset, - __global short *dst, int dst_step, int dst_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 2) + src1_offset); - int dst_index = mad24(y, dst_step, (x << 2) + dst_offset); - - short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index)); - int2 src_data2 = (int2)(src2.x, src2.y); - short2 dst_data = *((__global short2 *)((__global char *)dst + dst_index)); - - int2 tmp = convert_int2_sat(src_data1) - src_data2; - tmp = isMatSubScalar ? tmp : -tmp; - short2 data = convert_short2_sat(tmp); - - *((__global short2 *)((__global char *)dst + dst_index)) = data; - } -} -__kernel void arithm_s_sub_C2_D4 (__global int *src1, int src1_step, int src1_offset, - __global int *dst, int dst_step, int dst_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 3) + src1_offset); - int dst_index = mad24(y, dst_step, (x << 3) + dst_offset); - - int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index)); - int2 src_data2 = (int2)(src2.x, src2.y); - int2 dst_data = *((__global int2 *)((__global char *)dst + dst_index)); - - long2 tmp = convert_long2_sat(src_data1) - convert_long2_sat(src_data2); - tmp = isMatSubScalar ? tmp : -tmp; - int2 data = convert_int2_sat(tmp); - - *((__global int2 *)((__global char *)dst + dst_index)) = data; - } -} -__kernel void arithm_s_sub_C2_D5 (__global float *src1, int src1_step, int src1_offset, - __global float *dst, int dst_step, int dst_offset, - float4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 3) + src1_offset); - int dst_index = mad24(y, dst_step, (x << 3) + dst_offset); - - float2 src_data1 = *((__global float2 *)((__global char *)src1 + src1_index)); - float2 src_data2 = (float2)(src2.x, src2.y); - float2 dst_data = *((__global float2 *)((__global char *)dst + dst_index)); - - float2 tmp = src_data1 - src_data2; - tmp = isMatSubScalar ? tmp : -tmp; - - *((__global float2 *)((__global char *)dst + dst_index)) = tmp; - } -} - -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_s_sub_C2_D6 (__global double *src1, int src1_step, int src1_offset, - __global double *dst, int dst_step, int dst_offset, - double4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 4) + src1_offset); - int dst_index = mad24(y, dst_step, (x << 4) + dst_offset); - - double2 src_data1 = *((__global double2 *)((__global char *)src1 + src1_index)); - double2 src_data2 = (double2)(src2.x, src2.y); - double2 dst_data = *((__global double2 *)((__global char *)dst + dst_index)); - - double2 data = src_data1 - src_data2; - data = isMatSubScalar ? data : -data; - - *((__global double2 *)((__global char *)dst + dst_index)) = data; - } -} -#endif - -__kernel void arithm_s_sub_C4_D0 (__global uchar *src1, int src1_step, int src1_offset, - __global uchar *dst, int dst_step, int dst_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 2) + src1_offset); - int dst_index = mad24(y, dst_step, (x << 2) + dst_offset); - - uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index)); - - int4 tmp = convert_int4_sat(src_data1) - src2; - tmp = isMatSubScalar ? tmp : -tmp; - uchar4 data = convert_uchar4_sat(tmp); - - *((__global uchar4 *)(dst + dst_index)) = data; - } -} -__kernel void arithm_s_sub_C4_D2 (__global ushort *src1, int src1_step, int src1_offset, - __global ushort *dst, int dst_step, int dst_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 3) + src1_offset); - int dst_index = mad24(y, dst_step, (x << 3) + dst_offset); - - ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index)); - - int4 tmp = convert_int4_sat(src_data1) - src2; - tmp = isMatSubScalar ? tmp : -tmp; - ushort4 data = convert_ushort4_sat(tmp); - - *((__global ushort4 *)((__global char *)dst + dst_index)) = data; - } -} -__kernel void arithm_s_sub_C4_D3 (__global short *src1, int src1_step, int src1_offset, - __global short *dst, int dst_step, int dst_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 3) + src1_offset); - int dst_index = mad24(y, dst_step, (x << 3) + dst_offset); - - short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index)); - - int4 tmp = convert_int4_sat(src_data1) - src2; - tmp = isMatSubScalar ? tmp : -tmp; - short4 data = convert_short4_sat(tmp); - - *((__global short4 *)((__global char *)dst + dst_index)) = data; - } -} -__kernel void arithm_s_sub_C4_D4 (__global int *src1, int src1_step, int src1_offset, - __global int *dst, int dst_step, int dst_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 4) + src1_offset); - int dst_index = mad24(y, dst_step, (x << 4) + dst_offset); - - int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index)); - - long4 tmp = convert_long4_sat(src_data1) - convert_long4_sat(src2); - tmp = isMatSubScalar ? tmp : -tmp; - int4 data = convert_int4_sat(tmp); - - *((__global int4 *)((__global char *)dst + dst_index)) = data; - } -} -__kernel void arithm_s_sub_C4_D5 (__global float *src1, int src1_step, int src1_offset, - __global float *dst, int dst_step, int dst_offset, - float4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 4) + src1_offset); - int dst_index = mad24(y, dst_step, (x << 4) + dst_offset); - - float4 src_data1 = *((__global float4 *)((__global char *)src1 + src1_index)); - - float4 tmp = src_data1 - src2; - tmp = isMatSubScalar ? tmp : -tmp; - - *((__global float4 *)((__global char *)dst + dst_index)) = tmp; - } -} - -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_s_sub_C4_D6 (__global double *src1, int src1_step, int src1_offset, - __global double *dst, int dst_step, int dst_offset, - double4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 5) + src1_offset); - int dst_index = mad24(y, dst_step, (x << 5) + dst_offset); - - double4 src_data1 = *((__global double4 *)((__global char *)src1 + src1_index)); - - double4 data = src_data1 - src2; - data = isMatSubScalar ? data : -data; - - *((__global double4 *)((__global char *)dst + dst_index)) = data; - } -} -#endif diff --git a/modules/ocl/src/opencl/arithm_sub_scalar_mask.cl b/modules/ocl/src/opencl/arithm_sub_scalar_mask.cl deleted file mode 100644 index 9b758cf4c9..0000000000 --- a/modules/ocl/src/opencl/arithm_sub_scalar_mask.cl +++ /dev/null @@ -1,611 +0,0 @@ -/*M/////////////////////////////////////////////////////////////////////////////////////// -// -// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING. -// -// By downloading, copying, installing or using the software you agree to this license. -// If you do not agree to this license, do not download, install, -// copy or use the software. -// -// -// License Agreement -// For Open Source Computer Vision Library -// -// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved. -// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved. -// Third party copyrights are property of their respective owners. -// -// @Authors -// Jia Haipeng, jiahaipeng95@gmail.com -// -// Redistribution and use in source and binary forms, with or without modification, -// are permitted provided that the following conditions are met: -// -// * Redistribution's of source code must retain the above copyright notice, -// this list of conditions and the following disclaimer. -// -// * Redistribution's in binary form must reproduce the above copyright notice, -// this list of conditions and the following disclaimer in the documentation -// and/or other GpuMaterials provided with the distribution. -// -// * The name of the copyright holders may not be used to endorse or promote products -// derived from this software without specific prior written permission. -// -// This software is provided by the copyright holders and contributors as is and -// any express or implied warranties, including, but not limited to, the implied -// warranties of merchantability and fitness for a particular purpose are disclaimed. -// In no event shall the Intel Corporation or contributors be liable for any direct, -// indirect, incidental, special, exemplary, or consequential damages -// (including, but not limited to, procurement of substitute goods or services; -// loss of use, data, or profits; or business interruption) however caused -// and on any theory of liability, whether in contract, strict liability, -// or tort (including negligence or otherwise) arising in any way out of -// the use of this software, even if advised of the possibility of such damage. -// -//M*/ - -#if defined (DOUBLE_SUPPORT) -#ifdef cl_khr_fp64 -#pragma OPENCL EXTENSION cl_khr_fp64:enable -#elif defined (cl_amd_fp64) -#pragma OPENCL EXTENSION cl_amd_fp64:enable -#endif -#endif - -/**************************************sub with scalar with mask**************************************/ -__kernel void arithm_s_sub_with_mask_C1_D0 (__global uchar *src1, int src1_step, int src1_offset, - __global uchar *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 2; - - #define dst_align (dst_offset & 3) - int src1_index = mad24(y, src1_step, x + src1_offset - dst_align); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc); - - uchar4 src1_data = vload4(0, src1 + src1_index); - int4 src2_data = (int4)(src2.x, src2.x, src2.x, src2.x); - uchar4 mask_data = vload4(0, mask + mask_index); - - uchar4 data = *((__global uchar4 *)(dst + dst_index)); - int4 tmp = convert_int4_sat(src1_data) - src2_data; - tmp = isMatSubScalar ? tmp : -tmp; - uchar4 tmp_data = convert_uchar4_sat(tmp); - - data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x; - data.y = ((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y; - data.z = ((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z; - data.w = ((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w; - - *((__global uchar4 *)(dst + dst_index)) = data; - } -} -__kernel void arithm_s_sub_with_mask_C1_D2 (__global ushort *src1, int src1_step, int src1_offset, - __global ushort *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - - #define dst_align ((dst_offset >> 1) & 1) - int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc); - - ushort2 src1_data = vload2(0, (__global ushort *)((__global char *)src1 + src1_index)); - int2 src2_data = (int2)(src2.x, src2.x); - uchar2 mask_data = vload2(0, mask + mask_index); - - ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index)); - int2 tmp = convert_int2_sat(src1_data) - src2_data; - tmp = isMatSubScalar ? tmp : -tmp; - ushort2 tmp_data = convert_ushort2_sat(tmp); - - data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x; - data.y = ((mask_data.y) && (dst_index + 2 < dst_end )) ? tmp_data.y : data.y; - - *((__global ushort2 *)((__global uchar *)dst + dst_index)) = data; - } -} -__kernel void arithm_s_sub_with_mask_C1_D3 (__global short *src1, int src1_step, int src1_offset, - __global short *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - - #define dst_align ((dst_offset >> 1) & 1) - int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc); - - short2 src1_data = vload2(0, (__global short *)((__global char *)src1 + src1_index)); - int2 src2_data = (int2)(src2.x, src2.x); - uchar2 mask_data = vload2(0, mask + mask_index); - - short2 data = *((__global short2 *)((__global uchar *)dst + dst_index)); - int2 tmp = convert_int2_sat(src1_data) - src2_data; - tmp = isMatSubScalar ? tmp : -tmp; - short2 tmp_data = convert_short2_sat(tmp); - - data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x; - data.y = ((mask_data.y) && (dst_index + 2 < dst_end )) ? tmp_data.y : data.y; - - *((__global short2 *)((__global uchar *)dst + dst_index)) = data; - } -} -__kernel void arithm_s_sub_with_mask_C1_D4 (__global int *src1, int src1_step, int src1_offset, - __global int *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 2) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 2) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - int src_data1 = *((__global int *)((__global char *)src1 + src1_index)); - int src_data2 = src2.x; - int dst_data = *((__global int *)((__global char *)dst + dst_index)); - - long tmp = (long)src_data1 - (long)src_data2; - tmp = isMatSubScalar ? tmp : - tmp; - int data = convert_int_sat(tmp); - data = mask_data ? data : dst_data; - - *((__global int *)((__global char *)dst + dst_index)) = data; - } -} - -__kernel void arithm_s_sub_with_mask_C1_D5 (__global float *src1, int src1_step, int src1_offset, - __global float *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - float4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 2) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 2) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - float src_data1 = *((__global float *)((__global char *)src1 + src1_index)); - float src_data2 = src2.x; - float dst_data = *((__global float *)((__global char *)dst + dst_index)); - - float data = src_data1 - src_data2; - data = isMatSubScalar ? data : -data; - data = mask_data ? data : dst_data; - - *((__global float *)((__global char *)dst + dst_index)) = data; - } -} - - -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_s_sub_with_mask_C1_D6 (__global double *src1, int src1_step, int src1_offset, - __global double *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - double4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 3) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 3) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - double src_data1 = *((__global double *)((__global char *)src1 + src1_index)); - double src_data2 = src2.x; - double dst_data = *((__global double *)((__global char *)dst + dst_index)); - - double data = src_data1 - src_data2; - data = isMatSubScalar ? data : -data; - data = mask_data ? data : dst_data; - - *((__global double *)((__global char *)dst + dst_index)) = data; - } -} -#endif -__kernel void arithm_s_sub_with_mask_C2_D0 (__global uchar *src1, int src1_step, int src1_offset, - __global uchar *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - x = x << 1; - - #define dst_align ((dst_offset >> 1) & 1) - int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1)); - int mask_index = mad24(y, mask_step, x + mask_offset - dst_align); - - int dst_start = mad24(y, dst_step, dst_offset); - int dst_end = mad24(y, dst_step, dst_offset + dst_step1); - int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc); - - uchar4 src1_data = vload4(0, src1 + src1_index); - int4 src2_data = (int4)(src2.x, src2.y, src2.x, src2.y); - uchar2 mask_data = vload2(0, mask + mask_index); - - uchar4 data = *((__global uchar4 *)(dst + dst_index)); - int4 tmp = convert_int4_sat(src1_data) - src2_data; - tmp = isMatSubScalar ? tmp : -tmp; - uchar4 tmp_data = convert_uchar4_sat(tmp); - - data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy; - data.zw = ((mask_data.y) && (dst_index + 2 < dst_end )) ? tmp_data.zw : data.zw; - - *((__global uchar4 *)(dst + dst_index)) = data; - } -} -__kernel void arithm_s_sub_with_mask_C2_D2 (__global ushort *src1, int src1_step, int src1_offset, - __global ushort *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 2) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 2) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index)); - int2 src_data2 = (int2)(src2.x, src2.y); - ushort2 dst_data = *((__global ushort2 *)((__global char *)dst + dst_index)); - - int2 tmp = convert_int2_sat(src_data1) - src_data2; - tmp = isMatSubScalar ? tmp : -tmp; - ushort2 data = convert_ushort2_sat(tmp); - data = mask_data ? data : dst_data; - - *((__global ushort2 *)((__global char *)dst + dst_index)) = data; - } -} -__kernel void arithm_s_sub_with_mask_C2_D3 (__global short *src1, int src1_step, int src1_offset, - __global short *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 2) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 2) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index)); - int2 src_data2 = (int2)(src2.x, src2.y); - short2 dst_data = *((__global short2 *)((__global char *)dst + dst_index)); - - int2 tmp = convert_int2_sat(src_data1) - src_data2; - tmp = isMatSubScalar ? tmp : -tmp; - short2 data = convert_short2_sat(tmp); - data = mask_data ? data : dst_data; - - *((__global short2 *)((__global char *)dst + dst_index)) = data; - } -} -__kernel void arithm_s_sub_with_mask_C2_D4 (__global int *src1, int src1_step, int src1_offset, - __global int *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 3) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 3) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index)); - int2 src_data2 = (int2)(src2.x, src2.y); - int2 dst_data = *((__global int2 *)((__global char *)dst + dst_index)); - - long2 tmp = convert_long2_sat(src_data1) - convert_long2_sat(src_data2); - tmp = isMatSubScalar ? tmp : -tmp; - int2 data = convert_int2_sat(tmp); - data = mask_data ? data : dst_data; - - *((__global int2 *)((__global char *)dst + dst_index)) = data; - } -} -__kernel void arithm_s_sub_with_mask_C2_D5 (__global float *src1, int src1_step, int src1_offset, - __global float *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - float4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 3) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 3) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - float2 src_data1 = *((__global float2 *)((__global char *)src1 + src1_index)); - float2 src_data2 = (float2)(src2.x, src2.y); - float2 dst_data = *((__global float2 *)((__global char *)dst + dst_index)); - - float2 data = src_data1 - src_data2; - data = isMatSubScalar ? data : -data; - data = mask_data ? data : dst_data; - - *((__global float2 *)((__global char *)dst + dst_index)) = data; - } -} - -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_s_sub_with_mask_C2_D6 (__global double *src1, int src1_step, int src1_offset, - __global double *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - double4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 4) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 4) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - double2 src_data1 = *((__global double2 *)((__global char *)src1 + src1_index)); - double2 src_data2 = (double2)(src2.x, src2.y); - double2 dst_data = *((__global double2 *)((__global char *)dst + dst_index)); - - double2 data = src_data1 - src_data2; - data = isMatSubScalar ? data : -data; - data = mask_data ? data : dst_data; - - *((__global double2 *)((__global char *)dst + dst_index)) = data; - } -} -#endif - -__kernel void arithm_s_sub_with_mask_C4_D0 (__global uchar *src1, int src1_step, int src1_offset, - __global uchar *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 2) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 2) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index)); - uchar4 dst_data = *((__global uchar4 *)(dst + dst_index)); - - int4 tmp = convert_int4_sat(src_data1) - src2; - tmp = isMatSubScalar ? tmp : -tmp; - uchar4 data = convert_uchar4_sat(tmp); - - data = mask_data ? data : dst_data; - - *((__global uchar4 *)(dst + dst_index)) = data; - } -} -__kernel void arithm_s_sub_with_mask_C4_D2 (__global ushort *src1, int src1_step, int src1_offset, - __global ushort *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 3) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 3) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index)); - ushort4 dst_data = *((__global ushort4 *)((__global char *)dst + dst_index)); - - int4 tmp = convert_int4_sat(src_data1) - src2; - tmp = isMatSubScalar ? tmp : -tmp; - ushort4 data = convert_ushort4_sat(tmp); - - data = mask_data ? data : dst_data; - - *((__global ushort4 *)((__global char *)dst + dst_index)) = data; - } -} -__kernel void arithm_s_sub_with_mask_C4_D3 (__global short *src1, int src1_step, int src1_offset, - __global short *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 3) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 3) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index)); - short4 dst_data = *((__global short4 *)((__global char *)dst + dst_index)); - - int4 tmp = convert_int4_sat(src_data1) - src2; - tmp = isMatSubScalar ? tmp : -tmp; - short4 data = convert_short4_sat(tmp); - - data = mask_data ? data : dst_data; - - *((__global short4 *)((__global char *)dst + dst_index)) = data; - } -} -__kernel void arithm_s_sub_with_mask_C4_D4 (__global int *src1, int src1_step, int src1_offset, - __global int *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - int4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 4) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 4) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index)); - int4 dst_data = *((__global int4 *)((__global char *)dst + dst_index)); - - long4 tmp = convert_long4_sat(src_data1) - convert_long4_sat(src2); - tmp = isMatSubScalar ? tmp : -tmp; - int4 data = convert_int4_sat(tmp); - - data = mask_data ? data : dst_data; - - *((__global int4 *)((__global char *)dst + dst_index)) = data; - } -} -__kernel void arithm_s_sub_with_mask_C4_D5 (__global float *src1, int src1_step, int src1_offset, - __global float *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - float4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 4) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 4) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - float4 src_data1 = *((__global float4 *)((__global char *)src1 + src1_index)); - float4 dst_data = *((__global float4 *)((__global char *)dst + dst_index)); - - float4 data = src_data1 - src2; - data = isMatSubScalar ? data : -data; - - data = mask_data ? data : dst_data; - - *((__global float4 *)((__global char *)dst + dst_index)) = data; - } -} - -#if defined (DOUBLE_SUPPORT) -__kernel void arithm_s_sub_with_mask_C4_D6 (__global double *src1, int src1_step, int src1_offset, - __global double *dst, int dst_step, int dst_offset, - __global uchar *mask, int mask_step, int mask_offset, - double4 src2, int rows, int cols, int dst_step1, int isMatSubScalar) -{ - - int x = get_global_id(0); - int y = get_global_id(1); - - if (x < cols && y < rows) - { - int src1_index = mad24(y, src1_step, (x << 5) + src1_offset); - int mask_index = mad24(y, mask_step, x + mask_offset); - int dst_index = mad24(y, dst_step, (x << 5) + dst_offset); - - uchar mask_data = *(mask + mask_index); - - double4 src_data1 = *((__global double4 *)((__global char *)src1 + src1_index)); - double4 dst_data = *((__global double4 *)((__global char *)dst + dst_index)); - - double4 data = src_data1 - src2; - data = isMatSubScalar ? data : -data; - data = mask_data ? data : dst_data; - - *((__global double4 *)((__global char *)dst + dst_index)) = data; - } -} -#endif