//M*/
#include <stddef.h>
+#include <iostream>
#include "cuda_shared.hpp"
#include "cuda_runtime.h"
}
-extern "C" void cv::gpu::impl::set_to_with_mask(const DevMem2D& mat, const double * scalar, const DevMem2D& mask, int depth, int channels)
+extern "C" void cv::gpu::impl::set_to_with_mask(const DevMem2D& mat, const double * scalar, const DevMem2D& mask, int elemSize1, int channels)
{
- scalar_d[0] = scalar[0];
- scalar_d[1] = scalar[1];
- scalar_d[2] = scalar[2];
- scalar_d[3] = scalar[3];
+ // download scalar to constant memory
+ float data[4];
+ data[0] = scalar[0];
+ data[1] = scalar[1];
+ data[2] = scalar[2];
+ data[3] = scalar[3];
+ cudaMemcpyToSymbol(scalar_d, data, sizeof(data));
dim3 numBlocks(mat.rows * mat.step / 256, 1, 1);
dim3 threadsPerBlock(256);
if (channels == 1)
{
- if (depth == 1) ::mat_operators::kernel_set_to_with_mask<unsigned char, 1><<<numBlocks,threadsPerBlock>>>(mat.ptr, (float *)mask.ptr);
- if (depth == 2) ::mat_operators::kernel_set_to_with_mask<unsigned short, 1><<<numBlocks,threadsPerBlock>>>((unsigned short *)mat.ptr, (float *)mask.ptr);
- if (depth == 4) ::mat_operators::kernel_set_to_with_mask<unsigned int, 1><<<numBlocks,threadsPerBlock>>>((unsigned int *)mat.ptr, (float *)mask.ptr);
+ if (elemSize1 == 1) ::mat_operators::kernel_set_to_with_mask<unsigned char, 1><<<numBlocks,threadsPerBlock>>>(mat.ptr, (float *)mask.ptr);
+ if (elemSize1 == 2) ::mat_operators::kernel_set_to_with_mask<unsigned short, 1><<<numBlocks,threadsPerBlock>>>((unsigned short *)mat.ptr, (float *)mask.ptr);
+ if (elemSize1 == 4) ::mat_operators::kernel_set_to_with_mask<unsigned int, 1><<<numBlocks,threadsPerBlock>>>((unsigned int *)mat.ptr, (float *)mask.ptr);
}
if (channels == 2)
{
- if (depth == 1) ::mat_operators::kernel_set_to_with_mask<unsigned char, 2><<<numBlocks,threadsPerBlock>>>(mat.ptr, (float *)mask.ptr);
- if (depth == 2) ::mat_operators::kernel_set_to_with_mask<unsigned short, 2><<<numBlocks,threadsPerBlock>>>((unsigned short *)mat.ptr, (float *)mask.ptr);
- if (depth == 4) ::mat_operators::kernel_set_to_with_mask<unsigned int, 2><<<numBlocks,threadsPerBlock>>>((unsigned int *)mat.ptr, (float *)mask.ptr);
+ if (elemSize1 == 1) ::mat_operators::kernel_set_to_with_mask<unsigned char, 2><<<numBlocks,threadsPerBlock>>>(mat.ptr, (float *)mask.ptr);
+ if (elemSize1 == 2) ::mat_operators::kernel_set_to_with_mask<unsigned short, 2><<<numBlocks,threadsPerBlock>>>((unsigned short *)mat.ptr, (float *)mask.ptr);
+ if (elemSize1 == 4) ::mat_operators::kernel_set_to_with_mask<unsigned int, 2><<<numBlocks,threadsPerBlock>>>((unsigned int *)mat.ptr, (float *)mask.ptr);
}
if (channels == 3)
{
- if (depth == 1) ::mat_operators::kernel_set_to_with_mask<unsigned char, 3><<<numBlocks,threadsPerBlock>>>(mat.ptr, (float *)mask.ptr);
- if (depth == 2) ::mat_operators::kernel_set_to_with_mask<unsigned short, 3><<<numBlocks,threadsPerBlock>>>((unsigned short *)mat.ptr, (float *)mask.ptr);
- if (depth == 4) ::mat_operators::kernel_set_to_with_mask<unsigned int, 3><<<numBlocks,threadsPerBlock>>>((unsigned int *)mat.ptr, (float *)mask.ptr);
+ if (elemSize1 == 1) ::mat_operators::kernel_set_to_with_mask<unsigned char, 3><<<numBlocks,threadsPerBlock>>>(mat.ptr, (float *)mask.ptr);
+ if (elemSize1 == 2) ::mat_operators::kernel_set_to_with_mask<unsigned short, 3><<<numBlocks,threadsPerBlock>>>((unsigned short *)mat.ptr, (float *)mask.ptr);
+ if (elemSize1 == 4) ::mat_operators::kernel_set_to_with_mask<unsigned int, 3><<<numBlocks,threadsPerBlock>>>((unsigned int *)mat.ptr, (float *)mask.ptr);
}
}
-extern "C" void cv::gpu::impl::set_to_without_mask(const DevMem2D& mat, const double * scalar, int depth, int channels)
+extern "C" void cv::gpu::impl::set_to_without_mask(const DevMem2D& mat, const double * scalar, int elemSize1, int channels)
{
- scalar_d[0] = scalar[0];
- scalar_d[1] = scalar[1];
- scalar_d[2] = scalar[2];
- scalar_d[3] = scalar[3];
+ float data[4];
+ data[0] = scalar[0];
+ data[1] = scalar[1];
+ data[2] = scalar[2];
+ data[3] = scalar[3];
+ cudaMemcpyToSymbol(scalar_d, data, sizeof(data));
int numBlocks = mat.rows * mat.step / 256;
if (channels == 1)
{
- if (depth == 1) ::mat_operators::kernel_set_to_without_mask<unsigned char, 1><<<numBlocks,threadsPerBlock>>>(mat.ptr);
- if (depth == 2) ::mat_operators::kernel_set_to_without_mask<unsigned short, 1><<<numBlocks,threadsPerBlock>>>((unsigned short *)mat.ptr);
- if (depth == 4) ::mat_operators::kernel_set_to_without_mask<unsigned int, 1><<<numBlocks,threadsPerBlock>>>((unsigned int *)mat.ptr);
+ if (elemSize1 == 1) ::mat_operators::kernel_set_to_without_mask<unsigned char, 1><<<numBlocks,threadsPerBlock>>>(mat.ptr);
+ if (elemSize1 == 2) ::mat_operators::kernel_set_to_without_mask<unsigned short, 1><<<numBlocks,threadsPerBlock>>>((unsigned short *)mat.ptr);
+ if (elemSize1 == 4) ::mat_operators::kernel_set_to_without_mask<unsigned int, 1><<<numBlocks,threadsPerBlock>>>((unsigned int *)mat.ptr);
}
if (channels == 2)
{
- if (depth == 1) ::mat_operators::kernel_set_to_without_mask<unsigned char, 2><<<numBlocks,threadsPerBlock>>>(mat.ptr);
- if (depth == 2) ::mat_operators::kernel_set_to_without_mask<unsigned short, 2><<<numBlocks,threadsPerBlock>>>((unsigned short *)mat.ptr);
- if (depth == 4) ::mat_operators::kernel_set_to_without_mask<unsigned int, 2><<<numBlocks,threadsPerBlock>>>((unsigned int *)mat.ptr);
+ if (elemSize1 == 1) ::mat_operators::kernel_set_to_without_mask<unsigned char, 2><<<numBlocks,threadsPerBlock>>>(mat.ptr);
+ if (elemSize1 == 2) ::mat_operators::kernel_set_to_without_mask<unsigned short, 2><<<numBlocks,threadsPerBlock>>>((unsigned short *)mat.ptr);
+ if (elemSize1 == 4) ::mat_operators::kernel_set_to_without_mask<unsigned int, 2><<<numBlocks,threadsPerBlock>>>((unsigned int *)mat.ptr);
}
if (channels == 3)
{
- if (depth == 1) ::mat_operators::kernel_set_to_without_mask<unsigned char, 3><<<numBlocks,threadsPerBlock>>>(mat.ptr);
- if (depth == 2) ::mat_operators::kernel_set_to_without_mask<unsigned short, 3><<<numBlocks,threadsPerBlock>>>((unsigned short *)mat.ptr);
- if (depth == 4) ::mat_operators::kernel_set_to_without_mask<unsigned int, 3><<<numBlocks,threadsPerBlock>>>((unsigned int *)mat.ptr);
+ if (elemSize1 == 1) ::mat_operators::kernel_set_to_without_mask<unsigned char, 3><<<numBlocks,threadsPerBlock>>>(mat.ptr);
+ if (elemSize1 == 2) ::mat_operators::kernel_set_to_without_mask<unsigned short, 3><<<numBlocks,threadsPerBlock>>>((unsigned short *)mat.ptr);
+ if (elemSize1 == 4) ::mat_operators::kernel_set_to_without_mask<unsigned int, 3><<<numBlocks,threadsPerBlock>>>((unsigned int *)mat.ptr);
}
}