generic-library/opencv
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

fixed gpu::sum* on CC1.0, updated some tests

Browse Source
This commit is contained in:
Alexey Spizhevoy
2011-02-02 07:23:55 +00:00
parent f7e62d89f8
commit 557dd39f03
6 changed files with 27 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
modules/gpu/src/cuda/matrix_reductions.cu
View File

@@ -1394,7 +1394,7 @@ namespace cv { namespace gpu { namespace mathfunc
smem[tid] = res.x;
smem[tid + nthreads] = res.y;
smem[tid + 2 * nthreads] = res.z;
smem[tid + 3 * nthreads] = res.z;
smem[tid + 3 * nthreads] = res.w;
__syncthreads();
sumInSmem<nthreads, R>(smem, tid);
@@ -1432,21 +1432,25 @@ namespace cv { namespace gpu { namespace mathfunc
src, (typename TypeVec<R, 1>::vec_t*)buf.ptr(0));
sumPass2Kernel<T, R, threads_x * threads_y><<<1, threads_x * threads_y>>>(
(typename TypeVec<R, 1>::vec_t*)buf.ptr(0), grid.x * grid.y);
break;
case 2:
sumKernel_C2<T, R, IdentityOp<R>, threads_x * threads_y><<<grid, threads>>>(
src, (typename TypeVec<R, 2>::vec_t*)buf.ptr(0));
sumPass2Kernel_C2<T, R, threads_x * threads_y><<<1, threads_x * threads_y>>>(
(typename TypeVec<R, 2>::vec_t*)buf.ptr(0), grid.x * grid.y);
break;
case 3:
sumKernel_C3<T, R, IdentityOp<R>, threads_x * threads_y><<<grid, threads>>>(
src, (typename TypeVec<R, 3>::vec_t*)buf.ptr(0));
sumPass2Kernel_C3<T, R, threads_x * threads_y><<<1, threads_x * threads_y>>>(
(typename TypeVec<R, 3>::vec_t*)buf.ptr(0), grid.x * grid.y);
break;
case 4:
sumKernel_C4<T, R, IdentityOp<R>, threads_x * threads_y><<<grid, threads>>>(
src, (typename TypeVec<R, 4>::vec_t*)buf.ptr(0));
sumPass2Kernel_C4<T, R, threads_x * threads_y><<<1, threads_x * threads_y>>>(
(typename TypeVec<R, 4>::vec_t*)buf.ptr(0), grid.x * grid.y);
break;
}
cudaSafeCall(cudaThreadSynchronize());