portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2024-11-10 07:55:35 +00:00
Code Releases Activity

The GPU reduction was not working for me and causing errors. Need to revisit.

Browse Source 
Gianluca is working on deterministic reduction/
This commit is contained in:
Peter Boyle 2019-06-08 13:39:11 +01:00
parent 6d77941990
commit dc5024e88c
1 changed files with 1 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
Grid/lattice/Lattice_reduction.h
View File

@ -59,8 +59,7 @@ inline ComplexD innerProduct(const Lattice<vobj> &left,const Lattice<vobj> &righ
auto left_v = left.View(); auto left_v = left.View();
auto right_v=right.View(); auto right_v=right.View();
#ifdef GRID_NVCC #if 0
typedef decltype(innerProduct(left_v[0],right_v[0])) inner_t; typedef decltype(innerProduct(left_v[0],right_v[0])) inner_t;
thrust::plus<inner_t> binary_sum; thrust::plus<inner_t> binary_sum;
innerProductFunctor<vobj,inner_t> binary_inner_p; innerProductFunctor<vobj,inner_t> binary_inner_p;
@ -69,7 +68,6 @@ inline ComplexD innerProduct(const Lattice<vobj> &left,const Lattice<vobj> &righ
// is there a way of using the efficient thrust reduction while maintaining memory coalescing? // is there a way of using the efficient thrust reduction while maintaining memory coalescing?
inner_t vnrm = thrust::inner_product(thrust::device, &left_v[0], &left_v[sN], &right_v[0], zero, binary_sum, binary_inner_p); inner_t vnrm = thrust::inner_product(thrust::device, &left_v[0], &left_v[sN], &right_v[0], zero, binary_sum, binary_inner_p);
nrm = Reduce(TensorRemove(vnrm));// sum across simd nrm = Reduce(TensorRemove(vnrm));// sum across simd
#else #else
thread_loop( (int thr=0;thr<grid->SumArraySize();thr++),{ thread_loop( (int thr=0;thr<grid->SumArraySize();thr++),{
int mywork, myoff; int mywork, myoff;