OilfieldToolsNet/LBPM
4
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

cooperative groups builds on summit

Browse Source
This commit is contained in:
James E McClure
2018-02-20 21:20:14 -05:00
parent 5a81875503
commit 030f997516
1 changed files with 2 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
gpu/D3Q19.cu
View File

@@ -1,4 +1,5 @@
#include <stdio.h>
#include <cooperative_groups.h>
#define NBLOCKS 1024
#define NTHREADS 256
@@ -65,7 +66,7 @@ __global__ void sum_kernel_block(double *sum, double *input, int n)
double my_sum = thread_sum(input, n);
extern __shared__ double temp[];
auto g = this_thread_block();
thread_group g = this_thread_block();
double block_sum = reduce_sum(g, temp, my_sum);
if (g.thread_rank() == 0) atomicAdd(sum, block_sum);