kanshan/nccl-tests
1
0
Fork 0
mirror of https://github.com/NVIDIA/nccl-tests.git synced 2026-02-22 02:34:42 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Change all_gather/reduce_scatter algbw to match the documentation.

Browse Source 
Fix #45 : All_gather and reduce_scatter algorithm bandwidth was
computed as time/count*(nranks-1) which is not consistent with the
way we compute it for other collectives.

This change makes algbw higher; busbw is unchanged.
This commit is contained in:
Sylvain Jeaugey 2020-06-19 10:40:33 -07:00
parent 07ac716c1a
commit ec1b5e22e6
2 changed files with 4 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
src/all_gather.cu
View File

@ -48,10 +48,10 @@ testResult_t AllGatherInitData(struct threadArgs* args, ncclDataType_t type, ncc
}
void AllGatherGetBw(size_t count, int typesize, double sec, double* algBw, double* busBw, int nranks) {
double baseBw = (double)(count * typesize * (nranks - 1)) / 1.0E9 / sec;
double baseBw = (double)(count * typesize * nranks) / 1.0E9 / sec;
*algBw = baseBw;
double factor = 1;
double factor = ((double)(nranks - 1))/((double)nranks);
*busBw = baseBw * factor;
}

4
src/reduce_scatter.cu
View File

@ -47,10 +47,10 @@ testResult_t ReduceScatterInitData(struct threadArgs* args, ncclDataType_t type,
}
void ReduceScatterGetBw(size_t count, int typesize, double sec, double* algBw, double* busBw, int nranks) {
double baseBw = (double)(count * typesize * (nranks - 1)) / 1.0E9 / sec;
double baseBw = (double)(count * typesize * nranks) / 1.0E9 / sec;
*algBw = baseBw;
double factor = 1;
double factor = ((double)(nranks - 1))/((double)nranks);
*busBw = baseBw * factor;
}