AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

removed hcc mentions (#1291)

Browse Source 
[ROCm/rccl commit: d6c317d6ae]
This commit is contained in:
akolliasAMD
2024-08-14 15:04:13 -06:00
committed by GitHub
parent 09fe7f9b97
commit 38e189bb1e
19 changed files with 44 additions and 45 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/rccl/src/enqueue.cc
+2 -2
View File
@@ -1645,7 +1645,7 @@ static ncclResult_t getChannnelThreadInfo(struct ncclInfo* collInfo) {
// Ring/Tree channel tuning
while (collInfo->nBytes < nc*nt*threadThreshold) {
if (nc >= 2) nc--;
#if defined(__HIP_PLATFORM_AMD__) || defined(__HCC__) || defined(__HIPCC__)
#if defined(__HIP_PLATFORM_AMD__) || defined(__HIPCC__)
// do not reduce threads count on VEGA
#else
else if ((nt % 128) == 0) nt/=2;
@@ -1653,7 +1653,7 @@ static ncclResult_t getChannnelThreadInfo(struct ncclInfo* collInfo) {
else break;
}
}
#if defined(__HIP_PLATFORM_AMD__) || defined(__HCC__) || defined(__HIPCC__)
#if defined(__HIP_PLATFORM_AMD__) || defined(__HIPCC__)
#else
if (collInfo->protocol == NCCL_PROTO_SIMPLE) {
if (collInfo->algorithm == NCCL_ALGO_RING) nt += WARP_SIZE; // Extra warp for sync