kanshan/nccl-tests
1
0
Fork 0
mirror of https://github.com/NVIDIA/nccl-tests.git synced 2026-04-23 16:08:20 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Add extra reserved space during maxBytes calculation

Browse Source 
Also, don't allow minBytes > maxBytes
This commit is contained in:
David Addison 2025-05-21 09:40:26 -07:00
parent e041d901e6
commit 598fc0060e
1 changed files with 8 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
src/common.cu
View File

@ -16,6 +16,9 @@
#include "../verifiable/verifiable.h"
#define DIVUP(x, y) \
(((x)+(y)-1)/(y))
int test_ncclVersion = 0; // init'd with ncclGetVersion()
#if NCCL_MAJOR >= 2
@ -1025,10 +1028,14 @@ testResult_t run() {
PRINT("%s", line);
#endif
// Reserve 1GiB of memory for each 16GiB installed, but limit to a max of 4GiB
const size_t GB = (1ULL << 30);
size_t reserveMem = std::min(DIVUP(maxMem, 16*GB) * 1*GB, 4*GB);
// We need sendbuff, recvbuff, expected (when datacheck enabled), plus 1G for the rest.
size_t memMaxBytes = (maxMem - (1<<30)) / (datacheck ? 3 : 2);
size_t memMaxBytes = (maxMem - reserveMem - 1*GB) / (datacheck ? 3 : 2);
if (maxBytes > memMaxBytes) {
maxBytes = memMaxBytes;
if (minBytes > maxBytes) minBytes = maxBytes;
if (proc == 0) printf("#\n# Reducing maxBytes to %ld due to memory limitation\n", maxBytes);
}