AI/rocm-systems
2
0
Garpu 0
Kode Masalah Tarik Permintaan Actions Packages Projects Rilis Wiki Kegiatan

extending the unit-tests for multi-rank support

Telusuri Sumber 
[ROCm/rccl commit: a87d61db2b]
This commit is contained in:
Edgar
2022-03-15 18:53:52 -04:00
committed by Edgar Gabriel
orang tua 8953f5b5ca
melakukan f7ef619ba7
6 mengubah file dengan 90 tambahan dan 36 penghapusan
Download Patch File Download Diff File Expand all files Collapse all files
projects/rccl/test/SendRecv_SinglePairs.cpp
+6 -4
Melihat File
@@ -19,13 +19,15 @@ namespace RcclUnitTesting
OptionalColArgs options;
bool isCorrect = true;
int totalRanks = testBed.ev.maxGpus;
int numGpus = testBed.ev.maxGpus;
for (int rpg=0; rpg < 2 && isCorrect; ++rpg)
for (int isMultiProcess = 0; isMultiProcess <= 1 && isCorrect; ++isMultiProcess)
{
if (!(testBed.ev.processMask & (1 << isMultiProcess))) continue;
int const numProcesses = isMultiProcess ? totalRanks : 1;
testBed.InitComms(TestBed::GetDeviceIdsList(numProcesses, totalRanks), 1);
int ranksPerGpu = rpg == 0 ? 1 : testBed.ev.maxRanksPerGpu;
int totalRanks = numGpus * ranksPerGpu;
int const numProcesses = isMultiProcess ? numGpus : 1;
testBed.InitComms(TestBed::GetDeviceIdsList(numProcesses, numGpus, ranksPerGpu), 1);
for (int dataIdx = 0; dataIdx < dataTypes.size() && isCorrect; ++dataIdx)
for (int numIdx = 0; numIdx < numElements.size() && isCorrect; ++numIdx)