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

GPU allocation for CPX Unit Tests using PCI bus id (#1403)

Browse Source 
* mapping devices wrt pci

* Gpu allocation by using pci mapping

* Passing gpuPriorityOrder in as an argument rather than making the functions non-static.

* Removing redundant testBed instance calling
This commit is contained in:
saurabhAMD
2024-11-04 10:51:00 -06:00
committed by GitHub
parent 1c45962273
commit 69b2b712ab
9 changed files with 119 additions and 19 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/SendRecvTests.cpp
+4 -2
View File
@@ -27,7 +27,8 @@ namespace RcclUnitTesting
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),
const std::vector<int>& gpuPriorityOrder = testBed.ev.GetGpuPriorityOrder();
testBed.InitComms(TestBed::GetDeviceIdsList(numProcesses, numGpus, ranksPerGpu, gpuPriorityOrder),
{1,2}, //two group, second group sendrecv to self, has 2 coll
testBed.GetNumStreamsPerGroup(1,2),
2);
@@ -119,7 +120,8 @@ namespace RcclUnitTesting
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),
const std::vector<int>& gpuPriorityOrder = testBed.ev.GetGpuPriorityOrder();
testBed.InitComms(TestBed::GetDeviceIdsList(numProcesses, numGpus, ranksPerGpu, gpuPriorityOrder),
{1,2}, //two group, second group sendrecv to self, has 2 coll
testBed.GetNumStreamsPerGroup(1,2),
2);