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

[ROCm/rccl commit: 69b2b712ab]
This commit is contained in:
saurabhAMD
2024-11-04 10:51:00 -06:00
committed by GitHub
parent ad1384bea1
commit 69d976532b
9 changed files with 119 additions and 19 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/rccl/test/AllToAllVTests.cpp
+4 -2
View File
@@ -73,7 +73,8 @@ namespace RcclUnitTesting
for (int isMultiProcess : testBed.ev.GetIsMultiProcessList())
{
int const numProcesses = isMultiProcess ? totalRanks : 1;
testBed.InitComms(TestBed::GetDeviceIdsList(numProcesses, totalRanks));
const std::vector<int>& gpuPriorityOrder = testBed.ev.GetGpuPriorityOrder();
testBed.InitComms(TestBed::GetDeviceIdsList(numProcesses, totalRanks, gpuPriorityOrder));
// Prepare AllToAllV options
std::vector<size_t> numInputElements;
@@ -130,7 +131,8 @@ namespace RcclUnitTesting
for (int isMultiProcess : testBed.ev.GetIsMultiProcessList())
{
int const numProcesses = isMultiProcess ? totalRanks : 1;
testBed.InitComms(TestBed::GetDeviceIdsList(numProcesses, totalRanks));
const std::vector<int>& gpuPriorityOrder = testBed.ev.GetGpuPriorityOrder();
testBed.InitComms(TestBed::GetDeviceIdsList(numProcesses, totalRanks, gpuPriorityOrder));
// Prepare AllToAllV options
std::vector<size_t> numInputElements;