From 403a91137cc33110cc0e4207e038865eb64fa5fe Mon Sep 17 00:00:00 2001
From: Nilesh M Negi <Nilesh.Negi@amd.com>
Date: Mon, 4 Dec 2023 16:20:22 -0600
Subject: [PATCH] Fix gcnArch bug in IFC mix build (#998) (#1002)

Signed-off-by: nileshnegi <Nilesh.Negi@amd.com>

[ROCm/rccl commit: bc44e3faa7acdca1edce9d2ff811011b31f734b0]
---
 projects/rccl/src/init.cc | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/projects/rccl/src/init.cc b/projects/rccl/src/init.cc
index b81fbddad8..36a8db183b 100644
--- a/projects/rccl/src/init.cc
+++ b/projects/rccl/src/init.cc
@@ -1700,7 +1700,7 @@ static ncclResult_t ncclCommInitRankFunc(struct ncclAsyncJob* job_) {
   // a CUDA memory reconfig on load (c.f. NVSHMEM issue)
 #ifdef USE_INDIRECT_FUNCTION_CALL
   CUDACHECK(hipGetDeviceProperties(&devProp, 0));
-  if (stackSize > 0 && ncclParamSetStackSize() == 1 && devProp.gcnArch != 940 && devProp.gcnArch != 941 && devProp.gcnArch != 942) {
+  if (stackSize > 0 && ncclParamSetStackSize() == 1 && strcmp(devProp.gcnArchName,"gfx940") != 0 && strcmp(devProp.gcnArchName, "gfx941") != 0 && strcmp(devProp.gcnArchName, "gfx942") != 0) {
     INFO(NCCL_INIT, "Setting cudaLimitStackSize to %zi maxLocalSizeBytes %zi", stackSize, maxLocalSizeBytes);
     CUDACHECKIGNORE(cudaDeviceSetLimit(cudaLimitStackSize, stackSize));
   }