From 7d25ecc65c76e9f5ae33fb06652e64e526334245 Mon Sep 17 00:00:00 2001
From: amd-jiali <jiali.li@amd.com>
Date: Mon, 29 Dec 2025 16:43:40 -0800
Subject: [PATCH] Add an environment variable to allow user explicitly turn off
 direct AllGather (#2119)

Co-authored-by: Jiali Li <jialili@amd.com>

[ROCm/rccl commit: 935208ad0914805cbb5adad37bd23f3ee0122b40]
---
 projects/rccl/src/rccl_wrap.cc | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/projects/rccl/src/rccl_wrap.cc b/projects/rccl/src/rccl_wrap.cc
index 4f7ee90a06..78d7adeb19 100644
--- a/projects/rccl/src/rccl_wrap.cc
+++ b/projects/rccl/src/rccl_wrap.cc
@@ -351,6 +351,17 @@ ncclResult_t rcclGetProtocolName(int protocol, const char** protocolName) {
 }
 
 bool rcclUseAllGatherDirect(struct ncclComm* comm, size_t& msgSize) {
+  // Check if user explicitly disabled direct AllGather
+  static int userDirectAllGatherInput = -2;
+  if (userDirectAllGatherInput == -2) {
+    const char *inputStr = getenv("RCCL_DIRECT_ALLGATHER_DISABLE");
+    userDirectAllGatherInput = !inputStr ? 0 : 1;
+  }
+  if (userDirectAllGatherInput == 1) {
+    INFO(NCCL_INIT, "RCCL DIRECT ALLGATHER has been disabled.");
+    return false;
+  }
+
   size_t threshold = rcclParamDirectAllGatherThreshold();
 
   if (IsArchMatch(comm->topo->nodes[GPU].nodes[0].gpu.gcn, "gfx950") && threshold != -1) {