From 15b572751ee510c8ed4add8371efb6882d8f92b4 Mon Sep 17 00:00:00 2001
From: Wenkai Du <43822138+wenkaidu@users.noreply.github.com>
Date: Thu, 7 Apr 2022 13:34:04 -0700
Subject: [PATCH] Increase chunk steps of broadcast and reduce (#528)

---
 src/include/collectives.h | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/src/include/collectives.h b/src/include/collectives.h
index d52d597063..518643cd14 100644
--- a/src/include/collectives.h
+++ b/src/include/collectives.h
@@ -123,11 +123,11 @@ extern __device__ void NCCL_ONERANK_REDUCE_NAME(PreMulSum, double)(struct ncclWo
 #define REDUCESCATTER_SLICESTEPS 2
 #define REDUCESCATTER_CHUNKSTEPS 4
 #define BROADCAST_SLICESTEPS 1
-#define BROADCAST_CHUNKSTEPS 2
+#define BROADCAST_CHUNKSTEPS 4
 #define REDUCE_SLICESTEPS 1
-#define REDUCE_CHUNKSTEPS 2
+#define REDUCE_CHUNKSTEPS 4
 #define SENDRECV_SLICEFACTOR 1
-#define NCCL_MAX_SLICE_PER_CHUNK 2  // max value for CHUNKSTEPS/SLICESTEPS, must accord with above
+#define NCCL_MAX_SLICE_PER_CHUNK 4  // max value for CHUNKSTEPS/SLICESTEPS, must accord with above
 #define ALLTOALL_PIVOT_SLICESTEPS 2
 #define ALLTOALL_PIVOT_CHUNKSTEPS 4