From 1450d42675be325cd3b7a684d4b231eedceb22fb Mon Sep 17 00:00:00 2001
From: Sylvain Jeaugey <sjeaugey@nvidia.com>
Date: Thu, 13 Dec 2018 15:56:12 -0800
Subject: 2.4.2-1

Add tree algorithms for allreduce to improve performance at scale.
Add ncclCommAbort() and ncclCommGetAsyncError() to properly handle
network errors and be permit recover.
Detect initial CPU affinity and no longer escape it.
---
 src/collectives/device/broadcast.cu | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)

(limited to 'src/collectives/device/broadcast.cu')

diff --git a/src/collectives/device/broadcast.cu b/src/collectives/device/broadcast.cu
index 4125de4..b83ee70 100644
--- a/src/collectives/device/broadcast.cu
+++ b/src/collectives/device/broadcast.cu
@@ -4,12 +4,8 @@
  * See LICENSE.txt for license information
  ************************************************************************/
 
-#include "common.h"
 #include "broadcast.h"
+#include "common.h"
 #include "collectives.h"
 
-#define UNROLL 4
-
-#if NCCL_OP == 0
-IMPL_COLL3(ncclBroadcast, copy, FuncSum, i8, int8_t, ncclCollBroadcast, ncclSum, ncclInt8);
-#endif
+IMPL_COLL_C(ncclBroadcast, ncclCollBroadcast);
-- 
cgit v1.2.3