2.12.7-1

Add network communication through another GPU connected with NVLink
(PXN).
Add aggregation of messages coming from different local GPUs through
PXN and going to the same destination.
Add new v5 plugin API with grouped receives and tags.
Add compat for v4 plugins.
Add naming of NCCL threads to help debugging.
Fix NVLink detection and avoid data corruption when some NVLinks are
down.
Add support for Relaxed Ordering for IB.
Add profiling and timing infrastructure.

src/collectives/device/common_kernel.h

@@ -1,5 +1,5 @@
 /*************************************************************************
- * Copyright (c) 2015-2021, NVIDIA CORPORATION. All rights reserved.
+ * Copyright (c) 2015-2022, NVIDIA CORPORATION. All rights reserved.
  *
  * See LICENSE.txt for license information
  ************************************************************************/
@@ -16,10 +16,11 @@
 // Define min for ssize_t
 static __device__ int min(int a, ssize_t b) { return (a < b) ? a : b; }
 
-template <typename T>
-inline __device__ void loadPtr(void** ptr, T* &v) {
-  asm volatile("ld.volatile.global.u64 %0, [%1];"
-      : "=l"(v) : "l"(ptr));
+inline __device__ int loadInt(int* ptr) {
+  int v;
+  asm volatile("ld.volatile.global.u32 %0, [%1];"
+      : "=r"(v) : "l"(ptr));
+  return v;
 }
 
 typedef uint64_t PackType;