2.21.5-1

Add support for IB SHARP 1PPN operation with user buffers.
Improve support for MNNVL, add NVLS support and multi-clique support.
 * Detect the NVLS clique through NVML
 * Exchange XML between peers in the same NVLS clique and fuse XMLs
   before creating the topology graph.
 * Rework bootstrap allgather algorithms to allow for large allgather
   operations intra-node (XML exchange).
Net/IB: add support for dynamic GID detection.
 * Automatically select RoCEv2/IPv4 interface by default. Allow to
   select IPv6 or even the network/mask.
Reduce NVLS memory usage.
 * Add stepSize as property of a connection to allow for different
   sizes on different peers; set it to 128K for NVLink SHARP.
Improve tuner loading
 * Look for more paths, be more consistent with the network device
   plugin.
 * Also search for tuner support inside the net plugin.
Improve tuner API
 * Add context to support multi-device per process.
Add magic number around comm object to detect comm corruption.
 * Add some basic check around communicators so that we can report a
   problem when a communicator gets corrupted or a wrong comm pointer
   is passed to NCCL.
Fix net/IB error path. Github PR #1164
Fix collnet rail mapping with split comm.
Fix packet reordering issue causing bootstrap mismatch
 * Use a different tag in ncclTransportP2pSetup for the connectInfo
   exchange and the following barrier.
Fix hang when crossNic is inconsistent between ranks.
Fix minCompCap/maxCompCap computation. Github issue #1184


[ROCm/rccl commit: ab2b89c4c3]

projects/rccl/src/include/device.h

@@ -84,6 +84,15 @@ static_assert(NCCL_LL_CLEAN_MASK % NCCL_STEPS == 0, "Invalid NCCL_LL_CLEAN_MASK
 #define NCCL_IPC_READ     0x10
 #define NCCL_NVLS_MIN_POLL 0x20
 
+#define NCCL_MAX_COLLNET_SIZE (1L << 29)
+
+enum ncclRegBufferType {
+  NCCL_REGULAR_BUFFER = 0,
+  NCCL_IPC_REG_BUFFER = 1,
+  NCCL_NVLS_REG_BUFFER = 2,
+  NCCL_COLLNET_REG_BUFFER = 3
+};
+
 struct ncclConnInfo {
   // Regular comm mechanism
   char *buffs[NCCL_NUM_PROTOCOLS]; // Local for recv, remote for send
@@ -93,6 +102,7 @@ struct ncclConnInfo {
 
   int flags;          // Direct communication / other flags
   int shared;         // Buffers are shared
+  int stepSize;       // Step size for the SIMPLE buffer
   void **ptrExchange; // Pointer exchange for direct communication
   uint64_t* redOpArgExchange; // PreOp scaler exchange for direct pull case
 
@@ -157,7 +167,7 @@ struct ncclDirect {
   int down[NCCL_MAX_DIRECT_ARITY];
 };
 
-#define NCCL_MAX_NVLS_ARITY 8
+#define NCCL_MAX_NVLS_ARITY 32
 #define NCCL_MAX_NVLS_TREE_ARITY 3
 struct ncclNvls {
   int out;
@@ -171,6 +181,12 @@ struct ncclNvls {
   int nNodes;
 };
 
+#if __CUDA_ARCH__ >= 900
+#define NCCL_MAX_ARITY NCCL_MAX_NVLS_ARITY
+#else
+#define NCCL_MAX_ARITY NCCL_MAX_DIRECT_ARITY
+#endif
+
 #define NCCL_MAX_CONNS 2
 struct ncclChannelPeer {
   struct ncclConnector send[NCCL_MAX_CONNS];
@@ -212,9 +228,10 @@ struct ncclWorkElem {
   union {
     uint8_t flagBits;
     struct {
-      uint8_t isUsed:1, redOpArgIsPtr:1, regUsed:1, oneNode:1;
+      uint8_t isUsed:1, redOpArgIsPtr:1, oneNode:1;
     };
   };
+  uint8_t regUsed;
   uint8_t nWarps;
   uint8_t direct;
   uint32_t root;