2.22.3-1

Rework core for NVIDIA Trusted Computing
 * Compress work structs so that they are shared between channels
 * Utilize the full amount of kernel argument space permitted (4k)
   before resorting to work fifo.
 * Rework the task preprocessing phase.
 * Use a separate abortDevFlag which is kept in sync with abortFlag
   using cudaMemcpy operations.
 * Rename src/include/align.h to src/include/bitops.h

Add lazy connection establishment for collective operations
 * Move buffer allocation and connection establishment to the first
   collective operation using that algorithm.
 * Accelerate init time and reduce memory usage.
 * Avoid allocating NVLS buffers if all calls are registered.
 * Compute algo/proto in ncclLaunchCollTasksInfo early on.
 * Connect peers in ncclCollPreconnectFunc if not connected already.
 * Also move shared buffer creation to the first send/recv call.

Accelerate intra-node NVLink detection
 * Make each rank only detect NVLinks attached to its GPU.
 * Fuse XMLs to reconstruct the full NVLink topology

Add init profiling to report time spend in different init phases.
 * Report timings of bootstrap, allgather, search, connect, etc.
 * Add new "PROFILE" category for NCCL_DEBUG_SUBSYS.

Add support for PCI p2p on split PCI switches
 * Detect split PCI switches through a kernel module exposing
   switch information.
 * Update the topology XML and graph to add those inter-switch
   connections.

Add cost estimation API
 * Add a new ncclGroupEndSimulate primitive to return the estimated
   time a group would take.

Net/IB: Add separate traffic class for fifo messages
 * Add NCCL_IB_FIFO_TC to control the traffic class of fifo messages
   independently from NCCL_IB_TC.
   Merges PR #1194

Net/IB: Add support for IB router
 * Use flid instead of lid if subnets do not match
 * Warn if flid is 0

Optimizations and fixes for device network offload (unpack)
 * Double the default number of channels
 * Cache netDeviceType
 * Fix save/increment head logic to enable Tree support.

Support ncclGroupStart/End for ncclCommAbort/Destroy
 * Allow Abort/Destroy to be called within a group when managing
   multiple GPUs with a single process.

Improve Tuner API
 * Provide to the plugin the original cost table so that the plugin
   can leave unknown or disabled algo/proto combinations untouched.
 * Remove nvlsSupport and collnetSupport.

Do not print version to stdout when using a debug file
 * Also print version from all processes with INFO debug level.
   Fixes issue #1271

Fix clang warnings in NVTX headers
 * Update NVTX headers to the latest version
   Fixes issue #1270

Disable port fusion in heterogeneous systems
 * Do not fuse ports if a mix of multi-port and single port are detected.

Fix NVLS graphs search for dual NICs.
 * Fix NVLS graph search when we have more than one NIC per GPU.

Fix crash with collnetDirect
 * Add separate graph search for collnetDirect, testing alltoall paths
   and working similarly to the NVLS search.

Fix hang when nodes have different CPU types
 * Add the CPU type to the rank peer info.
 * Align all ranks on the CPU type after the first allgather.
 * Only use the aligned CPU type for all tuning operations.
   Fixes issue #1136
   Fixes issue #1184

Fix performance of registered send/recv operations
 * Allow for single full size operations
 * Add INFO to confirm the registration of send/recv buffers.

Move all sync ops to finalize stage
 * Ensure ncclCommDestroy is non-blocking if ncclCommFinalize has
   been called.

Improve error reporting during SHM segment creation

Improve support of various compilers
   Merges PR #1177
   Merges PR #1228

Allow net and tuner plugins to be statically linked
 * Search for ncclNet or ncclTuner symbols in the main binary.
   Merges PR #979

Plugin examples includes cleanup
 * Harmonize err.h and common.h usage.
 * Add mixed plugin with both net and tuner.

src/include/info.h

@@ -8,28 +8,9 @@
 #define NCCL_INFO_H_
 
 #include "nccl.h"
-#include "device.h"
 #include "collectives.h"
 #include "core.h"
 #include "utils.h"
-#include "strongstream.h"
-#define NCCL_MAX_LOCAL_RANKS 64
-
-typedef enum : uint8_t {
-  ncclPatternRing,
-  ncclPatternRingTwice,
-  ncclPatternPipelineFrom,
-  ncclPatternPipelineTo,
-  ncclPatternTreeUp,
-  ncclPatternTreeDown,
-  ncclPatternTreeUpDown,
-  ncclPatternCollnetChain,
-  ncclPatternCollnetDirect,
-  ncclPatternNvls,
-  ncclPatternNvlsTree,
-  ncclPatternSend,
-  ncclPatternRecv
-} ncclPattern_t;
 
 // Used to pass NCCL call information between functions
 struct ncclInfo {
@@ -47,110 +28,6 @@ struct ncclInfo {
   // Algorithm details
   int chunkSteps;
   int sliceSteps;
-  // Computed later
-  ncclDevRedOpFull opFull;
-  ncclPattern_t pattern;
-  size_t nBytes;
-  size_t aggnBytes;
-  size_t workBytes;
-  size_t sendbuffSize;
-  size_t recvbuffSize;
-  int stepSize;
-  int chunkCount;
-  int chunkSize;
-  int channelId;
-  int workFuncIndex;
-  ncclRegBufferType regBufType;
-  void* regBufSend[NCCL_MAX_LOCAL_RANKS];
-  void* regBufRecv[NCCL_MAX_LOCAL_RANKS];
-  // collnet buffer reg handles
-  void* sendMhandle;
-  void* recvMhandle;
-  // Need to initialize
-  int nThreads;
-  int nChannels;
-  int algorithm;
-  int protocol;
-  bool userTuned;
-  struct ncclInfo *next;
-};
-
-inline ncclResult_t ncclInfoSetDerived(struct ncclInfo* info, int nRanks) {
-  info->nBytes = info->workBytes = info->count * ncclTypeSize(info->datatype);
-  if (info->coll == ncclFuncAllGather || info->coll == ncclFuncBroadcast) {
-    info->count = info->workBytes;
-    info->datatype = ncclInt8;
-  }
-  if (info->coll == ncclFuncAllGather || info->coll == ncclFuncReduceScatter) info->nBytes *= nRanks; // count is per rank
-
-  /* compute buffer size for NVLS buffer registration */
-  if (info->coll == ncclFuncAllGather) {
-    info->sendbuffSize = info->workBytes;
-    info->recvbuffSize = info->sendbuffSize * nRanks;
-  } else if (info->coll == ncclFuncReduceScatter) {
-    info->recvbuffSize = info->workBytes;
-    info->sendbuffSize = info->recvbuffSize * nRanks;
-  } else {
-    info->sendbuffSize = info->recvbuffSize = info->workBytes;
-  }
-  return ncclSuccess;
-}
-
-struct ncclTaskColl {
-  struct ncclTaskColl* next;
-  ncclFunc_t func;
-  void const* sendbuff;
-  void* recvbuff;
-  size_t count;
-  int root;
-  ncclDataType_t datatype;
-  ncclDevRedOpFull op;
-  int chunkSteps, sliceSteps;
-  struct ncclInfo info;
-};
-struct ncclTaskP2p {
-  ncclTaskP2p *next;
-  void *buff;
-  size_t bytes;
-  // Stateful chunk index. If a p2p gets "cut" over two plans this keeps track
-  // of where it left off.
-  int chunk;
-};
-
-struct ncclCudaStreamList {
-  struct ncclCudaStreamList *next;
-  cudaStream_t stream;
-};
-struct ncclTasks {
-  struct Peer {
-    bool sendSeen, recvSeen;
-    struct ncclIntruQueue<struct ncclTaskP2p, &ncclTaskP2p::next> sendQueue;
-    struct ncclIntruQueue<struct ncclTaskP2p, &ncclTaskP2p::next> recvQueue;
-  };
-  struct ncclIntruQueue<struct ncclInfo, &ncclInfo::next> collQueue;
-  // Queue for user-tuned executed collectives
-  struct ncclIntruQueue<struct ncclInfo, &ncclInfo::next> collTunedQueue;
-  // Queue for continuous bytes distribution (CBD) collectives
-  struct ncclIntruQueue<struct ncclInfo, &ncclInfo::next> collCBDQueue;
-  // Queue for collnet
-  struct ncclIntruQueue<struct ncclInfo, &ncclInfo::next> collnetQueue;
-  size_t workBytesTotal;
-  int usableChannels;
-  bool sorted;
-  struct Peer* peers/*[nRanks]*/;
-  int *p2pSendOrder, *p2pRecvOrder;
-  int p2pOrderSteps;
-  int nTasksColl, nTasksP2p;
-
-  // The list of user streams aggregated over all tasks present.
-  struct ncclCudaStreamList* streams;
-  // The most recent user stream. Ignored if streams==nullptr
-  cudaStream_t streamRecent;
-  // The graph capturing all user streams or invalid if none. Thus we restrict the
-  // user that all streams must be captured in the same graph or not captured
-  // at all. Technically we could probably relax this, but that would mean
-  // collecting a different `ncclTasks` per graph and one for non-graph.
-  struct ncclCudaGraph capturingGraph;
 };
 
 #endif