2.20.3-1

Add support for alternating rings, allow for cross-nic rings without cross-rail communication. Add support for user buffer registration for network send/recv. Optimize aggregated operations to better utilize all channels. Add flattening for BCM PCI gen5 switches. Add support for inter-node NVLink communication Add support for port fusion in NET/IB. Add support for ReduceScatter and AllGather using Collnet. Update net API to v8. Fix hang during A2A connection. [ROCm/rccl commit: b6475625fb]
2024-02-05 05:06:02 -08:00
parent 382ae08419
commit 2ab8a3a750
74 changed files with 4632 additions and 2165 deletions
@@ -0,0 +1,42 @@
+#ifndef NCCL_REGISTER_H_
+#define NCCL_REGISTER_H_
+
+enum {
+  NET_REG_COMPLETE = 0x01,
+  NVLS_REG_COMPLETE = 0x02,
+  NVLS_REG_POSSIBLE = 0x04,
+  NVLS_REG_NO_SUPPORT = 0x08
+};
+
+struct ncclReg {
+  // common attributes
+  size_t pages;
+  int refs;
+  uintptr_t addr;
+  uint32_t state;
+  // net reg
+  int nDevs;
+  int devs[MAXCHANNELS];
+  void** handles;
+  // nvls reg
+  uintptr_t baseAddr;
+  size_t baseSize;
+  CUdeviceptr regAddr;
+  size_t regSize;
+  int dev;
+  CUmemGenericAllocationHandle mcHandle;
+  uintptr_t caddrs[NCCL_MAX_LOCAL_RANKS]; /* use to check if NVLS buffers match among intra-node ranks */
+};
+
+struct ncclRegCache {
+  struct ncclReg **slots;
+  int capacity, population;
+  uintptr_t pageSize;
+  void* sComms[MAXCHANNELS];
+  void* rComms[MAXCHANNELS];
+};
+
+ncclResult_t ncclRegCleanup(struct ncclComm* comm);
+ncclResult_t ncclRegFind(struct ncclComm* comm, const void* data, size_t size, struct ncclReg** reg);
+
+#endif