Sylvain Jeaugey
68b542363f
Add scalable init API * Add new ncclCommInitRankScalable to allow for passing multiple unique IDs to the init function. * Spreads the load onto multiple bootstrap roots, allowing for constant bootstrap time. * Requires multiple ranks to create a unique ID, and the CPU-side ID exchange code to call allgather[v] instead of broadcast. Accelerate init bootstrap operations * Reduce the number of calls to allgather. * Allow roots to reply early to ranks when information is already available. * Add an option to use ncclNet instead of sockets to perform bootstrap allgather operations. Add PAT algorithms for Allgather and ReduceScatter * Parallel Aggregated Trees, variation of Bruck algorithm. * Logarithmic number of network steps for small sizes at scale. * Only supports one rank per node at the moment. Add support for registered buffers for intra-node communication. * Allow registered user buffers to be accessed directly intra-node * Avoids extra copies in algorithms which permit it, saving memory bandwidth and helping with compute overlap. Add profiler plugin API * New plugin API for profiling * Supports various levels of profiling, with a hierarchy. Asynchronous graph allocation * Make calls to cudaMalloc and cudaMemcpy during graph allocation asynchronous. * Significantly speeds up graph capture. Use fatal IB asynchronous events to stop network operation * Avoids many other error messages * Only fatal errors are affected; potentially transient errors (e.g. port down) do not cause an immediate stop. Set P2P level to PXB on AMD CPUs when using more than 2 GPUs per node * P2P would cause a significant performance degradation when using many GPUs, and therefore many interleaved data flows. * Disable P2P through the CPU when we have 3+ GPUs per node; keep it enabled when we only have 2 GPUs. Improve the init logs to report the real NCCL function. * Make the log report ncclCommInitRank or ncclCommSplit, rather than the generic ncclCommInitRankFunc. Add a parameter to set the location of the user configuration file. * Add NCCL_CONF_FILE environment variable to set where the user's configuration file resides. Increase default IB timeout * Increase IB timeout value from 18 to 20. * Should help avoid fatal errors on large RoCE systems. Add new check for nvidia peermem * On linux kernels 6.6+, /sys/kernel/mm/memory_peers is no longer present; check for /sys/module/nvidia_peermem/version instead. Fix old performance regression when mixing small and large operations. * Improves distribution of work on channels. Fix crash when NUMA IDs are equal to -1. * Can happen when a NIC is a virtual NIC, or when linux doesn't know which NUMA node a device is attached to * Issue NVIDIA/nccl-tests#233 Fix tree graph search when NCCL_CROSS_NIC is set to 1. * Would force NCCL to use the balanced_tree pattern, thereby disabling LL128 on platforms with 1 GPU+1 NIC per PCI switch. * Would also try to use alternate rings even though it was not needed. Compiler tweaks and fixes * PR #1177 * PR #1228 Fix stack smash * PR #1325 Fixes for multi-node NVLink + IB operation Coverity fixes and comments.
70 rindas
2.3 KiB
C
70 rindas
2.3 KiB
C
/*************************************************************************
|
|
* Copyright (c) 2015-2020, NVIDIA CORPORATION. All rights reserved.
|
|
*
|
|
* See LICENSE.txt for license information
|
|
************************************************************************/
|
|
|
|
#include <stdlib.h>
|
|
|
|
#ifndef NCCL_P2P_H_
|
|
#define NCCL_P2P_H_
|
|
|
|
#include <cuda.h>
|
|
#include <cuda_runtime.h>
|
|
|
|
#include "core.h"
|
|
|
|
#if CUDART_VERSION < 12030
|
|
// MNNVL: FABRIC handle support lifted from CUDA 12.3
|
|
#define CU_DEVICE_ATTRIBUTE_HANDLE_TYPE_FABRIC_SUPPORTED ((CUdevice_attribute)128)
|
|
#define CU_MEM_HANDLE_TYPE_FABRIC ((CUmemAllocationHandleType)0x8ULL)
|
|
#define CU_IPC_HANDLE_SIZE 64
|
|
typedef struct CUmemFabricHandle_st {
|
|
unsigned char data[CU_IPC_HANDLE_SIZE];
|
|
} CUmemFabricHandle_v1;
|
|
typedef CUmemFabricHandle_v1 CUmemFabricHandle;
|
|
#endif
|
|
|
|
typedef union {
|
|
uint64_t data; // Needs to hold a CUmemGenericAllocationHandle for UDS fd support
|
|
CUmemFabricHandle handle;
|
|
} ncclCuDesc;
|
|
|
|
typedef union {
|
|
// Legacy CUDA IPC
|
|
cudaIpcMemHandle_t devIpc;
|
|
// cuMem API support
|
|
struct {
|
|
ncclCuDesc cuDesc;
|
|
CUmemGenericAllocationHandle memHandle;
|
|
};
|
|
} ncclIpcDesc;
|
|
|
|
enum ncclIpcRegType {
|
|
NCCL_IPC_SENDRECV = 0,
|
|
NCCL_IPC_COLLECTIVE = 1
|
|
};
|
|
|
|
struct ncclIpcImpInfo {
|
|
void* rmtRegAddr;
|
|
bool legacyIpcCap;
|
|
uintptr_t offset;
|
|
};
|
|
|
|
struct ncclIpcRegInfo {
|
|
int peerRank;
|
|
void* baseAddr;
|
|
struct ncclProxyConnector* ipcProxyconn;
|
|
struct ncclIpcImpInfo impInfo;
|
|
};
|
|
|
|
ncclResult_t ncclP2pAllocateShareableBuffer(size_t size, int directMap, ncclIpcDesc *ipcDesc, void **ptr);
|
|
ncclResult_t ncclP2pFreeShareableBuffer(ncclIpcDesc *ipcDesc);
|
|
ncclResult_t ncclP2pImportShareableBuffer(struct ncclComm *comm, int peer, size_t size, ncclIpcDesc *ipcDesc, void **devMemPtr);
|
|
ncclResult_t ncclIpcLocalRegisterBuffer(ncclComm* comm, const void* userbuff, size_t buffSize, int* peerRanks, int nPeers, ncclIpcRegType type, int* regBufFlag, uintptr_t* offsetOut, uintptr_t** peerRmtAddrsOut);
|
|
ncclResult_t ncclIpcGraphRegisterBuffer(ncclComm* comm, const void* userbuff, size_t buffSize, int* peerRanks, int nPeers, ncclIpcRegType type, int* regBufFlag, uintptr_t* offsetOut, uintptr_t** peerRmtAddrsOut, void* cleanupQueuePtr, int* nCleanupQueueElts);
|
|
|
|
ncclResult_t ncclIpcDeregBuffer(struct ncclComm* comm, struct ncclIpcRegInfo* regInfo);
|
|
|
|
#endif
|