Sylvain Jeaugey
920dbe5b35
Optimization for Tree allreduce on A100. Improve aggregation performance. Use shared buffers for inter-node send/recv. Add NVTX profiling hooks. Accelerate alltoall connections by merging communication for all channels. Add support for one hop communication through NVLink, for faster send/recv communication on cubemesh topologies like DGX-1. Improve alltoall scheduling to better balance intra/inter node communication. Increase send/recv parallelism by 8x, each warp sending or receiving to a different peer. Net: move to v4. Net: make flush operation asynchronous to accelerate alltoall. Net: define maximum number of requests. Fix hang when using LL128 protocol after 2^31 steps. Fix #379 : topology injection failing when using less GPUs than described in the XML. Fix #394 : protocol mismatch causing hangs or crashes when using one GPU per node.
61 строка
1.5 KiB
C++
61 строка
1.5 KiB
C++
/*************************************************************************
|
|
* Copyright (c) 2015-2020, NVIDIA CORPORATION. All rights reserved.
|
|
*
|
|
* See LICENSE.txt for license information
|
|
************************************************************************/
|
|
|
|
#ifndef NCCL_CORE_H_
|
|
#define NCCL_CORE_H_
|
|
|
|
#include <pthread.h>
|
|
#include <unistd.h>
|
|
#include <stdlib.h>
|
|
#include <stdint.h>
|
|
#include <algorithm> // For std::min/std::max
|
|
#include "nccl.h"
|
|
|
|
#ifdef PROFAPI
|
|
#define NCCL_API(ret, func, args...) \
|
|
__attribute__ ((visibility("default"))) \
|
|
__attribute__ ((alias(#func))) \
|
|
ret p##func (args); \
|
|
extern "C" \
|
|
__attribute__ ((visibility("default"))) \
|
|
__attribute__ ((weak)) \
|
|
ret func(args)
|
|
#else
|
|
#define NCCL_API(ret, func, args...) \
|
|
extern "C" \
|
|
__attribute__ ((visibility("default"))) \
|
|
ret func(args)
|
|
#endif // end PROFAPI
|
|
|
|
static __inline__ int ncclTypeSize(ncclDataType_t type) {
|
|
switch (type) {
|
|
case ncclInt8:
|
|
case ncclUint8:
|
|
return 1;
|
|
case ncclFloat16:
|
|
return 2;
|
|
case ncclInt32:
|
|
case ncclUint32:
|
|
case ncclFloat32:
|
|
return 4;
|
|
case ncclInt64:
|
|
case ncclUint64:
|
|
case ncclFloat64:
|
|
return 8;
|
|
default:
|
|
return -1;
|
|
}
|
|
}
|
|
|
|
#include "debug.h"
|
|
#include "checks.h"
|
|
#include "alloc.h"
|
|
#include "utils.h"
|
|
#include "param.h"
|
|
#include "nvtx.h"
|
|
|
|
#endif // end include guard
|