AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

2.19.1-1

Browse Source 
Add local user buffer registration for NVLink SHARP.
Add tuning plugin support.
Increase net API to v7 to allow for device-side packet reordering;
remove support for v4 plugins.
Add support for RoCE ECE.
Add support for C2C links.
Better detect SHM allocation failures to avoid crash with Bus Error.
Fix missing thread unlocks in bootstrap (Fixes #936).
Disable network flush by default on H100.
Move device code from src/collectives/device to src/device.
This commit is contained in:
Sylvain Jeaugey
2023-09-26 05:47:28 -07:00
parent 559b70f86c
commit f9c3dc251e
108 changed files with 4749 additions and 2059 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ext-net/example/Makefile
+1 -1
View File
@@ -5,7 +5,7 @@
#
NCCL_HOME:=../../build/
CUDA_HOME:=/usr/local/cuda
INC:= -I$(NCCL_HOME)/include -I$(CUDA_HOME)/include
INC:= -I$(NCCL_HOME)/include -I$(CUDA_HOME)/include -Inccl
PLUGIN_SO:=libnccl-net.so
default: $(PLUGIN_SO)
ext-net/example/nccl/net.h
+1
View File
@@ -24,6 +24,7 @@ typedef enum {NCCL_INIT=1, NCCL_COLL=2, NCCL_P2P=4, NCCL_SHM=8, NCCL_NET=16, NCC
typedef void (*ncclDebugLogger_t)(ncclDebugLogLevel level, unsigned long flags, const char *file, int line, const char *fmt, ...);
#include "net_v7.h"
#include "net_v6.h"
#include "net_v5.h"
#include "net_v4.h"
ext-net/example/nccl/net_device.h
+31
View File
@@ -0,0 +1,31 @@
/*************************************************************************
* Copyright (c) 2023-2023, NVIDIA CORPORATION. All rights reserved.
*
* See LICENSE.txt for license information
************************************************************************/
#ifndef NET_DEVICE_H_
#define NET_DEVICE_H_
#include "net_device.h"
#define NCCL_NET_DEVICE_INVALID_VERSION 0x0
#define NCCL_NET_MTU_SIZE 4096
// Arbitrary version number - A given NCCL build will only be compatible with a single device networking plugin
// version. NCCL will check the supplied version number from net->getProperties() and compare to its internal version.
#define NCCL_NET_DEVICE_UNPACK_VERSION 0x7
typedef enum {NCCL_NET_DEVICE_HOST=0, NCCL_NET_DEVICE_UNPACK=1} ncclNetDeviceType;
typedef struct {
ncclNetDeviceType netDeviceType; // Network offload type
int netDeviceVersion; // Version number for network offload
void* handle;
size_t size;
int needsProxyProgress;
} ncclNetDeviceHandle_v7_t;
typedef ncclNetDeviceHandle_v7_t ncclNetDeviceHandle_t;
#endif
ext-net/example/nccl/net_v6.h
-2
View File
@@ -18,8 +18,6 @@ typedef struct {
int maxRecvs; // Maximum number of grouped receives.
}ncclNetProperties_v6_t;
typedef ncclNetProperties_v6_t ncclNetProperties_t;
typedef struct {
// Name of the network (mainly for logs)
const char* name;
ext-net/example/nccl/net_v7.h
+79
View File
@@ -0,0 +1,79 @@
/*
* Copyright (c) 2017-2022, NVIDIA CORPORATION. All rights reserved.
*/
#ifndef NCCL_NET_V7_H_
#define NCCL_NET_V7_H_
#include "net_device.h"
typedef struct {
char* name; // Used mostly for logging.
char* pciPath; // Path to the PCI device in /sys.
uint64_t guid; // Unique identifier for the NIC chip. Important for
// cards with multiple PCI functions (Physical or virtual).
int ptrSupport; // [NCCL_PTR_HOST|NCCL_PTR_CUDA|NCCL_PTR_DMABUF]
int speed; // Port speed in Mbps.
int port; // Port number.
float latency; // Network latency
int maxComms; // Maximum number of comms we can create
int maxRecvs; // Maximum number of grouped receives.
ncclNetDeviceType netDeviceType; // Network offload type
int netDeviceVersion; // Version number for network offload
} ncclNetProperties_v7_t;
typedef ncclNetProperties_v7_t ncclNetProperties_t;
typedef struct {
// Name of the network (mainly for logs)
const char* name;
// Initialize the network.
ncclResult_t (*init)(ncclDebugLogger_t logFunction);
// Return the number of adapters.
ncclResult_t (*devices)(int* ndev);
// Get various device properties.
ncclResult_t (*getProperties)(int dev, ncclNetProperties_v7_t* props);
// Create a receiving object and provide a handle to connect to it. The
// handle can be up to NCCL_NET_HANDLE_MAXSIZE bytes and will be exchanged
// between ranks to create a connection.
ncclResult_t (*listen)(int dev, void* handle, void** listenComm);
// Connect to a handle and return a sending comm object for that peer.
// This call must not block for the connection to be established, and instead
// should return successfully with sendComm == NULL with the expectation that
// it will be called again until sendComm != NULL.
ncclResult_t (*connect)(int dev, void* handle, void** sendComm, ncclNetDeviceHandle_v7_t** sendDevComm);
// Finalize connection establishment after remote peer has called connect.
// This call must not block for the connection to be established, and instead
// should return successfully with recvComm == NULL with the expectation that
// it will be called again until recvComm != NULL.
ncclResult_t (*accept)(void* listenComm, void** recvComm, ncclNetDeviceHandle_v7_t** recvDevComm);
// Register/Deregister memory. Comm can be either a sendComm or a recvComm.
// Type is either NCCL_PTR_HOST or NCCL_PTR_CUDA.
ncclResult_t (*regMr)(void* comm, void* data, int size, int type, void** mhandle);
/* DMA-BUF support */
ncclResult_t (*regMrDmaBuf)(void* comm, void* data, size_t size, int type, uint64_t offset, int fd, void** mhandle);
ncclResult_t (*deregMr)(void* comm, void* mhandle);
// Asynchronous send to a peer.
// May return request == NULL if the call cannot be performed (or would block)
ncclResult_t (*isend)(void* sendComm, void* data, int size, int tag, void* mhandle, void** request);
// Asynchronous recv from a peer.
// May return request == NULL if the call cannot be performed (or would block)
ncclResult_t (*irecv)(void* recvComm, int n, void** data, int* sizes, int* tags, void** mhandles, void** request);
// Perform a flush/fence to make sure all data received with NCCL_PTR_CUDA is
// visible to the GPU
ncclResult_t (*iflush)(void* recvComm, int n, void** data, int* sizes, void** mhandles, void** request);
// Test whether a request is complete. If size is not NULL, it returns the
// number of bytes sent/received.
ncclResult_t (*test)(void* request, int* done, int* sizes);
// Close and free send/recv comm objects
ncclResult_t (*closeSend)(void* sendComm);
ncclResult_t (*closeRecv)(void* recvComm);
ncclResult_t (*closeListen)(void* listenComm);
// Copy the given mhandle to a dptr in a format usable by this plugin's device code
ncclResult_t (*getDeviceMr)(void* comm, void* mhandle, void** dptr_mhandle);
// Notify the plugin that a recv has completed by the device
ncclResult_t (*irecvConsumed)(void* recvComm, int n, void* request);
} ncclNet_v7_t;
#endif // end include guard
ext-net/example/plugin.c
+48 -13
View File
@@ -4,7 +4,7 @@
* See LICENSE.txt for license information
************************************************************************/
#include <nccl/net.h>
#include "net.h"
#define __hidden __attribute__ ((visibility("hidden")))
@@ -15,14 +15,14 @@ __hidden ncclResult_t pluginDevices(int* ndev) { *ndev = 0; return ncclSuccess;
__hidden ncclResult_t pluginPciPath(int dev, char** path) { return ncclInternalError; }
__hidden ncclResult_t pluginPtrSupport(int dev, int* supportedTypes) { return ncclInternalError; }
__hidden ncclResult_t pluginGetProperties(int dev, ncclNetProperties_v6_t* props) {
__hidden ncclResult_t pluginGetProperties(int dev, ncclNetProperties_v7_t* props) {
//pluginPciPath(dev, &props.pciPath);
//pluginPtrSupport(dev, &props.ptrSupport);
return ncclInternalError;
}
__hidden ncclResult_t pluginListen(int dev, void* handle, void** listenComm) { return ncclInternalError; }
__hidden ncclResult_t pluginConnect(int dev, void* handle, void** sendComm) { return ncclInternalError; }
__hidden ncclResult_t pluginAccept(void* listenComm, void** recvComm) { return ncclInternalError; }
__hidden ncclResult_t pluginConnect(int dev, void* handle, void** sendComm, ncclNetDeviceHandle_v7_t** sendDevComm) { return ncclInternalError; }
__hidden ncclResult_t pluginAccept(void* listenComm, void** recvComm, ncclNetDeviceHandle_v7_t** recvDevComm) { return ncclInternalError; }
__hidden ncclResult_t pluginRegMr(void* collComm, void* data, int size, int type, void** mhandle) { return ncclInternalError; }
__hidden ncclResult_t pluginRegMrDmaBuf(void* collComm, void* data, size_t size, int type, uint64_t offset, int fd, void** mhandle) { return ncclInternalError; }
__hidden ncclResult_t pluginDeregMr(void* collComm, void* mhandle) { return ncclInternalError;}
@@ -33,10 +33,12 @@ __hidden ncclResult_t pluginTest(void* request, int* done, int* size) { return n
__hidden ncclResult_t pluginCloseSend(void* sendComm) { return ncclInternalError; }
__hidden ncclResult_t pluginCloseRecv(void* recvComm) { return ncclInternalError; }
__hidden ncclResult_t pluginCloseListen(void* listenComm) { return ncclInternalError; }
__hidden ncclResult_t pluginIrecvConsumed(void* recvComm, int n, void* request) { return ncclInternalError; }
__hidden ncclResult_t pluginGetDeviceMr(void* comm, void* mhandle, void** dptr_mhandle) { return ncclInternalError; }
#define PLUGIN_NAME "Plugin"
const ncclNet_v6_t ncclNetPlugin_v6 = {
const ncclNet_v7_t ncclNetPlugin_v7 = {
.name = PLUGIN_NAME,
.init = pluginInit,
.devices = pluginDevices,
@@ -54,6 +56,37 @@ const ncclNet_v6_t ncclNetPlugin_v6 = {
.closeSend = pluginCloseSend,
.closeRecv = pluginCloseRecv,
.closeListen = pluginCloseListen,
.getDeviceMr = pluginGetDeviceMr,
.irecvConsumed = pluginIrecvConsumed,
};
__hidden ncclResult_t pluginGetProperties_v6(int dev, ncclNetProperties_v6_t* props) {
//pluginPciPath(dev, &props.pciPath);
//pluginPtrSupport(dev, &props.ptrSupport);
return ncclInternalError;
}
__hidden ncclResult_t pluginConnect_v6(int dev, void* handle, void** sendComm) { return ncclInternalError; }
__hidden ncclResult_t pluginAccept_v6(void* listenComm, void** recvComm) { return ncclInternalError; }
const ncclNet_v6_t ncclNetPlugin_v6 = {
.name = PLUGIN_NAME,
.init = pluginInit,
.devices = pluginDevices,
.getProperties = pluginGetProperties_v6,
.listen = pluginListen,
.connect = pluginConnect_v6,
.accept = pluginAccept_v6,
.regMr = pluginRegMr,
.regMrDmaBuf = pluginRegMrDmaBuf,
.deregMr = pluginDeregMr,
.isend = pluginIsend,
.irecv = pluginIrecv,
.iflush = pluginIflush,
.test = pluginTest,
.closeSend = pluginCloseSend,
.closeRecv = pluginCloseRecv,
.closeListen = pluginCloseListen
};
/* v5 Compat */
@@ -61,10 +94,10 @@ const ncclNet_v5_t ncclNetPlugin_v5 = {
.name = PLUGIN_NAME,
.init = pluginInit,
.devices = pluginDevices,
.getProperties = pluginGetProperties,
.getProperties = pluginGetProperties_v6,
.listen = pluginListen,
.connect = pluginConnect,
.accept = pluginAccept,
.connect = pluginConnect_v6,
.accept = pluginAccept_v6,
.regMr = pluginRegMr,
.deregMr = pluginDeregMr,
.isend = pluginIsend,
@@ -79,7 +112,7 @@ const ncclNet_v5_t ncclNetPlugin_v5 = {
/* v4 Compat */
static ncclResult_t pluginGetProperties_v4(int dev, ncclNetProperties_v4_t* props) {
ncclNetProperties_v6_t props_v6;
ncclResult_t ret = pluginGetProperties(dev, &props_v6);
ncclResult_t ret = pluginGetProperties_v6(dev, &props_v6);
if (ret != ncclSuccess) return ret;
props->name = props_v6.name;
props->pciPath = props_v6.pciPath;
@@ -103,14 +136,16 @@ static ncclResult_t pluginIflush_v4(void* recvComm, void* data, int size, void*
static ncclResult_t pluginConnect_v4(int dev, void* handle, void** sendComm) {
ncclResult_t ret;
do {
ret = pluginConnect(dev, handle, sendComm);
ncclNetDeviceHandle_v7_t* handle = NULL;
ret = pluginConnect(dev, handle, sendComm, &handle);
} while (ret == ncclSuccess && *sendComm == NULL);
return ret;
}
static ncclResult_t pluginAccept_v4(void* listenComm, void** recvComm) {
ncclResult_t ret;
do {
ret = pluginAccept(listenComm, recvComm);
ncclNetDeviceHandle_v7_t* handle = NULL;
ret = pluginAccept(listenComm, recvComm, &handle);
} while (ret == ncclSuccess && *recvComm == NULL);
return ret;
}
@@ -151,12 +186,12 @@ static ncclResult_t pluginInit_v3(ncclDebugLogger_t logFunction) {
static ncclResult_t pluginListen_v3(int dev, void* handle, void** listenComm) {
char pluginHandle[NCCL_NET_HANDLE_MAXSIZE];
ncclResult_t ret = pluginListen(dev, &pluginHandle, listenComm);
memcpy(handle, &pluginHandle, NCCL_NET_HANDLE_MAXSIZE_V3);
memcpy(handle, &pluginHandle, NCCL_NET_HANDLE_MAXSIZE_V4);
return ret;
}
static ncclResult_t pluginConnect_v3(int dev, void* handle, void** sendComm) {
char pluginHandle[NCCL_NET_HANDLE_MAXSIZE];
memcpy(&pluginHandle, handle, NCCL_NET_HANDLE_MAXSIZE_V3);
memcpy(&pluginHandle, handle, NCCL_NET_HANDLE_MAXSIZE_V4);
return pluginConnect_v4(dev, &pluginHandle, sendComm);
}
const ncclNet_v3_t ncclNetPlugin_v3 = {