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.
102 lines
3.3 KiB
Makefile
102 lines
3.3 KiB
Makefile
#
|
|
# Copyright (c) 2015-2020, NVIDIA CORPORATION. All rights reserved.
|
|
#
|
|
# See LICENSE.txt for license information
|
|
#
|
|
|
|
CUDA_HOME ?= /usr/local/cuda
|
|
PREFIX ?= /usr/local
|
|
VERBOSE ?= 0
|
|
KEEP ?= 0
|
|
DEBUG ?= 0
|
|
TRACE ?= 0
|
|
PROFAPI ?= 0
|
|
NVTX ?= 1
|
|
|
|
NVCC = $(CUDA_HOME)/bin/nvcc
|
|
|
|
CUDA_LIB ?= $(CUDA_HOME)/lib64
|
|
CUDA_INC ?= $(CUDA_HOME)/include
|
|
CUDA_VERSION = $(strip $(shell which $(NVCC) >/dev/null && $(NVCC) --version | grep release | sed 's/.*release //' | sed 's/\,.*//'))
|
|
#CUDA_VERSION ?= $(shell ls $(CUDA_LIB)/libcudart.so.* | head -1 | rev | cut -d "." -f -2 | rev)
|
|
CUDA_MAJOR = $(shell echo $(CUDA_VERSION) | cut -d "." -f 1)
|
|
CUDA_MINOR = $(shell echo $(CUDA_VERSION) | cut -d "." -f 2)
|
|
#$(info CUDA_VERSION ${CUDA_MAJOR}.${CUDA_MINOR})
|
|
|
|
|
|
# You should define NVCC_GENCODE in your environment to the minimal set
|
|
# of archs to reduce compile time.
|
|
CUDA8_GENCODE = -gencode=arch=compute_35,code=sm_35 \
|
|
-gencode=arch=compute_50,code=sm_50 \
|
|
-gencode=arch=compute_60,code=sm_60 \
|
|
-gencode=arch=compute_61,code=sm_61
|
|
CUDA9_GENCODE = -gencode=arch=compute_70,code=sm_70
|
|
CUDA11_GENCODE = -gencode=arch=compute_80,code=sm_80
|
|
|
|
CUDA8_PTX = -gencode=arch=compute_61,code=compute_61
|
|
CUDA9_PTX = -gencode=arch=compute_70,code=compute_70
|
|
CUDA11_PTX = -gencode=arch=compute_80,code=compute_80
|
|
|
|
# Include Ampere support if we're using CUDA11 or above
|
|
ifeq ($(shell test "0$(CUDA_MAJOR)" -ge 11; echo $$?),0)
|
|
NVCC_GENCODE ?= $(CUDA8_GENCODE) $(CUDA9_GENCODE) $(CUDA9_PTX) $(CUDA11_GENCODE) $(CUDA11_PTX)
|
|
# Include Volta support if we're using CUDA9 or above
|
|
else ifeq ($(shell test "0$(CUDA_MAJOR)" -ge 9; echo $$?),0)
|
|
NVCC_GENCODE ?= $(CUDA8_GENCODE) $(CUDA9_GENCODE) $(CUDA9_PTX)
|
|
else
|
|
NVCC_GENCODE ?= $(CUDA8_GENCODE) $(CUDA8_PTX)
|
|
endif
|
|
#$(info NVCC_GENCODE is ${NVCC_GENCODE})
|
|
|
|
CXXFLAGS := -DCUDA_MAJOR=$(CUDA_MAJOR) -DCUDA_MINOR=$(CUDA_MINOR) -fPIC -fvisibility=hidden \
|
|
-Wall -Wno-unused-function -Wno-sign-compare -std=c++11 -Wvla \
|
|
-I $(CUDA_INC) \
|
|
$(CXXFLAGS)
|
|
# Maxrregcount needs to be set accordingly to NCCL_MAX_NTHREADS (otherwise it will cause kernel launch errors)
|
|
# 512 : 120, 640 : 96, 768 : 80, 1024 : 60
|
|
# We would not have to set this if we used __launch_bounds__, but this only works on kernels, not on functions.
|
|
NVCUFLAGS := -ccbin $(CXX) $(NVCC_GENCODE) -std=c++11 -Xptxas -maxrregcount=96 -Xfatbin -compress-all
|
|
# Use addprefix so that we can specify more than one path
|
|
NVLDFLAGS := -L${CUDA_LIB} -lcudart -lrt
|
|
|
|
########## GCOV ##########
|
|
GCOV ?= 0 # disable by default.
|
|
GCOV_FLAGS := $(if $(filter 0,${GCOV} ${DEBUG}),,--coverage) # only gcov=1 and debug =1
|
|
CXXFLAGS += ${GCOV_FLAGS}
|
|
NVCUFLAGS += ${GCOV_FLAGS:%=-Xcompiler %}
|
|
LDFLAGS += ${GCOV_FLAGS}
|
|
NVLDFLAGS += ${GCOV_FLAGS:%=-Xcompiler %}
|
|
# $(warning GCOV_FLAGS=${GCOV_FLAGS})
|
|
########## GCOV ##########
|
|
|
|
ifeq ($(DEBUG), 0)
|
|
NVCUFLAGS += -O3
|
|
CXXFLAGS += -O3 -g
|
|
else
|
|
NVCUFLAGS += -O0 -G -g
|
|
CXXFLAGS += -O0 -g -ggdb3
|
|
endif
|
|
|
|
ifneq ($(VERBOSE), 0)
|
|
NVCUFLAGS += -Xptxas -v -Xcompiler -Wall,-Wextra,-Wno-unused-parameter
|
|
CXXFLAGS += -Wall -Wextra
|
|
else
|
|
.SILENT:
|
|
endif
|
|
|
|
ifneq ($(TRACE), 0)
|
|
CXXFLAGS += -DENABLE_TRACE
|
|
endif
|
|
|
|
ifeq ($(NVTX), 0)
|
|
CXXFLAGS += -DNVTX_DISABLE
|
|
endif
|
|
|
|
ifneq ($(KEEP), 0)
|
|
NVCUFLAGS += -keep
|
|
endif
|
|
|
|
ifneq ($(PROFAPI), 0)
|
|
CXXFLAGS += -DPROFAPI
|
|
endif
|