David Addison
17c8317cb1
Added detection of IBM/Power NVLink bridge device.
Add NUMA support to PCI distance calculations.
Added NCCL_IGNORE_CPU_AFFINITY env var.
Fix memory leaks; GithubIssue#180
Compiler warning fix; GithubIssue#178
Replace non-standard variable length arrays. GithubIssue#171
Fix Tree+Shared Memory crash. GithubPR#185
Fix LL cleanup hang during long running DL jobs.
Fix NCCL_RINGS environment variable handling.
Added extra checks to catch repeat calls to ncclCommDestroy() GithubIssue#191
Improve bootstrap socket connection reliability at scale.
Fix hostname hashing issue. GithubIssue#187
Code cleanup to rename all non device files from *.cu to *.cc
[ROCm/rccl commit: f40ce73e89]
72 строки
2.1 KiB
Makefile
72 строки
2.1 KiB
Makefile
#
|
|
# Copyright (c) 2015-2019, NVIDIA CORPORATION. All rights reserved.
|
|
#
|
|
# See LICENSE.txt for license information
|
|
#
|
|
|
|
include ../../../makefiles/common.mk
|
|
include ../../../makefiles/version.mk
|
|
|
|
BUILDDIR ?= $(abspath ../../../build)
|
|
OBJDIR := $(BUILDDIR)/obj/collectives/device
|
|
|
|
LIBSRCFILES := all_reduce.cu broadcast.cu reduce.cu all_gather.cu reduce_scatter.cu
|
|
|
|
LIBSRCFILES += functions.cu
|
|
|
|
DEPFILES := $(patsubst %.cu, $(OBJDIR)/%.d, $(LIBSRCFILES))
|
|
DEPENDFILES:= $(DEPFILES:%.d=%.dep)
|
|
STATICLIB := $(OBJDIR)/colldevice.a
|
|
DEVOBJ := $(OBJDIR)/devlink.o
|
|
RULESFILE := $(OBJDIR)/Makefile.rules
|
|
|
|
NVCUFLAGS += -I. -I.. -I$(BUILDDIR)/include -I../../include --compiler-options "-fPIC -fvisibility=hidden"
|
|
|
|
|
|
all: $(STATICLIB)
|
|
|
|
# Dummy rule so that the extra dependency (%.dep) files are preserved by make
|
|
all_deps: $(DEPENDFILES)
|
|
|
|
# Auto-generating the rules per op/reduction/datatype/algorithm
|
|
$(RULESFILE) :
|
|
@printf "Generating %-35s > %s\n" rules $@
|
|
@mkdir -p $(OBJDIR)
|
|
@./gen_rules.sh $(OBJDIR) > $@
|
|
|
|
-include $(RULESFILE)
|
|
|
|
LIBOBJ := $(GENOBJS) $(OBJDIR)/functions.o
|
|
|
|
-include $(DEPFILES)
|
|
|
|
$(STATICLIB): $(LIBOBJ) $(DEVOBJ)
|
|
@printf "Archiving %-35s > %s\n" objects $@
|
|
ar cr $@ $^
|
|
|
|
# We do not want make to build *.d when running make clean.
|
|
# So we only provide targets for .dep which will produce .dep and .d,
|
|
# with only .d being included, and .dep keeping track of what needs to
|
|
# be regenerated.
|
|
$(OBJDIR)/%.dep : %.cu
|
|
@mkdir -p $(OBJDIR)
|
|
@$(NVCC) $(NVCUFLAGS) -M $< -o $@.tmp
|
|
@sed "0,/^.*:/s//$(subst /,\/,$@):/" $@.tmp > $@
|
|
@sed -e 's/.*://' -e 's/\\$$//' < $@.tmp | fmt -1 | \
|
|
sed -e 's/^ *//' -e 's/$$/:/' >> $@
|
|
@rm -f $@.tmp
|
|
@cp $@ $(@:.dep=.d)
|
|
|
|
# Compiled kernels and collectives with relocatable device code ...
|
|
$(OBJDIR)/functions.o : functions.cu $(OBJDIR)/functions.dep
|
|
@printf "Compiling %-35s > %s\n" $< $@
|
|
mkdir -p `dirname $@`
|
|
$(NVCC) $(NVCUFLAGS) -dc $< -o $@
|
|
|
|
# ... and create the device-side linked object with all those.
|
|
$(DEVOBJ) : $(LIBOBJ)
|
|
$(NVCC) $(NVCUFLAGS) -dlink $^ -o $@
|
|
|
|
clean:
|
|
rm -f $(LIBOBJ) $(DEVOBJ) $(DEPFILES) $(DEPENDFILES) $(STATICLIB) test
|