Add a debug level to NCCL and CUDA versions at init

Makefile

@@ -62,23 +62,30 @@ else
 endif
 
 
+NCCL_MAJOR   := 1
+NCCL_MINOR   := 5
+NCCL_PATCH   := 3
+CXXFLAGS  += -DNCCL_MAJOR=$(NCCL_MAJOR) -DNCCL_MINOR=$(NCCL_MINOR) -DNCCL_PATCH=$(NCCL_PATCH)
+
+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 $(CUDA_VERSION) | cut -d "." -f 2)
+CXXFLAGS  += -DCUDA_MAJOR=$(CUDA_MAJOR) -DCUDA_MINOR=$(CUDA_MINOR)
+
 .PHONY : lib clean debclean test mpitest install
 .DEFAULT : lib
 
 INCEXPORTS  := nccl.h
 LIBSRCFILES := libwrap.cu core.cu all_gather.cu all_reduce.cu broadcast.cu reduce.cu reduce_scatter.cu
 LIBNAME     := libnccl.so
-VER_MAJOR   := 1
-VER_MINOR   := 2
-VER_PATCH   := 3
 
 INCDIR := $(BUILDDIR)/include
 LIBDIR := $(BUILDDIR)/lib
 OBJDIR := $(BUILDDIR)/obj
 
 INCTARGETS := $(patsubst %, $(INCDIR)/%, $(INCEXPORTS))
-LIBSONAME  := $(patsubst %,%.$(VER_MAJOR),$(LIBNAME))
+LIBSONAME  := $(patsubst %,%.$(NCCL_MAJOR),$(LIBNAME))
-LIBTARGET  := $(patsubst %,%.$(VER_MAJOR).$(VER_MINOR).$(VER_PATCH),$(LIBNAME))
+LIBTARGET  := $(patsubst %,%.$(NCCL_MAJOR).$(NCCL_MINOR).$(NCCL_PATCH),$(LIBNAME))
 LIBLINK    := $(patsubst lib%.so, -l%, $(LIBNAME))
 LIBOBJ     := $(patsubst %.cu, $(OBJDIR)/%.o, $(filter %.cu, $(LIBSRCFILES)))
 DEPFILES   := $(patsubst %.o, %.d, $(LIBOBJ)) $(patsubst %, %.d, $(TESTBINS)) $(patsubst %, %.d, $(MPITESTBINS))
@@ -171,10 +178,6 @@ $(MPITSTDIR)/% : test/mpi/%.cu $(TSTDEP)
 
 #### PACKAGING ####
 
-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)
-
 DEB_GEN_IN := $(shell ls debian/*.in)
 DEB_GEN    := $(DEB_GEN_IN:.in=)
 
@@ -192,9 +195,9 @@ debclean :
 
 debian/% : debian/%.in
 	@printf "Generating %-25s > %-24s\n" $< $@
-	sed -e "s/\$${nccl:Major}/$(VER_MAJOR)/g" \
+	sed -e "s/\$${nccl:Major}/$(NCCL_MAJOR)/g" \
-	    -e "s/\$${nccl:Minor}/$(VER_MINOR)/g" \
+	    -e "s/\$${nccl:Minor}/$(NCCL_MINOR)/g" \
-	    -e "s/\$${nccl:Patch}/$(VER_PATCH)/g" \
+	    -e "s/\$${nccl:Patch}/$(NCCL_PATCH)/g" \
 	    -e "s/\$${cuda:Major}/$(CUDA_MAJOR)/g" \
 	    -e "s/\$${cuda:Minor}/$(CUDA_MINOR)/g" \
 	    -e "s/\$${nccl:Debian}/$(DEB_REVISION)/g" \

src/core.cu

@@ -501,6 +501,8 @@ static void initDebug() {
   const char* nccl_debug = getenv("NCCL_DEBUG");
   if (nccl_debug == NULL) {
     ncclDebugLevel = NONE;
+  } else if (strcmp(nccl_debug, "VERSION") == 0) {
+    ncclDebugLevel = VERSION;
   } else if (strcmp(nccl_debug, "WARN") == 0) {
     ncclDebugLevel = WARN;
   } else if (strcmp(nccl_debug, "INFO") == 0) {
@@ -654,8 +656,19 @@ static ncclResult_t commUnlinkHostMem(ncclComm_t comm, ncclUniqueId commId, int
   return shmUnlink(rankname);
 }
 
+static void showVersion() {
+  static int shown = 0;
+  if (shown == 0 && ncclDebugLevel >= VERSION) {
+    printf("NCCL version %d.%d.%d compiled with CUDA %d.%d\n", NCCL_MAJOR, NCCL_MINOR, NCCL_PATCH, CUDA_MAJOR, CUDA_MINOR);
+    fflush(stdout);                                              \
+    shown = 1;
+  }
+}
+
 extern "C" DSOGLOBAL
 ncclResult_t ncclCommInitRank(ncclComm_t* newcomm, int ndev, ncclUniqueId commId, int myrank) {
+  if (myrank == 0) showVersion();
+
   if (strlen(commId.internal) < 1 ||
       strlen(commId.internal) >= NCCL_UNIQUE_ID_BYTES) {
     WARN("rank %d invalid commId", myrank);
@@ -735,6 +748,8 @@ extern "C" DSOGLOBAL
 ncclResult_t ncclCommInitAll(ncclComm_t* comms, int ndev, int* devlist) {
   initDebug();
 
+  showVersion();
+
   ncclResult_t res;
   int savedDevice;
   RankEntry* ranks = NULL;

src/core.h

@@ -110,7 +110,7 @@ struct ncclComm {
   ncclNodeRef ptrs[1];
 };
 
-typedef enum {NONE=0, WARN=1, INFO=2, ABORT=3} DebugLevel;
+typedef enum {NONE=0, VERSION=1, WARN=2, INFO=3, ABORT=4} DebugLevel;
 extern DebugLevel ncclDebugLevel;
 
 #define WARN(...) do {                                           \