Resync with NCCL 2.11

New operator: mulsum
New test: gather

src/Makefile

@@ -70,7 +70,7 @@ NVLDFLAGS += $(LIBRARIES:%=-l%)
 DST_DIR := $(BUILDDIR)
 SRC_FILES := $(wildcard *.cu)
 OBJ_FILES := $(SRC_FILES:%.cu=${DST_DIR}/%.o)
-BIN_FILES_LIST := all_reduce all_gather broadcast reduce_scatter reduce alltoall scatter sendrecv hypercube
+BIN_FILES_LIST := all_reduce all_gather broadcast reduce_scatter reduce alltoall scatter gather sendrecv hypercube
 BIN_FILES := $(BIN_FILES_LIST:%=${DST_DIR}/%_perf)
 
 build: ${BIN_FILES}