Merge remote-tracking branch 'github/master'

README.md

@@ -85,6 +85,7 @@ All tests support the same set of arguments :
   * `-S,--report_timestamps <0/1>` Add timestamp (`"%Y-%m-%d %H:%M:%S"`) to each performance report line. Default : 0.
   * `-J,--output_file <file>` Write [JSON] output to filepath. Infer type from suffix (only `json` supported presently).
   * `-T,--timeout <time in seconds>` timeout each test after specified number of seconds. Default : disabled.
+  * `-M,--memory_report <0/1>` enable memory usage report. Default : 0
 
 ### Running multiple operations in parallel
 

src/common.h

@@ -6,7 +6,7 @@
 #ifndef __COMMON_H__
 #define __COMMON_H__
 
-#define NCCL_TESTS_VERSION "2.17.8"
+#define NCCL_TESTS_VERSION "2.17.9"
 
 #include "nccl.h"
 #if NCCL_VERSION_CODE >= NCCL_VERSION(2,28,0)

src/common.mk

@@ -66,7 +66,7 @@ NVCC_GENCODE ?= -gencode=arch=compute_35,code=sm_35 \
                 -gencode=arch=compute_70,code=compute_70
 endif
 
-NVCUFLAGS  := -ccbin $(CXX) $(NVCC_GENCODE) $(CXXSTD)
+NVCUFLAGS  := -ccbin $(CXX) $(NVCC_GENCODE) $(CXXSTD) --extended-lambda
 CXXFLAGS   := $(CXXSTD)
 
 LDFLAGS    := -L${CUDA_LIB} -lcudart -lrt

src/util.cu

@@ -654,9 +654,9 @@ void writeResultFooter(const int errors[], const double bw[], double check_avg_b
     jsonKey("count");      jsonInt(errors[0]);
     jsonKey("okay");       jsonBool(errors[0] == 0);
     jsonFinishObject();
-    jsonKey("average_bus_bandwidith");
+    jsonKey("average_bus_bandwidth");
     jsonStartObject();
-    jsonKey("bandwidith"); jsonDouble(bw[0]);
+    jsonKey("bandwidth"); jsonDouble(bw[0]);
     jsonKey("okay");       check_avg_bw == -1 ? jsonStr("unchecked") : jsonBool(bw[0] >= check_avg_bw*(0.9));
     jsonFinishObject();
   }