gpu: nvgpu: unit: force gcov.sh to use a single thread

On some host OSes, the OS may allocate virtual addresses that are
larger than can be handled by NVGPU. As a workaround, we noticed
that the host OS will not allocate such large addresses when
running single threaded. There is no other way to prevent this
issue nor is there any guarantee that this workaround will always
work...

JIRA NVGPU-1246

Change-Id: I8c89cf6be9f80bd6fb18fd1688037b9144d2d21c
Signed-off-by: Nicolas Benech <nbenech@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/2135370
GVS: Gerrit_Virtual_Submit
Reviewed-by: Alex Waterman <alexw@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>

userspace/gcov.sh

@@ -60,7 +60,7 @@ find . -name "*.gcda" -exec rm {} +
 find $NVGPU_SRC_ROOT -name "*.gcda" -exec rm {} +
 
 #Run unit tests
-./unit.sh -j 2
+./unit.sh
 OUT=$?
 if [ ! $OUT -eq 0 ]; then
 	echo "ERROR: Unit test run failed."

userspace/unit.sh

@@ -41,7 +41,8 @@ if [ -f nvgpu_unit ]; then
 else
         # running on host
         LD_LIBRARY_PATH="build:build/units"
-        NVGPU_UNIT=build/nvgpu_unit
+        # On host, must run single-threaded to avoid high VAs
+        NVGPU_UNIT="./build/nvgpu_unit --num-threads 1"
 fi
 export LD_LIBRARY_PATH