Update MP UT to support arbitrary # of GPUs; multiple bugfixes (#16)

* Fixing temp file creation/deletion for Clique kernel mode.

* Refactoring of MP unit tests; include bugfixes and general support for any number of GPUs

* GroupCall MP UT properly quits when too many devices specified

* MP UT will programmatically set NCCL_COMM_ID if not specified; updated install script
This commit is contained in:
Stanley Tsang
2021-02-05 17:49:25 -07:00
committed by GitHub
parent 6dfdfef98f
commit d00b7d17bd
23 changed files with 538 additions and 716 deletions
+2 -2
View File
@@ -200,10 +200,10 @@ if ($run_tests); then
if (test -f "./test/UnitTests"); then
if ($run_tests_all); then
./test/UnitTests
NCCL_COMM_ID=$HOSTNAME:55512 ./test/UnitTestsMultiProcess
./test/UnitTestsMultiProcess
else
./test/UnitTests --gtest_filter="BroadcastCorrectnessSweep*:*float32*"
NCCL_COMM_ID=$HOSTNAME:55512 ./test/UnitTestsMultiProcess --gtest_filter="BroadcastMultiProcessCorrectnessSweep*:*float32*"
./test/UnitTestsMultiProcess --gtest_filter="BroadcastMultiProcessCorrectnessSweep*:*float32*"
fi
else
echo "Unit tests have not been built yet; please re-run script with -t to build unit tests."