Fix and add MXFP8 GEMM test failures (#326)

* Fix MXFP8 GEMM test
* Fix uninitialized var in GEMM code
* Add Dequantize+GEMM test to check MXFP8 scaling tensor layout

Author: ipanfilo

ci/core.sh

@@ -31,14 +31,14 @@ fi
 check_test_filter "nongemm"
 if [ $? -eq 0 ]; then
     echo ===== Run non GEMM tests =====
-    ctest --test-dir build -j"$n_parallel_jobs" -V --output-on-failure -E "OperatorTest/GEMMTestSuite"
+    ctest --test-dir build -j"$n_parallel_jobs" -V --output-on-failure -E "GEMMTestSuite"
     test $? -eq 0 || test_run_error "non-GEMM"
 fi
 
 check_test_filter "gemm"
 if [ $? -eq 0 ]; then
     echo  ===== Run GEMM tests =====
-    ctest --test-dir build -j"$n_parallel_jobs" -V --output-on-failure -R "OperatorTest/GEMMTestSuite"
+    ctest --test-dir build -j"$n_parallel_jobs" -V --output-on-failure -R "GEMMTestSuite"
     test $? -eq 0 || test_run_error "GEMM"
 fi
 

tests/cpp/CMakeLists.txt

@@ -64,8 +64,7 @@ else()
   project(transformer_engine_tests LANGUAGES HIP CXX)
   # Ask hcc to generate device code during compilation so we can use
   # host linker to link.
-  set(HIP_HCC_FLAGS "${HIP_HCC_FLAGS} -fno-gpu-rdc -Wno-defaulted-function-deleted")
-  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${HIP_HCC_FLAGS}")
+  set(CMAKE_HIP_FLAGS "${CMAKE_HIP_FLAGS} -fno-gpu-rdc -Wno-defaulted-function-deleted -Wno-unused-result")
 endif()
 
 add_subdirectory(../../3rdparty/googletest ${PROJECT_BINARY_DIR}/googletest)