Set NVHPC maximum optimization level to -O1 for now (#3800)

* Set NVHPC maximum optimization level to -O1 for now

Compiling HDF5 with NVHPC 23.5 - 23.9 results in test failures in
4 different test files that need to be resolved. Since those tests
pass with an optimization level of -O1 (and -O0) and it is currently
unclear whether the test failures are due to issues in HDF5 or issues
in the 'nvc' compiler, set the maximum optimization level for NVHPC
to -O1 until the test failures are resolved.

* Disable nvhpc Java testing in CMake and amend known issues

* Re-enable testing of Autotools nvhpc

.github/workflows/nvhpc-auto.yml

@@ -67,11 +67,11 @@ jobs:
 
       # RUN TESTS
       # NORMAL
-#      - name: Autotools Run Tests
-#        run: |
-#          export PATH=/opt/nvidia/hpc_sdk/Linux_x86_64/23.7/comm_libs/openmpi4/bin:/opt/nvidia/hpc_sdk/Linux_x86_64/23.7/compilers/bin:$PATH
-#          make check -j
-#        working-directory: ${{ runner.workspace }}/build
+      - name: Autotools Run Tests
+        run: |
+          export PATH=/opt/nvidia/hpc_sdk/Linux_x86_64/23.7/comm_libs/openmpi4/bin:/opt/nvidia/hpc_sdk/Linux_x86_64/23.7/compilers/bin:$PATH
+          make check -j
+        working-directory: ${{ runner.workspace }}/build
 
       # INSTALL (note that this runs even when we don't run the tests)
       - name: Autotools Install

.github/workflows/nvhpc-cmake.yml

@@ -56,7 +56,7 @@ jobs:
           -DLIBAEC_USE_LOCALCONTENT=OFF \
           -DZLIB_USE_LOCALCONTENT=OFF \
           -DHDF5_BUILD_FORTRAN:BOOL=ON \
-          -DHDF5_ENABLE_ASSERTS:BOOL=ON \
+          -DHDF5_BUILD_JAVA:BOOL=OFF \
           -DMPIEXEC_MAX_NUMPROCS:STRING="2" \
           $GITHUB_WORKSPACE
           cat src/libhdf5.settings
@@ -69,8 +69,8 @@ jobs:
         working-directory: ${{ runner.workspace }}/build
 
       # RUN TESTS
-#      - name: CMake Run Tests
-#        shell: bash
-#        run: |
-#          ctest . --parallel 2 -C ${{ inputs.build_mode }} -V
-#        working-directory: ${{ runner.workspace }}/build
+      - name: CMake Run Tests
+        shell: bash
+        run: |
+          ctest . --parallel 2 -C ${{ inputs.build_mode }} -V
+        working-directory: ${{ runner.workspace }}/build

config/cmake/HDFCompilerFlags.cmake

@@ -56,6 +56,44 @@ if (CMAKE_C_COMPILER_ID STREQUAL "NVHPC" )
   else ()
     set (CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Mbounds -g")
   endif ()
+
+  # With at least NVHPC 23.5 - 23.9, compiling with -O2 or higher and -DNDEBUG
+  # appears to have issues that manifest in the tests as incorrect metadata
+  # checksums being read or memory being corrupted. Compiling without -DNDEBUG
+  # does not appear to have these issues, but is not ideal due to compiling in
+  # asserts and other library debug code. Compiling with -O1 also does not appear
+  # to have these issues, so set maximum optimization level to -O1 for now until
+  # it can be determined whether these issues are compiler-specific or issues
+  # in the library.
+  set (cmake_c_flags_minsizerel_edited "${CMAKE_C_FLAGS_MINSIZEREL}")
+  string (REPLACE "-O2" "" cmake_c_flags_minsizerel_edited "${cmake_c_flags_minsizerel_edited}")
+  string (REPLACE "-O3" "" cmake_c_flags_minsizerel_edited "${cmake_c_flags_minsizerel_edited}")
+  string (REPLACE "-O4" "" cmake_c_flags_minsizerel_edited "${cmake_c_flags_minsizerel_edited}")
+  string (REPLACE "-Ofast" "" cmake_c_flags_minsizerel_edited "${cmake_c_flags_minsizerel_edited}")
+  string (REPLACE "-fast" "" cmake_c_flags_minsizerel_edited "${cmake_c_flags_minsizerel_edited}")
+  string (STRIP "${cmake_c_flags_minsizerel_edited}" cmake_c_flags_minsizerel_edited)
+  string (PREPEND cmake_c_flags_minsizerel_edited "-O1 ")
+  set (CMAKE_C_FLAGS_MINSIZEREL "${cmake_c_flags_minsizerel_edited}")
+
+  set (cmake_c_flags_release_edited "${CMAKE_C_FLAGS_RELEASE}")
+  string (REPLACE "-O2" "" cmake_c_flags_release_edited "${cmake_c_flags_release_edited}")
+  string (REPLACE "-O3" "" cmake_c_flags_release_edited "${cmake_c_flags_release_edited}")
+  string (REPLACE "-O4" "" cmake_c_flags_release_edited "${cmake_c_flags_release_edited}")
+  string (REPLACE "-Ofast" "" cmake_c_flags_release_edited "${cmake_c_flags_release_edited}")
+  string (REPLACE "-fast" "" cmake_c_flags_release_edited "${cmake_c_flags_release_edited}")
+  string (STRIP "${cmake_c_flags_release_edited}" cmake_c_flags_release_edited)
+  string (PREPEND cmake_c_flags_release_edited "-O1 ")
+  set (CMAKE_C_FLAGS_RELEASE "${cmake_c_flags_release_edited}")
+
+  set (cmake_c_flags_relwithdebinfo_edited "${CMAKE_C_FLAGS_RELWITHDEBINFO}")
+  string (REPLACE "-O2" "" cmake_c_flags_relwithdebinfo_edited "${cmake_c_flags_relwithdebinfo_edited}")
+  string (REPLACE "-O3" "" cmake_c_flags_relwithdebinfo_edited "${cmake_c_flags_relwithdebinfo_edited}")
+  string (REPLACE "-O4" "" cmake_c_flags_relwithdebinfo_edited "${cmake_c_flags_relwithdebinfo_edited}")
+  string (REPLACE "-Ofast" "" cmake_c_flags_relwithdebinfo_edited "${cmake_c_flags_relwithdebinfo_edited}")
+  string (REPLACE "-fast" "" cmake_c_flags_relwithdebinfo_edited "${cmake_c_flags_relwithdebinfo_edited}")
+  string (STRIP "${cmake_c_flags_relwithdebinfo_edited}" cmake_c_flags_relwithdebinfo_edited)
+  string (PREPEND cmake_c_flags_relwithdebinfo_edited "-O1 ")
+  set (CMAKE_C_FLAGS_RELWITHDEBINFO "${cmake_c_flags_relwithdebinfo_edited}")
 endif ()
 
 if (CMAKE_COMPILER_IS_GNUCC)

config/nvidia-flags

@@ -76,7 +76,8 @@ if test "X-nvc" = "X-$cc_vendor" -o "X-nvcc" = "X-$cc_vendor"; then
     ##############
 
     # NDEBUG is handled explicitly by the configure script
-    PROD_CFLAGS="-fast"
+    #PROD_CFLAGS="-fast"
+    PROD_CFLAGS=""           # -fast implies -O2 and -O2+ currently has test failures.
 
     #########
     # Debug #
@@ -106,7 +107,8 @@ if test "X-nvc" = "X-$cc_vendor" -o "X-nvcc" = "X-$cc_vendor"; then
     ################
 
     HIGH_OPT_CFLAGS="-O1"         # -O2+ currently has test failures.
-    DEBUG_OPT_CFLAGS="-gopt -O2"
+    #DEBUG_OPT_CFLAGS="-gopt -O2"
+    DEBUG_OPT_CFLAGS="-gopt -O1"  # -O2+ currently has test failures.
     NO_OPT_CFLAGS="-O0"
 
     #################

release_docs/RELEASE.txt

@@ -1158,6 +1158,28 @@ Platforms Tested
 Known Problems
 ==============
 
+    When HDF5 is compiled with NVHPC versions 23.5 - 23.9 (additional versions may
+    also be applicable) and with -O2 (or higher) and -DNDEBUG, test failures occur
+    in the following tests:
+
+        H5PLUGIN-filter_plugin
+        H5TEST-flush2
+        H5TEST-testhdf5-base
+        MPI_TEST_t_filters_parallel
+
+    Since these tests pass with an optimization level of -O1 (and -O0) and it is
+    currently unclear whether the test failures are due to issues in HDF5 or issues
+    in the 'nvc' compiler, the maximum optimization level for NVHPC has been set
+    to -O1 until the test failures can be resolved. Note that even at -O1 optimization
+    level, there still appears to be a sporadic test failure in the Java JUnit tests
+    that has occasionally been seen in JUnit-TestH5Pfapl and JUnit-TestH5D. It is also
+    unclear whether this is an issue in HDF5 or with the 'nvc' compiler. Finally, note
+    that NVHPC 23.9 will fail to compile the test/tselect.c test file with a compiler
+    error of 'use of undefined value' when the optimization level is -O2 or higher.
+    Nvidia is aware of this issue and has suggested lowering the optimization level to
+    -O1 for the time being:
+    https://forums.developer.nvidia.com/t/hdf5-no-longer-compiles-with-nv-23-9/269045.
+
     IEEE standard arithmetic enables software to raise exceptions such as overflow,
     division by zero, and other illegal operations without interrupting or halting
     the program flow. The HDF5 C library intentionally performs these exceptions.