jrood-nrel
commented
1 month ago The segfault with the Intel compiler itself has to do with OpenMP being enabled by default and AeroDyn_Inflow.f90 segfaults the compiler.
Closed marchdf closed 1 month ago
jrood-nrel
commented
1 month ago The segfault with the Intel compiler itself has to do with OpenMP being enabled by default and AeroDyn_Inflow.f90 segfaults the compiler.
marchdf
commented
1 month ago I tried this case again on kestrel and made sure openmp was for real disabled in openfast:
modified CMakeLists.txt
@@ -110,7 +110,7 @@ include(${CMAKE_SOURCE_DIR}/cmake/set_rpath.cmake)
# OpenMP
#-------------------------------------------------------------------------------
-if (OPENMP OR BUILD_FASTFARM OR BUILD_OPENFAST_CPP_API)
+if (OPENMP)
if (OPENMP)
FIND_PACKAGE(OpenMP REQUIRED)
else()(exawind-manager should do this automatically for us). And the case now works and doesn't segfault as noted above. Closing this issue.
A little over a month ago, I was told that people were having segfaults using amr-wind + openfast on Kestrel and using oneapi compilers. I was able to reproduce and get a stack trace:
It is a fairly annoying bug because it does not show up with
gccand (at the time) the oneapi compilers were not able to build in debug. I tried to reproduce earlier this week but now, the oneapi compilers won't build OpenFAST 3.5.3:The current "fix" AFAIK is to just use
gcc. I will be working on this next week hopefully.