Cannot figure out how to correct this; it is more subtle than use of kind attributes.
CHEM_STDPAR2_TGZ.txt (2.0 MB)
I have uploaded the compilation pkg including a Makefile. Thank you in advance.
nvfortran -DCFT_OPENMP -Mpreprocess -Mextend -r8 -Mbyteswapio -module Modules -O3 -fast -cuda -mp=multicore -gpu=mem:managed -gpu:fastmath -stdpar=gpu -acc=gpu -o chmrhs_dcpl_gpu3i.o -c chmrhs_dcpl_gpu3i.f
NVFORTRAN-W-0155-The type of FLOAT is now double precision with -r8 (chmrhs_dcpl_gpu3i.f: 92)
0 inform, 1 warnings, 0 severes, 0 fatal for chmrhs_gpu
NVFORTRAN-F-0000-Internal compiler error. Intrinsic already declared with different signature (chmrhs_dcpl_gpu3i.f: 296)
NVFORTRAN/arm64 Linux 24.11-0: compilation aborted
make: *** [Makefile:26: chmrhs_dcpl_gpu3i.o] Error 2
nvfortran questions have a separate subforum. I can move your question there if you like. You’re not using the NVCC compiler on this code.
Sorry, please move my question to the sub-forum. You’ll likely do it much smoother than
I could. Thank you.
Respectfully,
Don Kenzakowski
Hi Don,
That’s the second error I mentioned in your earlier post and is being tracked as TPR #36900.
The problem seems to be with integer(kind=1) being used as the exponent in pow operations. The work around is to change “kind=1” to “kind=4” for the “nurm” and “nupm” variables declared in “chm_simp3”.
integer(kind=4):: nurm,nupm
Did the work around not fix it?
Now I wasn’t able to recreate the error with a simple example, so there is more to the issue, but the change worked with your earlier version. If this is the case, I’ll investigate further with your new package.
-Mat
Yes, that was the problem. Thanks. I thought I had updated the fix you found. The error message and provided line number is very
confusing.
Respectfully,
Don Kenzakowski