- removed comment
GRHydro_UpdateMask takes too long to compile
GRHydro_UpdateMask takes more than an hour to compile. This is with Intel Version 12.1.1.256 Build 20111011 on a modern workstation Intel(R) Xeon(R) CPU X5675 @ 3.07GHz:
11110 eschnett 32 12 251m 206m 14m R 100 0.9 68:29.24 fortcom
Could we simplify the source code, e.g. moving the pointer assignments and the actual loops into different files?
Keyword:
Comments (6)
-
-
reporter - removed comment
I am using the options -xHost -safe_cray_ptr -traceback -pad -O3 -ip -no-prec-div. No, I don't see any warnings from the compiler.
-
- removed comment
Could you try if -O2 shows the same instead of -O3?
-
- removed comment
I had not realized it had gotten that bad. Erik had already pointed this out last year around August (to me, who did not do much about it since it does not happen with gcc). One thing to try as well would be to use Cray pointers rather than Fotran pointers. It might also be sufficient to just move the setup and loop part into different Fortran routines in the same file. I'll try and see if eg ifort on bethe shows the same issue.
-
- changed status to resolved
- removed comment
This has been fixed when switching to Cray pointers I believe.
-
- edited description
- changed status to closed
- Log in to comment
It would be interesting to know why this compiler needs to much time here. The code doesn't look particularly complicated and other compilers (other options?) don't have that problem. Do you get any debug/warnings from the compiler that would point to something specific? What are the compiler options you use?