Welcome, Guest
Username: Password: Remember me
HELPDESK

Here we can describe more what should be posted here

TOPIC: C_code

C_code 2 months 1 week ago #2371

  • Laura Rontu
  • Laura Rontu's Avatar
  • OFFLINE
  • Administrator
  • Finnish Meteorological Institute
  • Posts: 150
  • Thank you received: 8
This recent change in main.mk causes compilation crash at least for the most recent gcc compiler applied in a debian linux workstation for MUSC. OLD works, NEW doesn't. So this is not a real problem for MUSC, just perhaps good to know.

OLD:
< C_code.o: $(BLCOMP) blacklist/hirlam_blacklist.B
< set -e; cd $(ROOT); \
---
NEW:
> C_code.o: $(BLCOMP) blacklist/hirlam_blacklist.B blacklist/Gen_hirlam_blacklist
> set -e; cd $(ROOT)/blacklist; ./Gen_hirlam_blacklist
> cd $(ROOT); \

C_code 2 months 1 week ago #2372

  • Ulf Andrae
  • Ulf Andrae's Avatar
  • OFFLINE
  • Administrator
  • Posts: 283
  • Thank you received: 30
Laura,

Can you please try to lower the optimization for the blacklist compilation to -O0. I had to on our cluster. It may be that including all these historical blacklist files is not feasible after all.

Ulf

C_code 2 months 1 week ago #2373

  • Laura Rontu
  • Laura Rontu's Avatar
  • OFFLINE
  • Administrator
  • Finnish Meteorological Institute
  • Posts: 150
  • Thank you received: 8
In fact I already tried yesterday, did not see any effect. My version of gcc may be too new to adapt to the old tricks in our code, gcc (Debian 9.2.1-8) 9.2.1 20190909

C_code 2 months 1 week ago #2381

Hi,

As long this is a newer version of GCC my guess is that it might be related with the changed default language standard of the compiler, therefore, the lowered optimization level does not help any longer. What version of Hm/Musc are you trying to compile?

C_code 2 months 2 days ago #2382

Hi,

I tried a recent Fedora box with gcc version 9.2.1 20190827 (Red Hat 9.2.1-1) (GCC) (only the patch level just is a bit lower than yours)

C_code compiles with lowered C optimization levels with the "NEW" version without any problems

I'll check debian testing soon just in case
The following user(s) said Thank You: Laura Rontu

C_code 2 months 2 days ago #2383

  • Laura Rontu
  • Laura Rontu's Avatar
  • OFFLINE
  • Administrator
  • Finnish Meteorological Institute
  • Posts: 150
  • Thank you received: 8
Thank you Martynas for your efforts. I am working within cy43, and yes, debian testing with gcc (Debian 9.2.1-8) 9.2.1 20190909. Since my first posting, I did not have time to return to this issue. For me it is not crucial because in MUSC I definitely do not need odb for anything, so for the time being I am happy with whatever options its compilation includes.

C_code 2 months 1 day ago #2384

Hi,

Just to report that musc compilation of C_code went through with the latest Debian testing out of the box (same compiler version as Laura's) in case reduced optimization levels were used. Both old/new versions work.

If the message that you have seen was like :
gcc: fatal error: Killed signal terminated program cc1

(which better translates to "out of memory allocating something") it might be that it was something higher that -O0 on this part.

C_code 2 months 1 day ago #2385

  • Laura Rontu
  • Laura Rontu's Avatar
  • OFFLINE
  • Administrator
  • Finnish Meteorological Institute
  • Posts: 150
  • Thank you received: 8
Thanks Martynas again. So it seems that the problem is solved, must have been a temporary issue.
Time to create page: 0.085 seconds