PGI User Forum
 SearchSearch   MemberlistMemberlist     RegisterRegister   ProfileProfile    Log inLog in 

CUDA-x86.

Oddity in OpenACC
Goto page Previous  1, 2, 3
 
Post new topic   Reply to topic    PGI User Forum Forum Index -> Accelerator Programming
View previous topic :: View next topic  
Author Message
Maxim Milakov



Joined: 12 Apr 2012
Posts: 4

PostPosted: Tue Apr 17, 2012 5:43 am    Post subject: Reply with quote

Quote:
Note, if the code is too long or you don't want it posted on a public forum, please send it to PGI customer service (trs@pgroup.com) and ask them to forward it to me


Mat, it took me several weeks to build and the run the code. It is a large project with a lot of prerequisites to compile and run. I don't think it is a good idea to consider the whole project as reproduction example.

Quote:
If there is a backwards dependency, the code is not parallel.


I know. That is why I put the second OpenACC parallel directive INSIDE inner loop. There is actually yet another inner loop (the third one) inside that second OpenACC parallel directive.
Back to top
View user's profile
mkcolg



Joined: 30 Jun 2004
Posts: 5952
Location: The Portland Group Inc.

PostPosted: Tue Apr 17, 2012 8:43 am    Post subject: Reply with quote

Quote:
I don't think it is a good idea to consider the whole project as reproduction example.
Ok. If it is a problem with the compiler, then hopefully the problem is also seen elsewhere so we can get it fixed in a future version. If it's a problem with your code, then unfortunately there is not much we can do without an example.

- Mat
Back to top
View user's profile
mmarc__acc



Joined: 09 Dec 2011
Posts: 3

PostPosted: Wed Jul 11, 2012 9:00 pm    Post subject: Reply with quote

Hello,

Not sure if it is related, but I have the same error running the WRF 3.2.1 for PGI Accelerator, using PGI 12.3:

Code:

taskid: 0 hostname: cn15
 Namelist dfi_control not found in namelist.input. Using registry defaults for variables in dfi_control
 Namelist tc not found in namelist.input. Using registry defaults for variables in tc
 Namelist scm not found in namelist.input. Using registry defaults for variables in scm
 Namelist fire not found in namelist.input. Using registry defaults for variables in fire
  Ntasks in X             1 , ntasks in Y             3
 --- NOTE: sst_update is 0, setting io_form_auxinput4 = 0 and auxinput4_interval = 0 for all domains
 --- NOTE: grid_fdda is 0 for domain      1, setting gfdda interval and ending time to 0 for that domain.
 --- NOTE: both grid_sfdda and pxlsm_soil_nudge are 0 for domain      1, setting sgfdda interval and ending time to 0 for that domain.
 --- NOTE: obs_nudge_opt is 0 for domain      1, setting obs nudging interval and ending time to 0 for that domain.
 --- NOTE: num_soil_layers has been set to      5
 WRF V3.2.1 MODEL
  *************************************
  Parent domain
  ids,ide,jds,jde             1          425            1          300
  ims,ime,jms,jme            -4          430           -4          107
  ips,ipe,jps,jpe             1          425            1          100
  *************************************
 DYNAMICS OPTION: Eulerian Mass Coordinate
  alloc_space_field: domain             1 ,     911328204  bytes allocated
  RESTART run: opening wrfrst_d01_2001-10-24_03_00_00 for reading
Timing for processing lateral boundary for domain        1:    0.07996 elapsed seconds.
 WRF NUMBER OF TILES FROM OMP_GET_MAX_THREADS =   1
 WRF TILE   1 IS      1 IE    425 JS      1 JE    100
 WRF NUMBER OF TILES =   1
call to cuMemFree returned error 700: Launch failed
CUDA driver version: 4020


I'm guessing the Accelerator mode for WRF I'm trying to run is exactly the code developed by Michael & Craig, that was working for them with earlier versions of compiler.


- Dima.
Back to top
View user's profile
mkcolg



Joined: 30 Jun 2004
Posts: 5952
Location: The Portland Group Inc.

PostPosted: Thu Jul 12, 2012 11:15 am    Post subject: Reply with quote

Hi Dima,

I believe this a known issue where we unfortunately broke the Accelerator version of this code when implementing OpenACC. I know our team is working on fixing it but I don't know the details. Craig knows more about it but he's on vacation this week. I'll ask him about it when he gets back.

- Mat
Back to top
View user's profile
Display posts from previous:   
Post new topic   Reply to topic    PGI User Forum Forum Index -> Accelerator Programming All times are GMT - 7 Hours
Goto page Previous  1, 2, 3
Page 3 of 3

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum


Powered by phpBB © phpBB Group