Hello,
I'm running Pardiso with the out-of-core (OOC) option. When I'm done I call the PARDISO function with phase = -1. Most files generated during the out-of-core process are deleted, but 2 files remain: ABC.sin and ABC.sle, where "ABC" is what I set in MKL_PARDISO_OOC_PATH.
Why aren't these files also deleted? Is there a way to have them automatically deleted along with the other files?
Thanks.
I'm running Pardiso with the out-of-core (OOC) option. When I'm done I call the PARDISO function with phase = -1. Most files generated during the out-of-core process are deleted, but 2 files remain: ABC.sin and ABC.sle, where "ABC" is what I set in MKL_PARDISO_OOC_PATH.
Why aren't these files also deleted? Is there a way to have them automatically deleted along with the other files?
Thanks.
連結已複製
16 回應
I checked how it works with the latest update3 and all temporarily files were automatically deleted (with settingMKL_PARDISO_OOC_KEEP_FILE=1 and by the default). Win32 code (fortran) was used.
--Gennady
Gennady,
I built a Win32 version - still calling from C, though. I have the same problem: the files *.sin and *.sle are not deleted, but the files *.ind, *.jal, *lnz, and *.lup - which were created while the problem was running - were all deleted by Pardiso.
-Arthur
I built a Win32 version - still calling from C, though. I have the same problem: the files *.sin and *.sle are not deleted, but the files *.ind, *.jal, *lnz, and *.lup - which were created while the problem was running - were all deleted by Pardiso.
-Arthur
Gennady,
This is the version information I have:
Major Version= 10
Minor Version= 3
Update Version= 3
Product Status= Product
Build= 20110314
Processor Optimization= Intel Pentium 4 processor with SSE3
It appears I'm using the proper version, but I'll go ahead and download a new version - if my evaluation status allows it.
Thanks.
-Arthur
This is the version information I have:
Major Version= 10
Minor Version= 3
Update Version= 3
Product Status= Product
Build= 20110314
Processor Optimization= Intel Pentium 4 processor with SSE3
It appears I'm using the proper version, but I'll go ahead and download a new version - if my evaluation status allows it.
Thanks.
-Arthur
Gennady,
I clicked on my evaluation link in my email and I'm being prompted to download w_mkl_p_10.2.7.040.exe.
But when I first clicked on that link the file I downloaded was w_mkl_10.3.3.175; this is what I have installed. So it appears to me that I have the proper version installed and if I click on my email link I will end up installing an older, rather than newer version.
Can you let me know where I can download the very latest version from?
Thanks.
-Arthur
I clicked on my evaluation link in my email and I'm being prompted to download w_mkl_p_10.2.7.040.exe.
But when I first clicked on that link the file I downloaded was w_mkl_10.3.3.175; this is what I have installed. So it appears to me that I have the proper version installed and if I click on my email link I will end up installing an older, rather than newer version.
Can you let me know where I can download the very latest version from?
Thanks.
-Arthur
Gennady,
Do you at least see the .sin and .sle files before they are deleted?
I guess I don't know what else to try now. With MKL_PARDISO_OOC_KEEP_FILE set to 0 I see lots of other files left on the disk. If I set it o 1 then the .sin and .sle files still remain, while all others are gone.
Could it be that, in addition to MKL_PARDISO_OOC_KEEP_FILE set to 0 I need an additional flag to have these 2 files deleted and I'm missing this information?
-Arthur
Do you at least see the .sin and .sle files before they are deleted?
I guess I don't know what else to try now. With MKL_PARDISO_OOC_KEEP_FILE set to 0 I see lots of other files left on the disk. If I set it o 1 then the .sin and .sle files still remain, while all others are gone.
Could it be that, in addition to MKL_PARDISO_OOC_KEEP_FILE set to 0 I need an additional flag to have these 2 files deleted and I'm missing this information?
-Arthur
Yes, I see all tmp files ( like *.ind, *.lup, *,sln and etc) during the execution stage.
Quote:"Could it be that, in addition to MKL_PARDISO_OOC_KEEP_FILE set to 0 I need an additional flag to have these 2 files deleted and I'm missing this information?"
no, You don't need to do else...
I will ask the component owner to help with clarifying that problem.
--Gennady
Hello, Arthur!
We faced with similar problem in past. Now it should be resolved. Could you provide as with following info: What does input parameter iparm(60) equal? How much number of threads (MKL_NUM_THREADS) do you use? Please, print out the iparm(57) and iparm(64) after reordering step.
Additionally it would be nice if you set msglvl=1 before reordering and provide us with output information.
Sergey,
I ran with MKL_NUM_THREADS=2 (although I got a message #processors = 1; I have another thread here on this topic). Also,
iparm(60) = 2 on input
On output, iparm(57) = 0 and iparm(64) = 103000115
Thanks.
-Arthur
I ran with MKL_NUM_THREADS=2 (although I got a message #processors = 1; I have another thread here on this topic). Also,
iparm(60) = 2 on input
On output, iparm(57) = 0 and iparm(64) = 103000115
Thanks.
-Arthur
Arthur,
We have finally reproduced the problem with the latest version 10.3 Update3. One Note - on our side the problem is reprocuded withMKL_NUM_THREADS ==1 only. The component owner is confirmed the issue and will work to fix.
This issue has been
submitted to our internal development tracking database for further
investigation, we will inform you once a new update becomes available.
Here is a bug tracking number for your reference: DPD200210337.
--Gennady
Here is a bug tracking number for your reference: DPD200210337.
--Gennady
Gennady,
Thanks for getting to the bottom of this. I appreciate your help and patience throughout this process.
Should I get an update on this thread when the bug is fixed, or what is the procedure for tracking this bug?
-Arthur
Thanks for getting to the bottom of this. I appreciate your help and patience throughout this process.
Should I get an update on this thread when the bug is fixed, or what is the procedure for tracking this bug?
-Arthur
