[mpich-discuss] Debugging MPI Window

[Palmer, Bruce J]

Thanks for all the suggestions. I’ll try these and see how I do.

Bruce

From: Jeff Hammond via discuss <discuss at mpich.org>
Sent: Thursday, April 18, 2019 10:26 PM
To: discuss at mpich.org
Cc: Jeff Hammond <jeff.science at gmail.com>
Subject: Re: [mpich-discuss] Debugging MPI Window

MPI_Win_{get,set}_name might be an even simpler way to do that.

Jeff

On Thu, Apr 18, 2019 at 1:04 AM Joachim Protze via discuss <discuss at mpich.org<mailto:discuss at mpich.org>> wrote:
Hi Bruce,

I think, you could use the keyval and set/get-attr functions for that
purpose and it should even be portable :)
Without testing the code, I would so something like the following:


#define S1(x) #x
#define S2(x) S1(x)
#define LOCATION __FILE__ " : " S2(__LINE__)
int win_debug_string_key;
const int max_win_debug_string_len = 30;

int My_copy_attr_function(
   MPI_Win oldwin, int win_keyval,
   void *extra_state, void *attribute_val_in,
   void *attribute_val_out, int *flag)
{
   attribute_val_out = strndup(attribute_val_in, *(int*)extra_state);
   flag = 1; /* or set 0 if a copy is intended to be different */
}

int My_delete_attr_function(
   MPI_Win win, int win_keyval,
   void *attribute_val, void *extra_state)
{
   free(attribute_val);
}



/* Define the keyval once */
MPI_Win_create_keyval( My_copy_attr_function, My_delete_attr_function,
&win_debug_string_key, &max_win_debug_string_len );


/* When creating a window */
MPI_Win_create( ..., &win );

/* be aware of the extra byte for the \0 byte */
char * win_debug_string = malloc(sizeof(char)*(max_win_debug_string_len+1));
/* Write the current location in there: */
strncpy(win_debug_string, LOCATION);
/* Bind the debug-string to the new window */
MPI_Win_set_attr( win, win_debug_string_key, win_debug_string );


Best
Joachim


On 4/18/19 12:32 AM, Palmer, Bruce J via discuss wrote:
> Hi,
>
> Strictly for debugging purposes, is there a way in MPICH to determine if
> a MPI_Win created at point A in a code is the same window being accessed
> at point B? It doesn’t need to be portable, I’m just interested in using
> it to fix some bugs (that are probably not associated with MPI).
>
> Bruce Palmer
>
> Computer Scientist
>
> Pacific Northwest National Laboratory
>
> (509) 375-3899
>
>
> _______________________________________________
> discuss mailing list     discuss at mpich.org<mailto:discuss at mpich.org>
> To manage subscription options or unsubscribe:
> https://lists.mpich.org/mailman/listinfo/discuss
>


--
Dipl.-Inf. Joachim Protze

IT Center
Group: High Performance Computing
Division: Computational Science and Engineering
RWTH Aachen University
Seffenter Weg 23
D 52074  Aachen (Germany)
Tel: +49 241 80- 24765
Fax: +49 241 80-624765
protze at itc.rwth-aachen.de<mailto:protze at itc.rwth-aachen.de>
www.itc.rwth-aachen.de<http://www.itc.rwth-aachen.de>

_______________________________________________
discuss mailing list     discuss at mpich.org<mailto:discuss at mpich.org>
To manage subscription options or unsubscribe:
https://lists.mpich.org/mailman/listinfo/discuss
--
Jeff Hammond
jeff.science at gmail.com<mailto:jeff.science at gmail.com>
http://jeffhammond.github.io/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mpich.org/pipermail/discuss/attachments/20190419/93ce5881/attachment.html>