Charm++ parallel programming system

[Nikhil Jain <nikhil.life AT gmail.com>]

I do not think we have output redirection to individual files based on
PE. For LRTS based layer, machine-common-core contains code for
redirecting to a common file using option +stdout
(http://charm.cs.illinois.edu/manuals/html/charm++/26.html#SECTION04440000000000000000).
For netlrts, this redirection also required changes in the machine
file - essentially no forwarding to charmrun when redirecting.
--Nikhil

On Thu, Jan 9, 2014 at 2:28 AM, Miller, Philip B 
<mille121 AT illinois.edu>
 wrote:
> On Wed, Jan 8, 2014 at 2:55 PM, Jim Phillips 
> <jim AT ks.uiuc.edu>
>  wrote:
>>
>>
>> You can probably write a runscript to do this.  -Jim
>>
>>
> I don't think a wrapper script would be effective, at least not on a net*
> build, because the runtime internally redirects those FDs through a pipe
> that it then sends to charmrun. The problem is that when the program
> crashes, there can be unflushed data in that pipe that one will never see.
> In my case, I'm using AddressSanitizer and expecting output as the process
> dies, and I'm guaranteed not to see that reach charmrun.
>
>>
>> On Wed, 8 Jan 2014, Phil Miller wrote:
>>
>>> Before I go digging into the source code, is there an easy and
>>> already-implemented way of telling the runtime system to direct each
>>> process's standard output and standard error streams to a separate file?
>>>
>



-- 
"Success may ditch you most of the times, but if you ditch it once :
your job is done."
Nikhil Jain, 
nikhil.life AT gmail.com,
 +1-217-979-0918