Re: GDB / SIM 7.3.1 Cosmetic patch for profile title

[Mike Frysinger <vapier@gentoo.org>]

[-- Attachment #1: Type: Text/Plain, Size: 1458 bytes --]

On Thursday 20 October 2011 11:59:18 Joel Brobecker wrote:
> > since profile_printf() takes a cpu arg, this fails to build once the
> > header printf is pulled out from the loop where "cpu" is declared.
> 
> Ah, correct. A quick look at that function reveals that it only needs
> the "cpu" to get access to the associaed PROFILE_FILE. Not sure what
> this is and why it is necessary, rather than doing a simple printf.
> 
> In any case, sounded like a good idea, but maybe it wasn't. I'll leave
> it to you guys if you want to pursue it.

the sim profile code has a --profile-file flag so that you can send the profile 
output to a dedicated file.  this is really useful when you're trying to pick 
apart all the other things the sim may output -- tracing, output from the 
simulated program itself, etc...  when you've got like 4+ different sources of 
information sent to one or two fd's, you can see how dedicated file flags are a 
hard necessity :).

fundamentally, the profile code operates on a per-cpu basis which makes sense 
in that the execution/profile of each cpu looks very different.  and the profile 
output code follows this convention.

atm, in practice, all output is sent to the same file handle, and most (all?) 
sims i know of only support one cpu.  so it's a bit of a wash, but i think we 
should stick to supporting the multi-cpu concept everywhere so that future 
work is much easier.

</braindump>
-mike

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 836 bytes --]