[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Libunwind-devel] problem with ms_abi on x86-64
From: |
Ariel Burton |
Subject: |
Re: [Libunwind-devel] problem with ms_abi on x86-64 |
Date: |
Mon, 20 Jun 2011 12:12:28 -0400 |
User-agent: |
Mutt/1.4.1i |
The larger footprint is certainly a reasonable
explanation for a slow down.
Are those figures for local processing (i.e.,
introsepection), or from an external process?
What do you mean by `apply the patch, but diable it
by default'? Does that mean that a compile time
switch or a runtime switch?
Ariel
==
On Sat 18 Jun 20:49, Arun Sharma wrote:
> On Sat, Jun 18, 2011 at 10:07 AM, Lassi Tuura <address@hidden> wrote:
> > Compared to current head, I see consistent slightly over 5% slow-down with
> > the patch
> > in our memory profiler, using the same test I used previously. The 387M
> > stack walks
> > in ~960 secs become about ~1011 seconds with the patch. I don't think there
> > are any
> > ms_abi frames.
>
> Thanks for testing. Probably the effect of a bigger dcache footprint.
> I'm inclined to apply the patch, but disable it by default.
>
> >
> > Under what conditions would ms_abi support / preserving SSE register state
> > be enabled?
> > The matches I found in google codesearch were basically cygwin, wine and
> > some EFI code.
>
> Ariel?
>
> -Arun