[prev in list] [next in list] [prev in thread] [next in thread] 

List:       gdb
Subject:    Re: JIT debugging (Attach and speed)
From:       Yichao Yu <yyc1992 () gmail ! com>
Date:       2016-03-23 20:51:19
Message-ID: CAMvDr+S1h46g=dxnKP5UBGrVnSfrf=S84ZmNtE4LU_D1SUKBXw () mail ! gmail ! com
[Download RAW message or body]

On Wed, Mar 23, 2016 at 3:48 PM, Pedro Alves <palves@redhat.com> wrote:
> On 03/23/2016 07:32 PM, Yichao Yu wrote:
> 
> > > So, as a quick hack, if you make inline_frame_sniffer bail out early
> > > with 0, does it improve things?
> > 
> 
> > The timing looks the same. =(
> 
> Bummer.  :-/
> 
> Having now looked at the gdb.base/jit.exp test, I see that
> we can easily reproduce this with:
> 
> gdb ---args ./testsuite/outputs/gdb.base/jit/jit-main \
> ./testsuite/outputs/gdb.base/jit/jit-solib.so 2000 
> (It'd be nice to turn this into a check-perf test...)

The new profile after disabling that line[1].

Am I reading it correctly that it skips the one I disabled and fails
to the next option which is basically as expensive?

Also, it looks like gdb is trying to get info about the current frame?
Can this be disabled for this breakpoint since it's not a "real" one?

[1] http://i.imgur.com/NHdoka8.jpg

> 
> > What do you mean by "loader"? since you mentioned call JIT_REGISTER I
> > assume you mean JIT compiler?
> 
> Yes, sorry.  I meant whatever code ends up "sending" a JIT_REGISTER
> event to gdb.
> 
> Thanks,
> Pedro Alves
> 


[prev in list] [next in list] [prev in thread] [next in thread]