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

List:       gdb
Subject:    Re: Bad performance in updating JIT debug symbols
From:       Yichao Yu <yyc1992 () gmail ! com>
Date:       2016-09-18 13:54:54
Message-ID: CAMvDr+QiB6kq=0HWZo++nCiYfPEnvbAPeNjjxikc1iaKdZEUrA () mail ! gmail ! com
[Download RAW message or body]

On Sat, Sep 17, 2016 at 2:43 PM, Fredrik Tolf <fredrik@dolda2000.com> wrote:
> On Sat, 17 Sep 2016, Duane Ellis wrote:
>>
>> So why not make a wrapper on the symbol lookup function
>>
>> Search the main list if that fails to find try the next list of syms.
>>
>> After you under 100 inserts resort your additional list. And start a new
>> list
>
>
> I can imagine a lot of ways to optimize the current code, but the main point
> was that GDB seems to be stuck in a path of the code that is not really
> intended to be used other than when GDB is "confused", so the main question
> is why that happens.

Ref https://sourceware.org/ml/gdb/2016-03/msg00038.html

With some of my profiles and a patch that helps somewhat (there's
still bad scaling) but also cause functional regression.

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