Re: [PATCH] gdb script performance

[Jean-Marc Saffroy <saffroy@gmail.com>]

On Wed, 29 Nov 2006, Daniel Jacobowitz wrote:

> On Thu, Nov 30, 2006 at 01:33:02AM +0100, Jean-Marc Saffroy wrote:
>> It seems the patches I posted yesterday on the gdb list have gone
>> completely unnoticed, so I guess I should resend them here.
>
> This is the right list for patches.  However, please be patient - there
> is a chronic scarcity of reviewers, but we all do what we can.

Given your usual responsiveness on the lists, I thought my mail may have 
been filtered somehow. Thanks for bearing with me!

>> This is still unfinished (hooks for invalidating the caches are missing,
>> and I'm sure performance can still be enhanced significantly), but before
>> going further, I'd like to know how you feel about integrating such
>> changes.
>
> I'm glad to see some of this stuff sped up.  However, I'm hopeful that
> there's a better way to do it - shouldn't there be a more efficiently
> searchable data structure for whatever you're caching in the first
> place?
>
> Maybe there isn't; just thinking out loud.

I asked myself the same question, and hoped someone would have the answer 
here. ;-)

For the PC cache, I guess that just about any address can be passed, so I 
doubt there is much room for improvement (except maybe a cache with more 
than one entry, should it be useful in some cases, but now it's probably 
unneeded).

For the symbol cache, I think referencing all symbols in a global hash 
table could be a solution; actually I was surprised there was not one 
already. That's a more radical solution than my current patch, it could 
take a fair amount of memory, and also would take me much more time to 
implement, given that I'm still new to gdb internals. But I'd like this 
approach too, if it makes things cleaner and yields good performance.


-- 
saffroy@gmail.com