Just a short remark here (I'm just doing a quick mail scan from home): I
submitted my patch to golem 3 times before I landed it, and apart from our
usual heavy flakiness, there were no significant changes in the benchmark
numbers, as expected. How did you measure the 25%? Did you measure against
different baselines?

Regarding memory: Although I had to bump the cctest limits a little bit,
instrumenting the OS::foo allocation methods showed no significant change.
A little bit more memory usage is expected, of course, because we have more
maps with the change. I think it is at least a bit debatable what the /proc
FS actually tells us about memory usage. IMHO it is good as a sanity check
to catch huge unintended leaks etc., but I wouldn't worry too much about
minor changes.

On Tue, May 8, 2012 at 2:47 PM, <[email protected]> wrote:

> On 2012/05/07 10:52:13, Erik Corry wrote:
>
>> This change regressed the context create benchmark 25% and increased the
>>
> memory
>
>> use.  It would be nice to understand both issues.
>>
>
> Ouch, probably both related to what's serialized into the snapshot I guess.
> Unfortunately Sven is on leave and I'm unsure when he will return.
>
> https://chromiumcodereview.**appspot.com/10238005/<https://chromiumcodereview.appspot.com/10238005/>
>

-- 
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev

Reply via email to