Re: [PROF RFC 1/2] profiling of libcurl by curl-loader, steady state
Date: Wed, 18 Apr 2007 22:48:25 +0200 (CEST)
On Wed, 18 Apr 2007, Robert Iakobashvili wrote:
>> I would rather recommend you using CURLOPT_ERRORBUFFER instead, as it would
>> then also reduce the amount of *printf()s etc within libcurl.
> Thank you for your advise. Indeed, this is a valid option.
> Still, we need libcurl verbose debugging option at least for the detailed
> statistics and counters, that we are generating, and even more fore future.
Right, but then you'll sacrifize transfer performance to some degree.
> To the best of my knowledge CURLOPT_ERRORBUFFER requires a buffer per
> CURL-handle of size 256 bytes. Planning dealing with tens and hundreds
> thousand of virtual clients - this is the memory worth to save.
That's a valid remark but you instead get a slower operation. A trade-off of
course. But you're probably already using around 60-80K per client, so adding
256 bytes to that should make a very big difference...
I don't like the debug callback being used for that purpose so I'm not very
keen at applying that patch to the main sources.
-- Commercial curl and libcurl Technical Support: http://haxx.se/curl.htmlReceived on 2007-04-18