On Thu, 22 Sep 2011, vineeth wrote:

> 1:
> - Do a multi-perform in a single thread.

> 2:
> - Do a easy_perform in each spawned thread.

> Now, my question is: I believe that case-1 will not make full utilization
> of the bandwidth and case-2 will be faster.

Why? And shouldn't you pretty much saturate whatever bottle-neck you have
using either method?

> Can others pls share their views on the two approaches and which is the
> better one. Any pointers will be appreciated.

It is very rarely a question of which is best. It is more a matter of you can
do it either way, use the way you think works best with your application or
with how you vision it should work.

If you really think one is faster than the other, then I think you should
write up two small test applications and run a comparison of them against each
other and pick the winner!

There are of course implementation details such as shared caches and pools
that you may consider too.

-- 
  / daniel.haxx.se
-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette:  http://curl.haxx.se/mail/etiquette.html