Re: [web-devel] new field for Network.HTTP.Types.Status
Hello,
What if you try just the bang pattern and no unboxing?
Here it is: bang for three fields: 41346.8 req/s bang and unpack for Int only: 41761.7 req/s So, the current merged code is the best. --Kazu
Hi,
could you share the scripts used for benchmarking? I'd like to
reproduce these numbers to better grasp the meaning of the changes.
Thanks!
L.
On Thu, Aug 30, 2012 at 4:16 AM, Kazu Yamamoto
Hello,
What if you try just the bang pattern and no unboxing?
Here it is: bang for three fields: 41346.8 req/s bang and unpack for Int only: 41761.7 req/s
So, the current merged code is the best.
--Kazu
_______________________________________________ web-devel mailing list web-devel@haskell.org http://www.haskell.org/mailman/listinfo/web-devel
Hi,
could you share the scripts used for benchmarking? I'd like to reproduce these numbers to better grasp the meaning of the changes.
I use this one. httperf --hog --num-conns 1000 --num-calls 1000 --rate 1000000 --burst-length 20 --server localhost --port 8000 --uri / Note that each value of parameter is carefully chosen based on my experience. --Kazu
Thanks. This is the "client-side" part. What about the "server-side"
part, i.e. the haskell code?
Cheers,
L
On Fri, Aug 31, 2012 at 12:59 PM, Kazu Yamamoto
Hi,
could you share the scripts used for benchmarking? I'd like to reproduce these numbers to better grasp the meaning of the changes.
I use this one.
httperf --hog --num-conns 1000 --num-calls 1000 --rate 1000000 --burst-length 20 --server localhost --port 8000 --uri /
Note that each value of parameter is carefully chosen based on my experience.
--Kazu
_______________________________________________ web-devel mailing list web-devel@haskell.org http://www.haskell.org/mailman/listinfo/web-devel
Thanks. This is the "client-side" part. What about the "server-side" part, i.e. the haskell code?
I'm using mighttpd2. But I did not release new libraries which includes MANY optimizations. So, I guess that you cannot gain the performace which I showed. I believe that http-types optimization is effective only with new warp whose HTTP response composer is completely re-written. https://github.com/kazu-yamamoto/wai/tree/refactoring/warp --Kazu
participants (2)
-
Kazu Yamamoto -
Lorenzo Bolla