couchdb-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Adam Kocoloski (JIRA)" <>
Subject [jira] Commented: (COUCHDB-544) Make CouchDB View server take use of GPU capabilities (// computing)
Date Tue, 27 Oct 2009 14:45:59 GMT


Adam Kocoloski commented on COUCHDB-544:

So, in my limited experience with GPGPU-style programming it really shines if you have a "small
data, big computation" problem.  Matrix manipulations, stuff like that.  CouchDB's view functions
are much heavier on the I/O side; I doubt there is much to be gained from shipping the data
out to the GPU and back.  But hey, I could certainly be wrong, don't let me discourage someone
who wants to give this a shot.

> Make CouchDB View server take use of GPU capabilities (// computing)
> --------------------------------------------------------------------
>                 Key: COUCHDB-544
>                 URL:
>             Project: CouchDB
>          Issue Type: New Feature
>          Components: Database Core, JavaScript View Server
>         Environment: All
>            Reporter: Maxime Garcia
>            Priority: Minor
> Nowadays, GPUs provide a considerable amount of power in parallel computing of shaders...
in screen rendering. For example, a standard GPU like ATI's HD 2900 XT gives 500 Gflops power
whereas a classic CPU gives only 8 Gflops.
> nVidia CUDA, OpenCL... it is underway.
> It is a computing mode that permits massive parallel computing on isolated data.
> o_O That sounds like docs in view building ?
> Yeah, exactly.
> The main idea is to make CouchDB View engine use these GPU (if available).
> Can you imagine the power of view building on a super CouchDB server with 4 PCI-Express
GPU cards ?
> Some infos (for example) :
> -
> (for FR :
> -
> (for FR :
> PS : This issue is submitted to make discussions happen on this crazy idea. Relax.
> I don't know how, but i think this could be a great improvement in the world of database,
and the Couch fits perfectly.
> What do you think about ?

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

View raw message