[sldev] Optimization target: Avatar skinning, LLViewerJointMesh, matrix multiply

[James Cook]

Here's a little test app I've been using to time the matrix math
routines.  I'll post my SSE vector3 matrix4 multiply routine later
today.  Sadly, it's no faster than the compiler-generated non-SSE code.

I was thinking of taking the code gcc generates on the Mac and using it
as in-line assembly with Visual Studio for PC.

James

Callum Lerwick wrote:
> On Sat, 2007-04-14 at 16:43 -0700, James Cook wrote:
>> Yeah, it's those pesky 1 GHz Athlon Thunderbirds that I worry about. 
>> I'll dig into our CPU statistics and see if anyone still uses those with 
>> SL.  If not, I'll turn on SSE instructions in the builds.
> 
> I do.
> 
> The other advantage of autovectorization is you get support for PPC
> AltiVec and x86_64 and whatever else may come about for free. I wonder
> if spot has tried compiling slviewer on a SPARC yet...
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Click here to unsubscribe or manage your list subscription:
> /index.html

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 249 bytes
Desc: OpenPGP digital signature
Url : http://lists.secondlife.com/pipermail/sldev/attachments/20070415/001e0c67/signature.pgp