[prev in list] [next in list] [prev in thread] [next in thread] 

List:       mesa3d-dev
Subject:    Re: [Mesa3d-dev] get.c fix and a floating point question
From:       "Daniel Borca" <dborca () 3dfxzone ! it>
Date:       2004-12-30 7:56:26
Message-ID: 20041230075626.7084.qmail () webmail2 ! aruba ! it
[Download RAW message or body]

Hi, 

>  > On a side-note: I am having problems with gcc 3.3.x / gcc 3.4.x.
>  > It generates bad clipping in t_vb_render.c <- t_vb_cliptmp.h and 
> 
>  Any -march flags?  I hunted down a similar (same[0]?) problem when
>  compiling mesa with -mfpmath=sse.  At the end this worked well: 
> 
>  -O3 -march=i686 -msse -mfpmath=387 -DUSE_X86_ASM -DUSE_MMX_ASM
>  -DUSE_3DNOW_ASM -DUSE_SSE_ASM 
> 
>  Marcelo 
> 
>  [0] z-clipping was way off, either too far behind z_front or too far in
>      front of z_back.  The other planes had similar problems.

In early versions of gcc 3.x I tried -mcpu=athlon, but the result was
completely broken.  Since then, I stick to -mcpu=pentium.  I'm not a big
fan of -march, though - it breaks compatibility. 

I did try without -mcpu at all (this means pentium-pro, according to
`gcc -E -dM test.h`) without any success; yes, same clipping bugs. 

So basically, I have to either sanitize t_vb_cliptmp to use fi_type
instead of fi_type*, or use gcc 2.95.y.  I know that a particular flavor
of 2.95 has a very rare and odd bug, but I can't remember which one was. 

Regards,
Daniel Borca 



-------------------------------------------------------
The SF.Net email is sponsored by: Beat the post-holiday blues
Get a FREE limited edition SourceForge.net t-shirt from ThinkGeek.
It's fun and FREE -- well, almost....http://www.thinkgeek.com/sfshirt
_______________________________________________
Mesa3d-dev mailing list
Mesa3d-dev@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mesa3d-dev
[prev in list] [next in list] [prev in thread] [next in thread]