[Bf-committers] optimising the code

Arne Schmitz bf-committers@blender.org
Sun, 30 May 2004 18:49:02 +0200


--Boundary-02=_CChuApe0wiIddGB
Content-Type: text/plain;
  charset="iso-8859-15"
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline

Am Samstag, 29. Mai 2004 18:34 schrieb Jean-Luc Peuri=E8re:
> Blender structures use floats, so you must do float math
> when possible, which was the case before 2000 but is
> no more. Or switch all structures to use only doubles.
>
> Besides, SSE and Altivec code is float not double so if we
> either use those, that's the right road.

Plus double means also double memory and bandwidth requirements. And this c=
an=20
mean *a lot* for large scenes, or many operations. Rendering a scene with ~=
1=20
Mio vertices takes up twice as much in memory for the scene description whe=
n=20
using doubles and also needs twice as much bandwidth for loading and storin=
g=20
of the FPU registers.

Arne

=2D-=20
[--- PGP key FD05BED7 --- http://www.root42.de/ ---]

--Boundary-02=_CChuApe0wiIddGB
Content-Type: application/pgp-signature
Content-Description: signature

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)

iD8DBQBAuhCCzpP3gv0FvtcRArTYAJ9H4l3YPyVb/gk18kelLoLyeAPcLgCgmKaP
Ob1dvVwogM2aSw4kIxQYg+E=
=YJMn
-----END PGP SIGNATURE-----

--Boundary-02=_CChuApe0wiIddGB--