Time to revert bad commits?
der.fabe at gmx.net
Tue Jun 3 06:01:11 CDT 2008
On Tuesday 03 June 2008 10:42:39 you wrote:
> Am Dienstag, 3. Juni 2008 10:02:26 schrieb Fabian Bieler:
> > The shader compiler can also do better optimization with hard-coded
> > constants (particularly if the constants are 0 or 1). I've seen the
> > nvidia-compiler remove entire code paths because of this in a hl2-shader.
> Really? Do you have any evidence for this? In that case we might want to
> use Vitaliys original patch which just changed the precision of the printed
> float, since the patches I just sent break this ability. Henri and I
> considered them as theoretical, since the HLSL compiler or shader author
> would have removed those codepaths already
I did some testing using the Half-Life 2 intro and outro (dxlevel 81).
Attached is a spreadsheet with the instruction and register counts of the
compiled shaders (aquired via __GL_WriteProgramObjectAssembly). The
difference is not as big as I expected (I stumbled over one of the more
significant cases (fasm_164_0.txt of the outro: 13 vs. 9 instructions) some
time ago and made a wrong extrapolation ;-)).
-------------- next part --------------
A non-text attachment was scrubbed...
Name: half-life shader.ods
Size: 18541 bytes
Desc: not available
Url : http://www.winehq.org/pipermail/wine-devel/attachments/20080603/fa2d1630/attachment-0001.ods
More information about the wine-devel