I was just messing about in the software accel code implementing ALUT44
support. Sorry Dok couldn't resist ;) I've noticed a few small issues
unrelated to ALUT44.

1.
What's the idea behind the current implementation of Sop_lut8_to_Dacc()?
It processes 8 pixels per each loop iteration. On my system however I get
much better performance by processing only one pixel per iteration.
df_dok's fill blend benchmarks show a considerable speed imrovement
(fill_rect_blend 1.90 vs. 0.60 MPixels/s). Draw blend benchmarks also
benefit though not as much.

2.
Shouldn't Bop_PFI_SKto_Aop and Bop_PFI_Sto_Aop have _PFI suffix like
the non-stretch versions?


I forgot the rest already :)

-- 
Ville Syrj�l�
[EMAIL PROTECTED]
http://www.sci.fi/~syrjala/


--
Info:  To unsubscribe send a mail to [EMAIL PROTECTED] with
"unsubscribe directfb-dev" as subject.

Reply via email to