[hatari-devel] Regressions in 1.9.0 and current dev

[Anders Eriksson <ae@xxxxxx>]

Hello,

a few regressions have been found in 1.9.0 and the current 
development version.


1. The new demo by Ghost / Phalanx (yes, THAT Phalanx) released this 
weekend works flawlessly on 1.8.0 and on real hardware. But on 1.9.0 it 
would seem like the top border screws up now and then which in return 
takes the rest with it (it's a fullscreen). In the current dev version 
(from nov 3 that Bob compiled) the problem is the same, but much more 
severe. The dev-version uses the new WinUAE core.

Demozoo: http://demozoo.org/productions/148711/


2. In Hatari 1.9.0 When video recording the new demo by Spiceboys, the 
recording ends after the 3D-polygon screen (just before the Sinatra 
picture). The demo switches to medres at that time which could affect it. 
I got Hatari max size to 832*624 so the med/low thing shouldn't matter as 
the frames are the same size. In 1.8.0 this works fine. In the currect 
dev-version it seems to work as well. So it's a regression of 1.9.0 that 
is perhaps fixed now. Not sure if it's worth looking at anymore.

Demozoo: http://demozoo.org/productions/148713/


3. No regression, but a progression! Zorro / Ghost have a retina Macbook 
and the performance have been dreadful for Hatari. It's a quadcore i7 with 
Nvidia 650 (I think)  and it doesn't stand a chance to show Hatari without 
major frameskips. But now with Bobs SDL2 build, it runs fluid on the 
retina. Yehoo. There are many other issues with SDL2, but the performance 
is at least going the right way.


4. I noticed that the current build from Nov 3 is a lot slower than 1.9.0 
(and lower) to render a BMP AVI-stream. Perhaps it's just a matter of 
optimizations setting for the build.

 --
Anders Eriksson
ae@xxxxxx     http://www.dhs.nu/
ae@xxxxxxxxx  http://www.atari.org/