Odd gfx glitches with sdl_hello_world once r330 is applied

[GoogleCodeExporter]

What steps will reproduce the problem?
1. build sdl_hello_world as-is, can see some issue when frame limiter is 
enabled but it is not as pronounced
2. Run on Dingoo

What is the expected output? What do you see instead?

On my LCD MODULE: LCM_FAIR_ILI9331_3, with rev 295, get solid display. No 
flicker. With r362 there is significant flicker for about 1 sec, and then after 
that see flicker/shimmer about once every 1 sec.

I took a quick scan at r330 and I'm out of my depth. There is clearly an 
improvement in speed with this change so I'm not suggesting it be removed.

It maybe that the sdl demo needs work rather than the driver code. However I 
just tried the demo on a Windows machine and do not see any problems:

{{{
gcc main.c SFont.c  -I C:\Dev-Cpp\include\SDL -lmingw32 -lSDLmain -lSDL
}}}

Original issue reported on code.google.com by clac...@gmail.com on 21 Aug 2011 at 6:00

[GoogleCodeExporter]

For a less academic example, OhBoy is unusable with headrev code :-( . I'm 
currently using r329 for building OhBoy.

Issues with OhBoy are scrolling (fast) through menu items ends up with 
significant ghosting and flashing once scrolling stops.

Playing games with no scaling has an old buffer visible.

Original comment by clac...@gmail.com on 21 Aug 2011 at 6:26

• Added labels: Priority-High
• Removed labels: Priority-Medium

[GoogleCodeExporter]

FYI I'm using r329 for performing builds now.

Original comment by clac...@gmail.com on 21 Aug 2011 at 6:28

[GoogleCodeExporter]

I've had a scan at the code and I can't comment too much without testing it so 
that I can guess what the issue is. One thing that jumps out at me is that it 
waits for DMA to complete after starting it up, this is wrong, you should only 
block waiting for the DMA to finish before using it again (would be faster and 
more stable).

Original comment by Flatmush@googlemail.com on 22 Aug 2011 at 9:25

[GoogleCodeExporter]

For the moment it would make most sense to revert this code back to it's 
original form and make an "Enhancement" ticket to get this faster code to 
behave. I'd like to say that I would get around to fixing it but I'm not really 
in a position to promise I'll have the time at the minute unfortunately.

Original comment by Flatmush@googlemail.com on 22 Aug 2011 at 5:58