Re: [hatari-devel] new acia / ikbd emulation

[Eero Tamminen <oak@xxxxxxxxxxxxxx>]

Hi,

On perjantai 02 marraskuu 2012, Nicolas Pomarède wrote:
> > I added to profiler a new mode which will show addresses/instructions
> > that were used during profiling, in the address order (i.e. not sorted
> > by instruction/cycle count).  Results are attached, I hope they tell
> > what's wrong.
> 
> can't really see from the logs, I think it would be much easier to
> directly look at the Emutos code in C (if it's the case for this part)
> and see what logic was used to handle the keyboard. When source is
> available, it's often much better than using the disassembly output.

Sorry, I forgot to load EmuTOS symbols to the debugger,
latest EmuTOS 512k snapshot releases have them in Hatari
compatible format.

Attached output has also EmuTOS function names, from
that one can see which EmuTOS functions are still working
while the game appears frozen.

Based on how many times the first instructions of the functions
are called, int_acia() gets called a lot more than any other
EmuTOS function.

The code in question is from assembly file bios/aciavecs.S
which has quite a bit of ifdefs, so profiler output might
actually be easier to parse although it's lacking the local
symbols used in the assembly code:
http://emutos.cvs.sourceforge.net/viewvc/emutos/emutos/bios/aciavecs.S?revision=HEAD&view=markup

Comments in that file indicate that EmuTOS implementation
differs quite a lot from normal TOS...


	- Eero

Attachment: fokker.txt.gz
Description: GNU Zip compressed data