[hatari-devel] Little trace question :

Nicolas Pomarède npomarede at corp.free.fr
Tue Apr 12 22:54:24 CEST 2011


Le 12/04/2011 22:46, Laurent Sallafranque a écrit :
> Hi,
>
> I've done a trace of Lotus09 game.
> It loops forverer at startup.
>
> I can see a strange trace. (This appears many times, always at same HBL
> line).
>
>
> I just want to be sure I read correctly the trace:
>
> cpu video_cyc=133568 472 at 262 : $e00d60 : b0b9 0000 0466 cmp.l $466,d0
> cpu video_cyc=133580 484 at 262 : $e00d66 : 67f8 beq.s $e00d60
> cpu video_cyc=133584 488 at 262 : $e00d60 : b0b9 0000 0466 cmp.l $466,d0
> cpu video_cyc=133596 500 at 262 : $e00d66 : 67f8 beq.s $e00d60
> cpu video_cyc=133600 504 at 262 : $e00d60 : b0b9 0000 0466 cmp.l $466,d0
> cpu video_cyc=133612 133104 at 264 : $e00d66 : 67f8 beq.s $e00d60
> cpu video_cyc=133616 133108 at 264 : $e00d60 : b0b9 0000 0466 cmp.l $466,d0
> cpu video_cyc=133628 133120 at 264 : $e00d66 : 67f8 beq.s $e00d60
> cpu video_cyc=133632 133124 at 264 : $e00d60 : b0b9 0000 0466 cmp.l $466,d0
>
>
> If I understand it well, it means that at video cycle 133600, I'm in HBL
> 262, my instruction should be 12 cycles.
> But next instruction is executed (133104 - 504) cycles later ? (2 HBL
> later ?)
>
> If this is true, I think I've got a problem with something like
> do_specialties.
> May this be an artefact of video.c (something that happends at HBL 262,
> cycle 504) ?
>
> Regards
>
> Laurent
>
>

You only included cpu traces, what are the video traces ? Certainly a 
VBL might add some extra cycles between the 2 instructions.

I think this game is running at 60 Hz, and what you see is due to the 
VBL starting at hbl 263 and video cycle not yet reset to 0.

There's the same "problem" in 50 Hz on hbl 313 (just run --trace cpu_all 
while just booting to gem desktop), cycles on the last line are not 
correct in the trace, but in all cases this never was a problem and 
should not cause a bug (never see it until there).


Your problem might related to sthg else.



More information about the hatari-devel mailing list