|
Tuning model-2 tgp makes vf2 and dynamcop works, but frozen/crashes daytona...
#358069 - 08/28/16 09:51 AM Attachment: 0002.png 228 KB (0 downloads)
|
|
|
While analyzing some model-2/2a non-working games, it seems some timing issues between copro tgp and i960 cpu prevented vf2 and dynamcop from running.
Tried tuning up tgp copro's clock to some particular value, vf2 and dynamcop will run much further.
However, daytona seems very "tgp clock sensitive". Even 2MHz overclock will either frozen the whole mame emulation, or crash it "segment fault" if running in debug mode.
If carefully running debug mode with single vblank, it can be caught at about 160+ frames, happens just when 3D demo starts.
Another, the frozen only happens with daytona, other clones such as daytona93, daytonase are working well.
[ATTACHED IMAGE - CLICK FOR FULL SIZE]
|
|
|
|
Re: Tuning model-2 tgp makes vf2 and dynamcop works, but frozen/crashes daytona...
[Re: Naibo]
#358070 - 08/28/16 09:52 AM Attachment: 0000.png 70 KB (0 downloads)
|
|
|
dynamcop
[ATTACHED IMAGE]
|
|
|
|
Another, is there any known issues of i960/sharc emulation so far? I'm also working on the color-lighting of model-2
[Re: Naibo]
#358071 - 08/28/16 10:14 AM Attachment: xlat.png 11 KB (0 downloads)
|
|
|
Attatched is the "xlat" table taken from manxtt. The 1st pass the CPU writes near perfect curve, see the green track (using some float-point square-root algorithm I think, as some of i960 opcodes are unable to dis-assemble so far);
However, the CPU writes it the 2nd pass, this time all odds units are changed to a weired shape(see yellow track).
Any idea?
[ATTACHED IMAGE - CLICK FOR FULL SIZE]
|
|
|
|
Re: Tuning model-2 tgp makes vf2 and dynamcop works, but frozen/crashes daytona...
[Re: Naibo]
#358089 - 08/28/16 08:27 PM
|
|
|
Good to have you back, Naibo!
|
|
|
R. Belmont |
Cuckoo for IGAvania
|
|
|
Reged: 09/21/03
|
Posts: 9716
|
Loc: ECV-197 The Orville
|
|
Send PM
|
|
|
Re: Another, is there any known issues of i960/sharc emulation so far? I'm also working on the color-lighting of model-2
[Re: Naibo]
#360852 - 12/02/16 05:52 PM
|
|
|
> Attatched is the "xlat" table taken from manxtt. > The 1st pass the CPU writes near perfect curve, see the green track (using some > float-point square-root algorithm I think, as some of i960 opcodes are unable to > dis-assemble so far); > > However, the CPU writes it the 2nd pass, this time all odds units are changed to a > weired shape(see yellow track). > > Any idea?
There are definitely i960 instruction bugs. SHARC is fairly well trusted because a lot of Konami games use it for transform and lighting, but probably not bulletproof.
|
|
|