Loonies' "Hotstyle Takeover" breaks on 3.5.0

[Foebane]

http://www.pouet.net/prod.php?which=2306

This demo's always been a bit temperamental on earlier releases, but at least it always ran through to the end, but now it crashes about half-way through on 3.5.0, as well as issues with the opening animation (the woman's flying machine jumps around the screen, it shouldn't, and the beginning animation with the woman judders with JIT on). The crash happens during the tunnel sequence with rotating viewpoint after the woman animation.

Logs included.

EDIT: I have observed that the glitches (jumps) in the 3D animation in the beginning extends to the light in the brown tunnel and the corridors where the woman fights robots.

[Foebane]

Included here are the logs from 3.4.0, the version on which this demo works perfectly, for comparison.

[Toni Wilen]

Your config is too weird. JIT + cpu speed adjustments +cpu idle + vsync. This can result in version specific config.

Demo also has problems if CPU is too slow or too fast, quite common with many 68030+ demos.

[Foebane]

By "weird", do you mean the numeric keypad keys 0-3 for adjusting performance on the fly?

1. That is so I can run a wider range of demos.
2. I never had a problem with the demo in 3.4.0, something has obviously broken in 3.5.0.

In any case, I have tried various combinations of Vsync (do you mean Low Latency VSync?) and CPU idle, but the animated objects are definitely jumping around, like there are calculation errors, and it always crashes at the tunnel after the animation.

[Foebane]

Also:

1. My configuration may be weird to you, and I have to admit I don't fully know the intricacies of high-end Amiga emulation, but it has served me well and it has worked well with every single high-end demo I've thrown at it.
2. In any case, I've got 3.4.0 if I have further problems with 3.5.0. I can switch between fairly easily between them, even though they're being run from the same folder.

[amilo3438]

Yeah, in new WinUAE3600b4 I get cpu crash H10 after the tunnel part, on WinUAE3500 get only a guru at the same place, and no problem in WinUAE3400.

Tested in QuickStart A4000 w/68030 + 16MB RAM (motherboard), in fast mode and approximate +300 mode... in df0 attached WB3.1 and in HD attached as archive the demo file.

[Foebane]

Originally Posted by amilo3438

Yeah, in new WinUAE3600b4 I get cpu crash H10 after the tunnel part, on WinUAE3500 get only a guru at the same place, and no problem in WinUAE3400.

Yes, in 3.5.0 and on my 68040 config, I get a guru after the tunnel as well, but not on my 68060 config, which the demo recommends, the screen simply corrupts to a black screen with the occasional garbage.

EDIT: If it helps, the guru reports the following:

Error: 8000 000B Task: 07802228

[Toni Wilen]

Originally Posted by Foebane

Yes, in 3.5.0 and on my 68040 config, I get a guru after the tunnel as well, but not on my 68060 config, which the demo recommends, the screen simply corrupts to a black screen with the occasional garbage.

EDIT: If it helps, the guru reports the following:

I can't duplicate any crashes and I tried multiple configurations. Only demo glitches but they are normal.

EDIT: CPU libraries are 100% confirmed correctly installed and in use?

[Foebane]

Originally Posted by Toni Wilen

EDIT: CPU libraries are 100% confirmed correctly installed and in use?

How can I test this in WB? In any case, I've had no real problems in 3.4.0 with the 68060, although I have to turn JIT off because of horrendous juddering in the animated corridors. Also, I can still run the demo on 68040, so the 68060 issue is not really urgent.

However, the crash always happens in 3.5.0, even despite my trying various variations in the config, 040, 030 even (doesn't work properly there).

[Foebane]

Originally Posted by Toni Wilen

I can't duplicate any crashes and I tried multiple configurations.

Can you send me a hardware configuration that worked for you, Toni? So I can see if I can get the same results, that is.

[Toni Wilen]

Do you get same crashing behavior if you boot without startup-sequence, run setpatch and then the demo? Also try with "wrong" 68020+68882 config. (make sure "unimplemented cpu emu" IS ticked. This test makes possible cpu library issue irrelevant)

[Foebane]

Originally Posted by Toni Wilen

Do you get same crashing behavior if you boot without startup-sequence, run setpatch and then the demo? Also try with "wrong" 68020+68882 config. (make sure "unimplemented cpu emu" IS ticked. This test makes possible cpu library issue irrelevant)

I did all that with my 060 config (had to slow down the emulation with "Approx A500/A1200" so I was able to get the Early Boot Menu with both mouse buttons, then resumed at "Fastest possible" before selecting "boot without s-s". The demo ran the same as usual, but blacked out once again at the end of the tunnel section, with dark grey regular flashing.

It seems I already have "unimplemented cpu emu" ticked, but since JIT is also ticked, it's greyed out. I tried it without JIT, but the jerks in the 3D animation reappeared (they disappear with JIT and run just like on 3.4.0).

Would you like me to try without the unimplemented cpu emu?

[amilo3438]

Originally Posted by Toni Wilen

Also try with "wrong" 68020+68882 config. (make sure "unimplemented cpu emu" IS ticked. This test makes possible cpu library issue irrelevant)

In WinUAE3600b4 now get a guru and in WinUAE3400 still no problem!
(Tested w/o startup-sequence.)


@Foebane
In misc options tick "one second pause" to get the Early Boot Menu with JIT enabled!

[Foebane]

Originally Posted by amilo3438

@Foebane
In misc options tick "one second pause" to get the Early Boot Menu with JIT enabled!

Cool feature, I didn't expect that! Thx for the heads-up!

[StingRay]

I can confirm that there's a problem running the demo on 3.5! Demo works fine on 3.4, on 3.5 it crashes shortly after the "One frame effects are not possible" screen. I get a normal Amiga crash (guru 8000 000B so it's FPU related) though, WinUAE itself does not crash. Config used: 68030+68882 FPU, no JIT.

[amilo3438]

Now, someone should test WinUAE3500 betas to find in what beta version problem appears!

[Toni Wilen]

Originally Posted by StingRay

I can confirm that there's a problem running the demo on 3.5! Demo works fine on 3.4, on 3.5 it crashes shortly after the "One frame effects are not possible" screen. I get a normal Amiga crash (guru 8000 000B so it's FPU related) though, WinUAE itself does not crash. Config used: 68030+68882 FPU, no JIT.

Can you find out which instruction? (For example by setting f breakpoint to F-line vector address and then using H command when it breaks)

[amilo3438]

Originally Posted by amilo3438

Now, someone should test WinUAE3500 betas to find in what beta version problem appears!

WinUAE 3.4.1b1 (2017.02.11) - fine
WinUAE 3.4.1b2 (2017.02.25) - fine
WinUAE 3.4.1b3 (2017.03.13) - problem -> http://eab.abime.net/showpost.php?p=...7&postcount=29
WinUAE 3.4.1b4 (2017.03.25) - problem
WinUAE 3.4.1b5 (2017.04.02) - problem


PS. Btw. find test config I used for this demo in "The Zone"! (works w/o glitches)

[ross]

Originally Posted by StingRay View Post
I can confirm that there's a problem running the demo on 3.5! Demo works fine on 3.4, on 3.5 it crashes shortly after the "One frame effects are not possible" screen. I get a normal Amiga crash (guru 8000 000B so it's FPU related) though, WinUAE itself does not crash. Config used: 68030+68882 FPU, no JIT.

Originally Posted by Toni Wilen

Can you find out which instruction? (For example by setting f breakpoint to F-line vector address and then using H command when it breaks)

Interesting. For me is not a simple guru.
Crash is in the same position as Stingray, but with a 'freerun' CPU (in supervisor mode) in un-mapped memory and CHIP mem covered by a repeating 48byte pattern (on screen a static garbage image). Maybe Sting result is incidental?

[EDIT: tried a second time -> crash, 'dynamic' garbage on screen, CPU freerun, but this time CIA-OVERLAY reset so i cannot see the chip ram -> seem a totally random behaviour]

[StingRay]

Originally Posted by Toni Wilen

Can you find out which instruction? (For example by setting f breakpoint to F-line vector address and then using H command when it breaks)

The crash happens somewhere in this loop:

LOOP
    movem.w    (a0),d0-d3
    addq.l    #4,a0
lbC01BCC6
    move.w    #2,a6
    cmp.l    d1,d3
    beq.b    NEXT
    bgt.b    .noswap
    exg    d0,d2
    exg    d1,d3
    move.w    #1,a6
.noswap
    sub.l    d0,d2
    move.l    d3,d4
    asl.l    #8,d2
    sub.l    d1,d4
    asl.l    #8,d2
    addq.w    #8,d1
    addq.w    #8,d3
    ror.l    #4,d1
    lsr.w    #4,d3
    cmp.w    d1,d3
    ble.b    NEXT
    cmp.w    d3,a5
    bhs.b    lbC01BCF4
    move.w    d3,a5
lbC01BCF4
    fmove.l    d4,fp0
    fmove.l    fp1,d4
    fmove.l    d2,fp1
    movem.l    d0/d1/a6,(6,sp,d5.l)
    neg.l    d5
    fdiv.x    fp0,fp1
    movem.l    (6,sp,d5.l),d0/d1/d3
    move.l    (a1,d1.w*4),a6
    move.l    d1,d2
    move.w    a3,(a6)+
    asl.l    #8,d0
    move.w    d3,(a6)+
    clr.w    d2
    moveq    #$10,d3
    rol.l    #4,d2
    asl.l    #8,d0
    sub.l    d2,d3
    muls.l    d4,d3
    add.l    d3,d0
    add.l    #$80000,d0
    asr.l    #4,d0
    move.l    d0,(a6)+
    move.l    d4,(a6)+
    move.l    a6,(a1,d1.w*4)
    move.w    d6,a3
NEXT
    dbra    d7,LOOP

Is it possible to set breakpoints on register conditions? So that I could f.e. trigger a breakpoint when loopcounter (d7) contains a certain value.


Edit: It sure is. found "fo" command.