WinUAE 3.0.0 beta series (was 2.8.2/2.9.0)

[Toni Wilen]

Please, step by step instruction. "Boot with config <config file>, change option a and b using GUI, speed is weird and it seems to hang" or similar.

[amilo3438]

When using BPPC config in latest Beta 8 test zip in SysInfo got allocated memory space 15MB on $68xxxxxx and 16MB on $08xxxxxx ?! (w/o and also when using MapROM)
(seems it happens from beta 5 but didnt it noticed till now)

p.s. Btw. The easiest way to enter into BPPC/CSMK3/CSPPC boot menu is to watch power led and in moment when it lit quick press/release Esc key and then keep depressed until boot menu shows.

EDIT: Seems there is still problem with BPPC saving changes after changes in boot menu (use, save) ?!
Manage to do it once today but now, even when renaming new flash, got corrupted unworking flash ?! (WinUAE beta 8 test zip)

[vagrant]

Quickest way to reproduce:
- load A4000/030 quickstart config
- load workbench
- enable CE in GUI
- run sysinfo, run speed test

previous releases: test completes with a brief hang reporting CPU speed of 29.. normal behavior (see pic 1)

beta 8: test freezes during "computing mflops" for a huge amount of time then reports abnormal speed of 16.. or freezes indefinitely. Both times I tried this with quickstart it froze forever (see pic 2)

[Snake79]

Originally Posted by ED-209

Quickest way to reproduce:
- load A4000/030 quickstart config
- load workbench
- enable CE in GUI
- run sysinfo, run speed test

previous releases: test completes with a brief hang reporting CPU speed of 29.. normal behavior (see pic 1)

beta 8: test freezes during "computing mflops" for a huge amount of time then reports abnormal speed of 16.. or freezes indefinitely. Both times I tried this with quickstart it froze forever (see pic 2)

Confirmed, i have the same issue and my hardfile .hdf is booting veeery slow with this config also...
EDIT: This is only happening when using "Fastest possible" option with CE.

[SaphirJD]

No problems here with such a quickstart config and jump from fastest possible to Cycle Exact - using this latest Winuae Zip File here http://www.winuae.net/files/b/winuae.zip and one of the ClassicWBP96 hardfiles in combination with OS31

[Snake79]

Originally Posted by SaphirJD

No problems here with such a quickstart config and jump from fastest possible to Cycle Exact - using this latest Winuae Zip File here http://www.winuae.net/files/b/winuae.zip and one of the ClassicWBP96 hardfiles in combination with OS31

After using upper linked winuae sadly got the same... remember that this is only happening when using Fastest possible AND Cycle exact...

[Toni Wilen]

Originally Posted by amilo3438

When using BPPC config in latest Beta 8 test zip in SysInfo got allocated memory space 15MB on $68xxxxxx and 16MB on $08xxxxxx ?! (w/o and also when using MapROM)
(seems it happens from beta 5 but didnt it noticed till now)

Fixed.

EDIT: Seems there is still problem with BPPC saving changes after changes in boot menu (use, save) ?!
Manage to do it once today but now, even when renaming new flash, got corrupted unworking flash ?! (WinUAE beta 8 test zip)

Can't duplicate. Works fine here, multiple times changed.

Originally Posted by ED-209

Quickest way to reproduce:
- load A4000/030 quickstart config
- load workbench
- enable CE in GUI
- run sysinfo, run speed test

previous releases: test completes with a brief hang reporting CPU speed of 29.. normal behavior (see pic 1)

Hmm.. But in previous releases fastest possible + cycle exact didn't even work and I am quite sure you have this option selected, right?

beta 8: test freezes during "computing mflops" for a huge amount of time then reports abnormal speed of 16.. or freezes indefinitely. Both times I tried this with quickstart it froze forever (see pic 2)

This is fixed (which only happened with fastest possible + cycle exact). It can still cause slowdown and high CPU% during tight loops (like speed tests) but it won't hang anymore.

http://www.winuae.net/files/b/winuae.zip

[SaphirJD]

Ok, after finally being able to get that hang/freeze with the Cloanto OS3X wb files too, this new Zip solves the problem

[amilo3438]

Originally Posted by Toni Wilen

Can't duplicate. Works fine here, multiple times changed.

BPPC -> still after boot menu update got unusable file?! (attached below, unzip and rename it!)

(on a new 256k file just enter into boot menu and click on use, save - dont change anything - change to MapROM only and click on use, save)
(p.s. On 1st case got working 512 file and on 2nd got unusable file!)

[vagrant]

Originally Posted by Toni Wilen

Hmm.. But in previous releases fastest possible + cycle exact didn't even work and I am quite sure you have this option selected, right?

Hmm no, fastest possible wasn't selected during any tests.
When I enable CE via the GUI, if 'fastest possible' was set it automatically switches to 'approx a500/a1200 cycle exact speed'.
Maybe fastest possible was still on and GUI was displaying incorrectly?

That winuae.zip has fixed it though, no hang & speed back up to 29

[vagrant]

Doh! That Blood Money sound problem is back with the new winuae.zip.. (sorry tony )

beta 8 -the game works 100% every time
winuae.zip - title music is always half speed and small gfx glitch briefly on the ship during intro (see pic).

[Toni Wilen]

Originally Posted by ED-209

Doh! That Blood Money sound problem is back with the new winuae.zip.. (sorry tony )

beta 8 -the game works 100% every time
winuae.zip - title music is always half speed and small gfx glitch briefly on the ship during intro (see pic).

I don't think I really care, it is 68020+ cycle exact which isn't accurate enough and works accidentally only and I also don't really like using whdload games for any kind of compatibiity testing because for obvious reasons slave programmer can't test it with all kinds of configurations.

[Toni Wilen]

Second PPC status report:

PearPC CPU core does work. Instructions are executed, it does some poking to board's IO addresses but unfortunately I am not sure how addresses F60030 and F60038 work.

Only available information comes from netbsd and linux sources:

/* IPL_EMU 0x30 */
#define P5_DISABLE_INT 0x40 // if set: all CPU interrupts disabled
#define P5_M68K_IPL2 0x20
#define P5_M68K_IPL1 0x10
#define P5_M68K_IPL0 0x08
#define P5_PPC_IPL2 0x04
#define P5_PPC_IPL1 0x02
#define P5_PPC_IPL0 0x01
#define P5_IPL_MASK 0x07
/* INT_LVL 0x38 */
#define P5_LVL7 0x40
#define P5_LVL6 0x20
#define P5_LVL5 0x10
#define P5_LVL4 0x08
#define P5_LVL3 0x04
#define P5_LVL2 0x02
#define P5_LVL1 0x01

Something to do with interrupts but are they input or output or both? Mainboard inputs? Does writing cause interrupts? What?

But at least BPPC now boots to amigados and PPC CPU is stopped quite soon after it has been started.

Memory is currently directly accessed from PPC thread and all IO addresses are put to queue that m68k side handles and replies. Not very optimal but only way to make it work without extra complexity.

[amilo3438]

Originally Posted by Toni Wilen

Second PPC status report:

PearPC CPU core does work. Instructions are executed, it does some poking to board's IO addresses but unfortunately I am not sure how addresses F60030 and F60038 work.

To see if the PPC is alive you could run the PPCTest floppy -> http://powerup.amigaworld.de/index.php?lang=en&page=24

If you get two "OK"'s (at least on CSPPC) then your PPC is alive.

EDIT: Note: But WarpOS could be problematic to test on BPPC because of ... just reminds.

[Toni Wilen]

Originally Posted by amilo3438

To see if the PPC is alive you could run the PPCTest floppy -> http://powerup.amigaworld.de/index.php?lang=en&page=24

I know. Not much help when M68K and PPC can't communicate properly due to lack of correct interrupt handling. (and I am NOT going to disassemble any PPC code)

[amilo3438]

Originally Posted by Toni Wilen

(and I am NOT going to disassemble any PPC code)

But could disassemble CSPCC you have, right ?! ... piece by piece and compare.

(p.s. Didnt you already mentioned somewhere you still have one?)

Nobody said it would be easy, but it would be a real shame to give up now.

On some sites it already seen as a done deal -> http://www.indieretronews.com/2014/0...pc-update.html

In any case, you'll know the best is it worth the effort/trouble.

[mark_k]

Originally Posted by Toni Wilen

PearPC CPU core does work. Instructions are executed, it does some poking to board's IO addresses but unfortunately I am not sure how addresses F60030 and F60038 work.

Only available information comes from netbsd and linux sources:
...
Something to do with interrupts but are they input or output or both? Mainboard inputs? Does writing cause interrupts? What?

Does the file src/sys/arch/amigappc/amigappc/pic_amiga.c in the NetBSD source give any clues?
Also src/sys/arch/amigappc/amigappc/locore.S
and maybe src/sys/arch/amigappc/amigappc/machdep.c

[Toni Wilen]

It does not tell if PPC can cause m68k interrupt by writing to those registers or if it can be only used to read m68k interrupt level and to set virtual PPC interrupt level.

Netbsd/Linux does not need to care about m68k after switching to PPC.

EDIT: also function of "P5_ENABLE_IPL" is not that clear..

[amilo3438]

Originally Posted by Toni Wilen

I don't think I really care, it is 68020+ cycle exact which isn't accurate enough and works accidentally only

Does that mean you're also giving up from improved/better A1200 CE emulation? (hopefully not)

[Toni Wilen]

Originally Posted by amilo3438

Does that mean you're also giving up from improved/better A1200 CE emulation? (hopefully not)

No but there has not been any compatibility improvements in this beta, nothing has really happened except support for "fastest possible cycle exact" mode. Which obviously has nothing to do with compatibility