WinUAE 4.9.2+ beta series

[Toni Wilen]

Originally Posted by jotd

I remember you fixing issues with code wrongly detected as SMC even when caches were flushed. I don't know exactly in which version it was but it's pretty recent.

Beta thread requirement is to find the version that had the difference. Otherwise the usual response is "works for me" or something as useless.

I think I know what you mean but you really shouldn't use RAM above 16M with SMCD or if you really need more RAM, use mainboard RAM space ($08000000) which is much lower in address space then Z3 RAM.

Next beta has different logic that does not require whole buffer clear when cache is flushed which should help.

Originally Posted by Ponki1986

I made speed measurements using SortBench. I hope this program is more reliable than SysInfo.

Could you check also AIBB? Does all tests show similar reduction in performance or only some?

Was that JIT enabled or not? Does both show similar difference?

I can't see anything obvious in b5 changes (interrupt was the most obvious but it can't explain it because sysinfo tests are running with interrupts disabled). One possibility is compiler update or something similar. I also noticed odd behavior in sysinfo, sometimes tests results can be -10/+10%, rerunning the tests always returns mostly same result but shutting down winuae and restarting sometimes can return different results.. Possibly reason might be winuae main thread running on different core with different max boost limit? (Ryzen 7950X)

[Ponki1986]

Originally Posted by Toni Wilen

Can't duplicate. Probably need config and log.
EDIT: of course I managed to duplicate it after posting this..

winuae.7z fixed.



I'll upload test version tomorrow.

Can I download now?
Same link: https://download.abime.net/winuae/files//b/winuae.7z ?

[Toni Wilen]

Originally Posted by freddy

Toni, can you activate the string for translation ?

Done

Originally Posted by Ponki1986

Can I download now?
Same link: https://download.abime.net/winuae/files//b/winuae.7z ?

Can do now. (Normally winuae.7z is not fully optimized build, full optimized uploaded about 10 minutes ago)

[rutra80]

It's not necessarily beta related, but since you discuss speed variations, there are wild speed swings in Full-windowed mode on my system. Exclusive Fullscreen has rock solid performance.

[Ponki1986]

Originally Posted by Toni Wilen

Done



Can do now. (Normally winuae.7z is not fully optimized build, full optimized uploaded about 10 minutes ago)

Now Overlay using Direct3D 9 is good.

Other things like PiP, MIDI playback, FloppyBridge, Direct3D 11 work fine.

The AIBB gave a score of 3888959 Dhrystones. Version 4.9.1 scored 449531 Dhrystones

[vulture]

Hi Toni,

DamienD has tested all 51 x games in his collection that have JIT enabled.

These are now no longer working and freeze with beta 8:

... Cave Story [RTG] [Windows conversion]
... Command & Conquer - Tiberian Sun & The Covert Operations Mission Pack [RTG] [Vanilla Conquer]
... Curse of Monkey Island, The [AGA] [ScummVM]
... Day of the Tentacle [AGA] [ScummVM]
... Descent - Freespace - The Great War [AmigaCD] [RTG]
... Descent [RTG] [D1X_Rebirth]
... Dig, The [AGA] [ScummVM]
... Feeble Files, The [AmigaCD] [RTG]
... Full Throttle [AGA] [ScummVM]
... Griffon Legend, The [RTG] [Windows conversion]
... Heart of Darkness [RTG] [Windows conversion]
... Heretic & Hexen [AmigaCD] [RTG] [MS-DOS conversion]
... Hydra Castle Labyrinth [RTG] [3DS conversion]
... Lands of Lore - The Throne of Chaos [AGA] [ScummVM]
... Legend of Kyrandia, The - Book 2 - The Hand of Fate [AGA] [ScummVM]
... Legend of Kyrandia, The - Book 3 - Malcom's Revenge [AGA] [ScummVM]
... Legend of Zelda, The - Navi's Quest [RTG] [Windows conversion]
... Legend of Zelda, The - Oni Link Begins [RTG] [Windows conversion]
... Legend of Zelda, The - Return of the Hylian (Enhanced) [RTG] [Windows conversion]
... Legend of Zelda, The - Return of the Hylian [RTG] [Windows conversion]
... Legend of Zelda, The - Time to Triumph [RTG] [Windows conversion]
... Mini Slug Project [RTG] [Arcade conversion]
... Retrobattle [RTG] [Windows conversion]
... Sam & Max Hit the Road [AGA] [ScummVM]
... Tyrian [RTG] [OpenTyrian]
... Ultima VII - The Black Gate + The Forge of Virtue & Serpent Isle + The Silver Seed [RTG] [Exult]

If you require configs, HDFs, logs or anything else then please let me know.

[Toni Wilen]

Originally Posted by vulture

Hi Toni,

DamienD has tested all 51 x games in his collection that have JIT enabled.

These are now no longer working and freeze with beta 8:

More information is needed. Which beta? Is it reproducible? JIT related problems usually depend on also host CPU speed which can make debugging it difficult without knowing which version work and which don't and if there are anything interesting in the logs. (and weird config is not allowed, especially modified CPU speed adjustments)

I don't have any of above but it also should not be too big. (PC ports are generally annoying to debug)

[vulture]

Hi Toni,

Beta8 is when the issue starts after the recent JIT changes you made. All games are fine in previous betas.

Definitely reproducible every time. Totally understand that PC ports are annoying to debug.

I've uploaded the following configurations / gamedirs for you:

... Descent - Freespace - The Great War [AmigaCD] [RTG]
... Feeble Files, The [AmigaCD] [RTG]
... Sam & Max Hit the Road [AGA] [ScummVM]

https://mega.nz/file/JqAQzTJS#Ol-j96...IV7c3qlaApuW58

Many thanks for your time, appreciated.

[Ponki1986]

Originally Posted by vulture

Hi Toni,

DamienD has tested all 51 x games in his collection that have JIT enabled...

... Descent
... Heart of Darkness
... Heretic & Hexen
... NBlood
... Mini Slug Project
... JFKDuke3D
... AmiQuake 2

I too have checked a few of the games in the list above and confirm that they work worse. Up to version 4.9.9 Beta 7 dated 26.11.2022 the games work fine. Later versions of the Beta make the sound in games seem intermittent, jerky and you can also hear a strange echo. It is then clear that there is not enough power to make the game run smoothly. Of course, someone can always say that if something runs slowly it is the hardware's fault. But up until Beta 7 of 26.11.2022 everything worked fine on my hardware. On version 4.9.1 all games ran even too fast. I am attaching the log and configuration files. Maybe they will be useful.

Logs...

[paraj]

SMC detector changes look very good , and this may already be fixed, but you've left in some debugging code (line 3364 in debug.cpp)

if (addr < 0x100 || addr > 0x200)

that makes it not work in the current preview version.
Not beta related, but spotted an issue a bit further down in the same function:

		if (regs.irc == 0x4e75 || regs.irc == 4e74 || regs.irc == 0x4e72 || regs.irc == 0x4e77)
			return; /* RTS, RTD, RTE, RTR */

Should check against 0x4e74 and 0x4e73 for the middle two ones.

And Test/x64 gives an asan hit in romlist_cleanup due to i ending up reading past the end of roms, changing the condition to

while (i < sizeof(roms)/sizeof(*roms) && roms[i].name)

fixes it.

[Toni Wilen]

Originally Posted by vulture

Beta8 is when the issue starts after the recent JIT changes you made. All games are fine in previous betas.

Fixed. Hopefully without no other side-effects. JIT requirements are difficult to keep when everything else changes around it..

past the end of roms, changing the condition to

while (i < sizeof(roms)/sizeof(*roms) && roms[i].name)

fixes it.

Thanks.

(i < j && roms[i].name) should be enough ("i < j" is needed or most roms get marked as not complete)

[vulture]

Now everything works fine!

Many thx for the fix Toni!

[Mclane]

Well thanks to Paraj in this case, Toni normally but nice to see others lending a hand..

[Ponki1986]

Originally Posted by Toni Wilen

Fixed...

I confirm that everything works great. A big thank you for your work and patience needed while reading our requests.

[paraj]

Thanks, everything seems fixed, except the super minor romlist_cleanup thing:

Originally Posted by Toni Wilen

(i < j && roms[i].name) should be enough ("i < j" is needed or most roms get marked as not complete)

Is still off by one (for my rom list). The i<j loop exits fine, but ends with i==0x215 then i is incremented once more an it's past the end.

[Toni Wilen]

Originally Posted by Ponki1986

I've decided to upload a picture of the JIT message here as it may prove useful. I have encountered a similar situation in this thread before. As recommended, I have set with StackAttack2 an even bigger stack than the author recommends. In contrast, now the error message showed up when I started SnoopDOS and enabled logging. I then started Voodoo-X for archiving and wanted to check if it was missing something, because when I unzip something the progress window is empty. Unfortunately while running Voodoo-X that SnoopDOS in the background everything crashed.

Unfortunately these kind of JIT problems don't tell anything useful except that either PC was corrupted or memory was corrupted (listed opcode is obviously bogus but is it corrupted code or data. You can try disassembling around the bogus instruction to see if there are valid code nearby)

Use stack command to increase stack. Those helper programs can't work in all situations. And try without JIT. If it crashes or misbehaves (usually differently than with JIT): it is Amiga side problem.

[Toni Wilen]

https://download.abime.net/winuae/fi...nuae_4990b9.7z
https://download.abime.net/winuae/fi...ae64_4990b9.7z

Beta 9 (RC1):

- Direct3D9 mode overlay mode didn't show Amiga screen in most configurations.
- JIT interrupt handling fix part 2.
- Debugger SMC detector cache flush does not anymore clear whole SMC state array (which can be huge if Amiga RAM is located in Z3 space). When state array needs to be reset, reset only RAM and ROM regions. Max SMC array size is now automatically same as highest RAM address.
- Autoresolution didn't do anything unless gfx_windowed_multi or gfx_fullscreen_multi had non-zero values (b1)
- Autoscale worked incorrectly in programmed modes. (b1)
- Disable GUI Wait for blitter checkbox in 68000/010 cycle-exact modes. Previously it was allowed but did nothing.
- CDTV/CD32 Quickstart didn't use new NVRAM Files -path (b1)
- Added Supra AMAB2 ROM to ROM scanner. (Seems to support at least 2000DMA and AutoBoot models). AMAB1 is still missing, previously added 2000DMA ROM most likely isn't AMAB1 because it only supports 2000DMA. 2000DMA ROM selection now also lists all AMAB roms because AMAB2-4 have 2000DMA support (AMAB5/6 dropped 2000DMA support) and ROM scanner selection logic does not support filtering ROMs that have same type.

[Maijestro]

Hi Toni,

AmigaOs 4.1 FE Update 2 no longer boots with WinUae 4.9.9 beta I only get a green screen when booting.

It runs fine with WinUae 4.9.1.

winuaebootlog.txt

[Toni Wilen]

Originally Posted by Maijestro

Hi Toni,

AmigaOs 4.1 FE Update 2 no longer boots with WinUae 4.9.9 beta I only get a green screen when booting.

It runs fine with WinUae 4.9.1.

Attachment 77363

Unfortunately bootlog.txt won't help. You need to include both config and winuaelog.txt (easiest way is to tick "Enable full logging" in Paths panel, then load your config and start emulation, when it hangs, open Paths panel again, click Save All to generate file that includes both logs and config)

[aribeth23]

Hey really like where this is going but I still have trouble with DblPAL screenmode (640x512) I cannot get a good scaling with any 4.9 betas.


With 4.4 I used these settings which provided the best experience for me:

gfx_colour_mode=32bit
gfx_width_fullscreen=native
gfx_height_fullscreen=native
gfx_autoresolution=0
gfx_lores=false
gfx_resolution=hires
gfx_fullscreen_amiga=fullwindow
gfx_center_horizontal=smart
gfx_center_vertical=smart
gfx_filter_keep_autoscale_aspect=1
gfx_filter_keep_aspect=tv
gfx_filter_autoscale=integer_auto
gfx_filter_aspect_ratio=4:3

But that produces a picture even more off than with these settings (fresh config created only in the WinUAE config)


Is this something that is on the todo list? Am wondering because latest beta is considered RC1.


I also had a crash playing with the settings on the Filter config screen (with the WinUAE created config).

The Attachment winuae4.4_codeblock-reference.jpg shows how it looks in 4.4 with the settings of the above code block

The last picture shows how my config looks with 4.9 Beta9 RC1

The screenmode I use is:
"DBLPAL:High Res No Flicker"
0xA9004