English Amiga Board

English Amiga Board (https://eab.abime.net/index.php)
-   support.WinUAE (https://eab.abime.net/forumdisplay.php?f=5)
-   -   WinUAE 4.9.0 beta series (Was 4.5.0) (https://eab.abime.net/showthread.php?t=104099)

Toni Wilen 31 October 2020 18:49

https://download.abime.net/winuae/fi...nuae_4500b7.7z
https://download.abime.net/winuae/fi...ae64_4500b7.7z

Beta 7:

- Fixed incorrect bitplane DMA overrun condition that made Zool 2 AGA flicker strangely (b1) (AGA game with 7 planes but FMODE=0. What a waste of DMA slots.)
- JIT FPU 80-bit host mode should work again (b1)
- Emulated clockport expansions (Prelude A1200 is currently the only one) now support 4 port clockport expander. "-" = no expander (Same device mapped to all 4 slots), "1" = expander slot 1 and so on.
- Added Zorro 2/A500/CDTV DMA controller 24-bit DMA checkbox in Expansion panel. (Unticked = full 32 addressing capable, ticked = 24-bit access only, real world behavior) Previously some emulated 24-bit DMA controllers had full 32-bit support, some didn't.
- All Z2/Z3 RAM banks now have DMA capable checkbox in RAM panel. If not ticked and DMA HD controller tries to access it, read returns all ones and writes do nothing. (In real world result could be memory corruption, hang etc). This can be used to quickly test if some real world DMA controller + accelerator combination with accelerator RAM set to no DMA causes similar problems in emulation. Other RAM banks only have config file support. Find RAM name in config file and then replace "*_size" with "*_options=nodma=true". For example "bogomem_options=nodma=true" makes slow ram not DMA capable. NOTE: Chipset DMA is not affected even if chip ram DMA is disabled, it only affects mainboard (CDTV and CD32) and expansion CD/HD DMA controllers.
- Added CD image select/eject keyboard shortcut. END+F5 = select CD image, END+SHIFT+F5 = eject CD. Statefile quick restore moved to END+F6. NOTE: currently END+F5 opens even if there is no emulated CD drive.
- If game controller was plugged in after config was loaded and loaded config had game controller configured in game ports panel, game controller was correctly remembered but type and autofire mode was set to default values.
- Hardware emulated Z3 RTG boards had wrong autoconfig size configuration. This caused access faults if MMU was enabled.
- SVGA chip vblank interrupt acknowledge forced unnecessary full display refresh.
- ALT+Space does not anymore open WinUAE window standard menu when input is captured.
- JIT + Hardware emulated RTG board should fully work again.
- JIT MOVEM/MOVE16 direct RAM access optimization (calculate address only once, then do moves vs do separate multiple calculate address+move combinations) is now automatically disabled if hardware emulated RTG board is enabled, most have byte/word swapping logic which would cause cause graphics corruption if same MOVEM/MOVE16 copy was first used for RAM reads or writes and then later (when it was still in translated format) it was used for VRAM reads or writes. At least CGX4 does this.
- JIT LSL/LSR fixed (temporary fix, better fix later that should also fix other shift instructions). jit_blacklist config entry should be removed now.

AZka 31 October 2020 18:52

Thank You!

Viceroy 31 October 2020 18:59

Many thanks ;-)

White 31 October 2020 19:15

Toni, thanks for the latest beta.

With JIT activated some icons disappear with the beta 7
with all color depths 15-16-24-32 bits
if I deactivate the JIT, the icons are seen again
using CVision3D zorro 3

Seiya 31 October 2020 19:55

good. it works with jit.

Toni Wilen 31 October 2020 20:07

Quote:

Originally Posted by White (Post 1438213)
Toni, thanks for the latest beta.

With JIT activated some icons disappear with the beta 7
with all color depths 15-16-24-32 bits
if I deactivate the JIT, the icons are seen again
using CVision3D zorro 3

Required information missing: Picasso96 or CGX? Default WB? Also if JIT without direct?

White 31 October 2020 20:26

JIT indrect and direct
cgx latest
workbench 3.9 default
the problem does not occur with beta 6
can the configuration be useful?

White 31 October 2020 20:35

1 Attachment(s)
config

Toni Wilen 31 October 2020 21:34

Quote:

Originally Posted by White (Post 1438231)
JIT indrect and direct
cgx latest
workbench 3.9 default
the problem does not occur with beta 6
can the configuration be useful?

Get rid of CSPPC first, try with 68020 CPU. Any difference?

White 31 October 2020 21:56

1 Attachment(s)
Toni
the configuration I posted is not the one I use
for the hurry I created a configuration with indirect and selected ppc I corrected it as you suggested it crashes here is the dump

AMIGASYSTEM 31 October 2020 23:28

Thanks Toni, Beta7 AfA One CVision3D 060/JIT 800x600x16, Doom2, works well :)

https://drive.google.com/file/d/1kU2...ew?usp=sharing

Mad-Matt 01 November 2020 00:50

2 Attachment(s)
I know you dont want to know about ppc issues but as this is a regression in the new beta, id like to mention that using CVision3D the colours are no longer correct when running a PPC App/Game window or fullscreen.

I post an example of nightlong ppc fullscreen and Gears window mode. PPC and 68k version side by side. Nightlong 68k runs fine. Using PicassoIV Nightlong ppc looks correct. Last working version - Beta6

CGXv4, NOJIT, 15bit wb screen

Toni Wilen 01 November 2020 09:29

Quote:

Originally Posted by White (Post 1438247)
Toni
the configuration I posted is not the one I use
for the hurry I created a configuration with indirect and selected ppc I corrected it as you suggested it crashes here is the dump

My crystal ball said you adjusted it on the fly. Don't do it.

Quote:

Originally Posted by Mad-Matt (Post 1438275)
I know you dont want to know about ppc issues but as this is a regression in the new beta, id like to mention that using CVision3D the colours are no longer correct when running a PPC App/Game window or fullscreen.

I post an example of nightlong ppc fullscreen and Gears window mode. PPC and 68k version side by side. Nightlong 68k runs fine. Using PicassoIV Nightlong ppc looks correct. Last working version - Beta6

CGXv4, NOJIT, 15bit wb screen

winuae.7z should fix it. RTG boards are now more JIT friendly which made it also "too PPC" friendly, allowing it to use direct mode even when board has byte/word swap hardwre.

EDIT: now it is really fixed.

Mad-Matt 01 November 2020 14:59

The issue is still present. (Confirming correct build dated 2020.11.01)

Toni Wilen 01 November 2020 15:46

But it is later build? Original fix didn't work. After this fix at least in my test setup gears-ppc has correct colors. (and strange colors without the fix)

EDIT: if it still happens, attach PPC address space information (appears when PPC starts in winuaelog.txt) and screenshot.

Mad-Matt 01 November 2020 16:22

1 Attachment(s)
Quote:

Originally Posted by Toni Wilen (Post 1438384)
But it is later build? Original fix didn't work. After this fix at least in my test setup gears-ppc has correct colors. (and strange colors without the fix)

EDIT: if it still happens, attach PPC address space information (appears when PPC starts in winuaelog.txt) and screenshot.

Screenshot is 68k normal and ppc miscoloured side by side on a 15bit screen. Nightlong ppc and other ppc demos are effected as well.


Code:

WinUAE 4.5.0 (Public Beta Beta 7, 2020.11.01) (10.0.18363 [1]) 64-bit 9.6.3C03 8 09:58 0

20-813 [661 110-482]: PPC: uae_ppc_cpu_reboot
20-814 [661 110-482]: PPC: Loading QEmu implementation
20-815 [661 110-482]: DLOPEN: Loaded plugin qemu-uae.dll
20-815 [661 110-482]: DLOPEN: Patching common functions
20-815 [661 110-482]: Loaded qemu-uae library at 501C0000
20-815 [661 110-482]: Imported qemu_uae_init
20-816 [661 110-482]: Imported qemu_uae_start
20-816 [661 110-482]: QEMU: Initializing
20-816 [661 110-482]: QEMU: Initialize QEMU-UAE (QEMU 2.2.0 + API 3.6.0)
20-816 [661 110-482]: QEMU: MODULE_INIT_QOM
20-818 [661 110-482]: Imported qemu_uae_ppc_init
20-818 [661 110-482]: Imported qemu_uae_ppc_in_cpu_thread
20-818 [661 110-482]: QEMU: ppc_cpu_init_with_model 604e => 604e_v2.4
20-818 [661 110-482]: QEMU: Initializing PPC CPU model 604e_v2.4
20-819 [661 110-482]: PPC: Creating thread CPU 0/TCG
20-819 [661 110-482]: QEMU: CPU PVR 0x00090204
20-819 [661 110-482]: QEMU: Starting main loop
20-820 [661 110-482]: PPC: Loaded qemu-uae library at 501C0000
20-820 [661 110-482]: QEMU: Main thread running
20-820 [661 110-482]: QEMU: Running main loop
20-820 [661 110-482]: Starting PPC thread.
20-820 [661 110-482]: PPC: uae_ppc_cpu_reset
20-821 [661 110-482]: PPC: Hard reset
20-822 [661 110-482]: QEMU: Map memory regions:
20-822 [661 110-482]: QEMU: 00 00000000 [+  200000]  =>  7FFF0000  "Chip memory")
20-823 [661 110-482]: QEMU: 01 00bf0000 [+  10000]  =>  00000000  "CIA")
20-823 [661 110-482]: QEMU: 02 00c00000 [+  100000]  =>  00000000  "Custom chipset")
20-823 [661 110-482]: QEMU: 03 00dc0000 [+  10000]  =>  00000000  "Battery backed up clock (RF5C01A)")
20-824 [661 110-482]: QEMU: 04 00dd0000 [+  10000]  =>  00000000  "A4000 IDE")
20-824 [661 110-482]: QEMU: 05 00de0000 [+    8000]  =>  00000000  "Motherboard Resources")
20-824 [661 110-482]: QEMU: 06 00df0000 [+  10000]  =>  00000000  "Custom chipset")
20-824 [661 110-482]: QEMU: 07 00e80000 [+  10000]  =>  00000000  "Autoconfig Z2")
20-824 [661 110-482]: QEMU: 08 00ea0000 [+  10000]  =>  00000000  "UAE Board")
20-825 [661 110-482]: QEMU: 09 00eb0000 [+  10000]  =>  00000000  "UAE Boot ROM")
20-825 [661 110-482]: QEMU: 10 00f00000 [+  40000]  =>  80EF0000  "CPUBoard F00000")
20-825 [661 110-482]: QEMU: 11 00f40000 [+    1000]  =>  00000000  "CyberStorm SCSI IO")
20-825 [661 110-482]: QEMU: 12 00f41000 [+    f000]  =>  80F31000  "CyberStorm SCSI RAM")
20-825 [661 110-482]: QEMU: 13 00f50000 [+  20000]  =>  00000000  "CPUBoard IO")
20-826 [661 110-482]: QEMU: 14 00f80000 [+  80000]  =>  80F70000  "Kickstart ROM")
20-826 [661 110-482]: QEMU: 15 08000000 [+ 8000000]  =>  87FF0000  "RAMSEY memory (high)")
20-826 [661 110-482]: QEMU: 16 44000000 [+  400000]  =>  C3FF0000  "PCem SVGA VRAM")
20-827 [661 110-482]: QEMU: 17 44400000 [+  400000]  =>  C3FF0000  "PCem SVGA VRAM")
20-827 [661 110-482]: QEMU: 18 44800000 [+  400000]  =>  C3FF0000  "PCem SVGA VRAM")
20-827 [661 110-482]: QEMU: 19 44c00000 [+  400000]  =>  C3FF0000  "PCem SVGA VRAM")
20-827 [661 110-482]: QEMU: 20 45000000 [+  10000]  =>  00000000  "PCem SVGA MMIO")
20-828 [661 110-482]: QEMU: 21 45800000 [+  10000]  =>  00000000  "PCem SVGA MMIO WORDSWAP")
20-828 [661 110-482]: QEMU: 22 47000000 [+  10000]  =>  00000000  "PCem SVGA MMIO WORDSWAP")
20-828 [661 110-482]: QEMU: 23 48000000 [+  10000]  =>  00000000  "PCem SVGA SPC")
20-828 [661 110-482]: QEMU: 24 4c000000 [+  40000]  =>  00000000  "PCem SVGA IO SWAP2")
20-829 [661 110-482]: QEMU: 25 4c0e0000 [+  40000]  =>  00000000  "PCem SVGA PCI")
20-829 [661 110-482]: QEMU: 26 ff000000 [+  10000]  =>  00000000  "Autoconfig Z3")
20-829 [661 110-482]: QEMU: 27 fff00000 [+  80000]  =>  8FF70000  "CPUBoard MAPROM")
20-829 [661 110-482]: QEMU: Reset CPU
20-829 [661 110-482]: QEMU: NIP = 0xfff00100
20-830 [661 110-482]: QEMU: Flushing all JIT translation blocks
20-830 [661 110-482]: QEMU: Set state 1
20-830 [661 110-482]: QEMU: Resumed!


Toni Wilen 01 November 2020 16:50

Quote:

Originally Posted by Mad-Matt (Post 1438392)
Code:

20-826 [661 110-482]: QEMU: 16 44000000 [+  400000]  =>  C3FF0000  "PCem SVGA VRAM")
20-827 [661 110-482]: QEMU: 17 44400000 [+  400000]  =>  C3FF0000  "PCem SVGA VRAM")
20-827 [661 110-482]: QEMU: 18 44800000 [+  400000]  =>  C3FF0000  "PCem SVGA VRAM")
20-827 [661 110-482]: QEMU: 19 44c00000 [+  400000]  =>  C3FF0000  "PCem SVGA VRAM")


C3FF0000 should be all zeros. Strange because "PPCIO" flag is always set..
Recheck the download, it is not that rare that old cached version gets downloaded. (It should be dated few hours ago only)

Mad-Matt 01 November 2020 18:45

Well that was strange. Downloaded on a different computer and got what i think is the latest fix build ;) Colours working again thx.

White 01 November 2020 19:35

Toni
fixed graphic imperfections with workbench 3.9 now everything works again
strangely the beta 6 made me think that it could be a winuae problem but now the beta 7 works well even with the JIT activated.
Beta 7 brought to light this problem that actually didn't exist with beta 6.
Only with the activation of the JIT did I notice.
Thanks again for your patience

Zilog 01 November 2020 19:45

Great Toni!


Thanks!


All times are GMT +2. The time now is 20:06.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.

Page generated in 0.13604 seconds with 11 queries