WinUAE 3.1.0 beta series

[Toni Wilen]

Originally Posted by guyfry

"24-bit address space is not supported with selected accelerator board configuration". -> 68060 running + PPC!?

Missed check, it always complains, even if 24-bit addressing is already disabled..

Tried the recent beta for testing my forthcoming ocs amiga demo, and glad that I did. One of the parts (twist scroll, that changes 15 palette colors and bitmap offsets per line) worked fine in winuae 3.0 but caused corruption on the new winuae. I fixed this by having a starting x offset on the copper to $d1 (instead of $d9) 15 colors get written from the copper from this position and then 4 lowords for the bitplane pointers (in a 288x186) screen. I guess to make sure I could have adjusted the pointers at a more safe place inbetween the color writes.

I should get a real amiga at some point as well :-)

[Toni Wilen]

Originally Posted by algorithm

Tried the recent beta for testing my forthcoming ocs amiga demo, and glad that I did. One of the parts (twist scroll, that changes 15 palette colors and bitmap offsets per line) worked fine in winuae 3.0 but caused corruption on the new winuae. I fixed this by having a starting x offset on the copper to $d1 (instead of $d9) 15 colors get written from the copper from this position and then 4 lowords for the bitplane pointers (in a 288x186) screen. I guess to make sure I could have adjusted the pointers at a more safe place inbetween the color writes.

I should get a real amiga at some point as well :-)

I'd like to see your non-working version to confirm it. No need for complete demo, just executable that runs copper list and all higher bitplane dma channels active (=bitplane dma). Bitplane data does not need to be correct either

One difference in recent betas is edge case updates when DDFSTRT/DDFSTOP is modified while old/new value would match at the same time.

[turrican3]

incredible work on this winuae 3.1, just amazing.
I'll test this FPS.Adj, one more great feature !
I really think that without your work the amiga couldn't have so much project in 2015.

Toni, sent you a pm with the link to the demo part (non-working version) which causes corruption on most recent beta of winuae

[Toni Wilen]

Originally Posted by algorithm

Toni, sent you a pm with the link to the demo part (non-working version) which causes corruption on most recent beta of winuae

Thanks. Real A500 has identical graphics problem. Perfect result

[Toni Wilen]

http://www.winuae.net/files/b/winuae_3100b10.zip

Beta 10:

- Added hd controller subtype selection.
- A590/A2091: DMAC-01/02 revision selection.
- GVP Series I: all revisions.
- Remaining non-DMA SupraDrives added.
- Masoboshi MC-302 (IDE-only) and MC-702 (IDE+SCSI). Note that only MC-302 is currently working, DMA also not yet supported. (Driver by default uses only PIO)
- Added A2090/A2090a emulation. (Currently shared with A590/A2091, it is not possible to use A2090 and A590/A2091 at the same time)
- Previous supradrive config entry name changed, config reset needed.
- A590/A2091 interrupt handling updates.
- 68000 address error stacked PC accuracy improved (MOVEM and NOT, CLR and other single operand instructions)
- Apparently GVP Series I Z2 RAM autoconfig id is 1761/8.
- Accelerator board without IDE and SCSI and SCSI (auto) or IDE (auto) selected: drive was incorrectly reserved for selected accelerator board.
- Reorganized accelerator board handling, built-in HD controller configuration source code special cases removed. ROM selection shows only ROM images compatible with selected accelerator board.
- Added accelerator rom config/jumper options. Currently only map rom for Blizzards (ROM panel MapROM emulation still also selects it) and OSMODE (J304) jumper for A26x0.
- WD33C93 emulation fix, A2630+A2090 Amix installation works.
- "24-bit address space is not supported with selected accelerator board configuration" error even when 24-bit was not enabled.
- Bogus 0x20 68882 stack frame version id removed, both 68881 and 68882 use 0x1f.
- Set/reset accelerator ROM correctly in Quickstart mode.
- Debugger fl was already reserved for listing active break points. Scanline fl renamed to fs.
- Debugger library and device list commands (Td, Tl) shows version, revision, opencount and id string.
- Debugger task list command (Tt) shows signals that task is waiting for and PC where task continues executing when Wait() returns. Also previously it showed extra bogus entry and usually also didn't show all tasks..
- Added extra prefetch hack for A26x0 ROM off switch code if 68030 MMU is enabled.
- On the fly switching from AGA to non-AGA mode didn't reset FMODE value.

A2090(a):

- Commodore's first Amiga harddrive controller and it shows..
- Custom partition table (No RDB), autoboot is supported.
- Very confusing and illogical partitioning system. (No, double clicking "Prep HD" icon won't do what you would expect it to do. At least with SCSI-only configuration.)
- 34.4 boot ROM added.
- At least 34.4 has broken format routine (few absolute addresses are not relocated correctly in rom code) for SCSI drives. Disable boot ROM and use binddrivers mounting when running prep-program as a workaround.
- ST-506 parts not emulated.

GVP Series I:
- Impact A2000-1/X (Autoconfig Product number 1)
- Impact A2000-HC (Product 2)
- Impact A2000-HC+2 (Product 3)
- Nearly identical hardware. Different SRAM buffer size and SRAM address offset.
- Early partition software created incompatible RDB blocks. (For example checksum fields are zeroed)
- v1.0 (1.16 driver version) ROM added.

SupraDrives:
- A500 ByteSync/XP (was already supported since b8)
- 2000 WordSync
- A500 Autoboot
- 4x4 (Non autoboot). Clock chip not yet emulated.
- Supra AMABx ROMs support all above models.
- All Supradrives except very rare 2000 DMA (which apparently had bad DMA implementation and uses very different driver) are now emulated.

[mark_k]

There seems to be an AutoConfig-related problem. Not sure whether this is specific to beta 10, but I noticed it just now with a config that has both GVP Series I and Supra SCSI controllers and a UAE shared folder.

On a cold boot I can hold both mouse buttons to bring up the boot menu. All partitions on the GVP- and Supra-connected HDFs are shown, as is the UAE shared folder. However if I select one to boot from, on clicking Use then Boot, the Supra and UAE AutoConfig devices vanish so the system only boots from the GVP HDF.

[Toni Wilen]

Originally Posted by mark_k

There seems to be an AutoConfig-related problem. Not sure whether this is specific to beta 10, but I noticed it just now with a config that has both GVP Series I and Supra SCSI controllers and a UAE shared folder.

On a cold boot I can hold both mouse buttons to bring up the boot menu. All partitions on the GVP- and Supra-connected HDFs are shown, as is the UAE shared folder. However if I select one to boot from, on clicking Use then Boot, the Supra and UAE AutoConfig devices vanish so the system only boots from the GVP HDF.

Supra autoconfig was not reset, after first reset autoconfig read supra's IO registers instead of autoconfig data.

[mark_k]

Also, I noticed some log messages on booting my A-Max HDF connected to Supra controller (after putting suprascsi.amhd in the Devs drawer). Booting to Workbench there were a few lines of
raw_scsi_put_data but bus phase is 3!

Running A-Max, when it was booting from the Mac partition there were many lines of
raw_scsi_put_data but bus phase is -2!

Maybe a Supra software bug, if not a WinUAE issue?

[Toni Wilen]

Originally Posted by mark_k

Also, I noticed some log messages on booting my A-Max HDF connected to Supra controller (after putting suprascsi.amhd in the Devs drawer). Booting to Workbench there were a few lines of
raw_scsi_put_data but bus phase is 3!

Running A-Max, when it was booting from the Mac partition there were many lines of
raw_scsi_put_data but bus phase is -2!

Maybe a Supra software bug, if not a WinUAE issue?

It is probably "soft scsi" emulator bug, it isn't perfect yet, it has some assumptions that are not exactly correct. Those messages should be harmless if program still runs normally.

[mark_k]

Load the attached config and boot to Workbench. Press F12 to open settings, change CPU type to 68000 then click OK. WinUAE crashes. (You don't have to boot from an HDF; booting from a WB 3.1 disk still shows the problem.)

[Toni Wilen]

Originally Posted by mark_k

Load the attached config and boot to Workbench. Press F12 to open settings, change CPU type to 68000 then click OK. WinUAE crashes. (You don't have to boot from an HDF; booting from a WB 3.1 disk still shows the problem.)

This is not new. Any CPU mode without "more compatible" enabled will crash if CPU's opcode fetch accesses non-existing memory. It is performance optimization.

It is near guaranteed when switching from 68010+ to 68000 CPU due to different exception/interrupt stack frames. Don't do it. It is never going to work, even if it isn't going to crash.

[Ariath]

I think A590 emulation is bugged in the Beta 10.

I installed today a new Workbench 2.1 system creating a new hardfile and selecting A590/A2091 as the controller. The A590 v7.0 ROM was correctly selected and the revision was the DMAC-01.

At first glance, everything was fine. I used HDToolBox to create one partition in the hardfile and installed Workbench, all OK.

The problems came after the reboot. Opening the Locale application caused an #8000000B error, the ShowConfig fails opening, IconEdit crashes, etc ...

I tried then installing two games: Elvira Mistress of the Dark and Eye of the Beholder.
Again, the setup was fine, but when i opened the installation drawers... the icons are all screwed up .

I have going back to the latest stable version 3.0.0, and the game icons are OK now.

See ya!

[Toni Wilen]

Originally Posted by Ariath

I think A590 emulation is bugged in the Beta 10.

I installed today a new Workbench 2.1 system creating a new hardfile and selecting A590/A2091 as the controller. The A590 v7.0 ROM was correctly selected and the revision was the DMAC-01.

At least your config is different now, previously DMAC-02 was always selected. I don't think DMAC-01 emulation is tested very well.

[mark_k]

Originally Posted by Toni Wilen

Beta 8:
- Added GVP G-Force 030 board. Same as GVP A530 but supports more RAM, located at 0x01000000.

Is that emulation of the G-Force 030 (Impact A2000-030 Combo Series II)? I have the earlier Impact A2000-030 Combo. If amiga.resource.cx is to be believed both boards are product number 11. Their capabilities are very similar (max memory capacity is the same). Board layout differs and the G-Force 030 uses PGA CPU & FPU vs QFP/PLCC for the earlier Combo board.

I'm not sure what software-visible differences there might be between the two, but at least the Impact A2000-030 Combo works with v3.x GVP ROMs as well as 4.x. Currently WinUAE only detects the 4.15 ROM when G-Force 030 is selected.

For what it's worth, the 3.15 GVP ROM supports up to 16MB RAM at $01000000 and 32MB at $02000000. If no RAM is present outside the 24-bit region its memory check routine expects $01000000 to be mirrored to address 0. Perhaps some/all G-Force 030 boards support 16MB GVP SIMMs? That would explain the maximum allowed by the ROM; 4MB on-board (which would be mapped to Zorro II space) plus three 16MB SIMMs.

[Toni Wilen]

Originally Posted by mark_k

Is that emulation of the G-Force 030 (Impact A2000-030 Combo Series II)? I have the earlier Impact A2000-030 Combo. If amiga.resource.cx is to be believed both boards are product number 11. Their capabilities are very similar (max memory capacity is the same). Board layout differs and the G-Force 030 uses PGA CPU & FPU vs QFP/PLCC for the earlier Combo board.

All GVP SCSI compatible turbo boards seem to use product number 11. Only differences seem to be DMA controller address space size and amount of RAM.

I'm not sure what software-visible differences there might be between the two, but at least the Impact A2000-030 Combo works with v3.x GVP ROMs as well as 4.x. Currently WinUAE only detects the 4.15 ROM when G-Force 030 is selected.

Works and was designed for it/manufactured with 3.x rom are two different things

For what it's worth, the 3.15 GVP ROM supports up to 16MB RAM at $01000000 and 32MB at $02000000. If no RAM is present outside the 24-bit region its memory check routine expects $01000000 to be mirrored to address 0. Perhaps some/all G-Force 030 boards support 16MB GVP SIMMs? That would explain the maximum allowed by the ROM; 4MB on-board (which would be mapped to Zorro II space) plus three 16MB SIMMs.

Real hardware information needed. Because same ROM is supported by multiple GVP boards, ROM data can't be used to guess hardware capabilities, like RAM size.

[mark_k]

Originally Posted by Toni Wilen

Works and was designed for it/manufactured with 3.x rom are two different things

Maybe I should have been a bit clearer. Some of these boards did ship with a version 3.x ROM.

My GVP Combo 030 board is the same version as shown at http://www.bigbookofamigahardware.co...e030_1_big.jpg. My board came with the same "OMNI3.5" ROM shown in the picture (gvpscsi.device 3.14). The rev.3 board pic at amiga.resource.cx shows ROM version 3.12. (Its label says VER.3.12S/3.2A; I wonder if that version also contains an IDE driver for the older GVP accelerators?)

Real hardware information needed. Because same ROM is supported by multiple GVP boards, ROM data can't be used to guess hardware capabilities, like RAM size.

That might be hard to obtain, e.g. testing whether a board works with 16MB SIMMs if that capability wasn't advertised. But that's not too critical for emulation.

[amilo3438]

Originally Posted by Toni Wilen

Beta 5:

- Some tweaks to writing to BPLxPT 1 cycle before BPL DMA fetches same bitplane. Nakudemo / Vision-X ECS only corruption is now correct (and also looks much worse than previously).

Oh yeah, in ECS really looks ugly... but what about OCS, is that corruption also correct? (see pic.1)

(tested in beta 10 A500 most common)

EDIT: In correct ECS corruption there is no interference in the form of vertical lines as in OCS! (pic.2)

[Arnie]

My CDTV set up with HD attached stops with a HALT2. working in WinUAE3.0