English Amiga Board

English Amiga Board (http://eab.abime.net/index.php)
-   support.WinUAE (http://eab.abime.net/forumdisplay.php?f=5)
-   -   OS4 compatible UAE expansion development status (http://eab.abime.net/showthread.php?t=81146)

Toni Wilen 16 January 2016 19:58

And even more bigger update!

Directory harddrives now work in OS4!

(Directory hardfiles didn't need library calls, only uaegfx and bsdsocket needs those)

ero__ 16 January 2016 20:57

Quote:

Originally Posted by Toni Wilen (Post 1063122)
And even more bigger update!

Directory harddrives now work in OS4!

(Directory hardfiles didn't need library calls, only uaegfx and bsdsocket needs those)

Great News! :crazy

Aegis 16 January 2016 22:55

Awesome news! Posted at amiga.org :)

AMIGASYSTEM 16 January 2016 23:15

Thank you Toni, pubblicato su Amigapage.it, amiga.ikirsector.it, aros-exec.org and other sites devoted to windows.

AnnaWu 17 January 2016 01:04

Quote:

Originally Posted by Toni Wilen (Post 1063122)
And even more bigger update!

Directory harddrives now work in OS4!

(Directory hardfiles didn't need library calls, only uaegfx and bsdsocket needs those)

Excellent job, Toni! :)

amilo3438 17 January 2016 01:43

Small reminder for others thrilled with the news and who have not yet donated... :)

Quote:

Originally Posted by Toni Wilen (Post 1062719)
Donations are accepted, doing OS4 related stuff is quite uninteresting (did you expect something else? :))

Nice, but honestly, personally care more about the classics Amiga! (sorry)

(otherwise it seems like a very interesting project... go ahead) :great

Toni Wilen 17 January 2016 10:48

Some notes, there is also possibility for "you are completely on your own" test version later today..

Currently supported UAE devices:

- Directory harddrives.
- uaehf.device (hardfiles). HD_SCSICMD is NOT supported yet.

Supported but not tested:

- uae.resource
- Some "uaelib" functions

Not yet supported (lots more work needed):

- uaegfx
- bsdsocket.library
- uaenative.library
- clipboard sharing
- uaescsi.device
- uaenet.device
- virtual mouse driver
- on the fly directory harddrive changes.

And everything else that was not mentioned.

Notes:

Host<>Amiga communication system is currently busy waiting, both PC and Amiga sides. Until it gets replaced with proper waits/interrupts (very complex task) at least triple core CPU is recommended.. (one for main thread, one for PPC, one for trap thread). I don't know if there is any difference but it is not my problem as usual :)

EDIT: Host side busy waiting is now replaced with waiting for signal from Amiga side.

Communication is also slower than normal host<>Amiga (which is practically instant) because every read or write that accesses Amiga memory needs to be done in Amiga side (due to non 1:1 logical/physical mapping). Lots of easy improvents to do here but optimizations come later

As usual, stable and working implementation is the most important and first requirement.
Any non-trivial optimizations will come (much) later.

Aegis 17 January 2016 11:47

Looking forward to testing (and breaking everything horribly :D)

Toni Wilen 17 January 2016 14:42

Very alpha quality test version: http://www.winuae.net/files/b/winuae_OS4UAE.7z

If you post about any problem: log and winuaelog.txt MUST BE included. Log illegal memory accesses in Misc panel must be checked.

Posts that are useless or annoying will disappear.

See my previous post for supported features.

Extra notes: shutting down emulation can crash if trap thread is still waiting. It is considered normal at this point.

Directory harddrive performance is not that good (dozens of context switches/dos packet). Make sure hardfile/directory harddrive boot priority is below CSPPC SCSI if booting from SCSI. Equal boot priorities = UAE device will boot.

Add "uaeboard=full+indirect" to config file to enable new trap system.

Aegis 17 January 2016 16:13

Thanks Tony - I've got my CyberstormPPC emulation booting off a uaehf.device hardfile just fine (although it seems a tad slower than SCSI at this time) - haven't had any luck trying to get a BlizzardPPC to boot yet though (possibly because I haven't tried setting that configuration up before).

*Edit* Tested a directory hard drive (just a 'shared' folder) with a partitioned uaehf.device boot hardfile and that's all working nicely. Didn't notice any speedups with the latest build. Still can't get a Blizzard PPC install to work so someone else will have to test that...

Toni Wilen 17 January 2016 16:24

I didn't optimize uaehf.device context switches at all, wanted to get directory harddrives working first. Update later..

EDIT: Quick uaehf.device optimization done.

Nickman1200 17 January 2016 18:57

Quote:

Originally Posted by Toni Wilen (Post 1063240)
Some notes, there is also possibility for "you are completely on your own" test version later today..

Currently supported UAE devices:

- Directory harddrives.
- uaehf.device (hardfiles). HD_SCSICMD is NOT supported yet.

Supported but not tested:

- uae.resource
- Some "uaelib" functions

Not yet supported (lots more work needed):

- uaegfx
- bsdsocket.library
- uaenative.library
- clipboard sharing
- uaescsi.device
- uaenet.device
- virtual mouse driver
- on the fly directory harddrive changes.

And everything else that was not mentioned.

Notes:

Host<>Amiga communication system is currently busy waiting, both PC and Amiga sides. Until it gets replaced with proper waits/interrupts (very complex task) at least triple core CPU is recommended.. (one for main thread, one for PPC, one for trap thread). I don't know if there is any difference but it is not my problem as usual :)

Communication is also slower than normal host<>Amiga (which is practically instant) because every read or write that accesses Amiga memory needs to be done in Amiga side (due to non 1:1 logical/physical mapping). Lots of easy improvents to do here but optimizations come later

As usual, stable and working implementation is the most important and first requirement.
Any non-trivial optimizations will come (much) later.

Did a quick donate to keep you a bit interested :bowdown

Toni Wilen 18 January 2016 18:31

Thanks :)

Update:

- stability improved.
- PC side busy looping is gone.
- Increased size of shared memory buffers.
- Memory leak fixed.

I can also confirm OS4 can fully boot from directory harddrive. Slowly..

Performance improvements and others need redesign of UAE fs interrupt handler, because it is not possible to Wait() in interrupt, separate worker task is needed. Later..

Aegis 18 January 2016 19:06

2 Attachment(s)
Great stuff! Just made a (small) donation - will make another one soon!

*Edit* Weird - finally got Blizzard PPC up and running (256MB yay!) but even using uaehf.device OS 4.1 seems massively slower than using the Cyberstorm emulation - I've attached my configs in case I'm missing something obvious?? :confused

Attachment 47044
Attachment 47045

Toni Wilen 19 January 2016 22:07

Wrong thread. That has nothing to do with UAE expansions! (It is known problem, no one knows the reason). Donations won't make it any easier to post offtopic posts ::)

Toni Wilen 19 January 2016 22:08

Possibly interesting screenshot attached..

EDIT: Screenshot removed, better one re-attached to later post.

Aegis 19 January 2016 22:18

Quote:

Originally Posted by Toni Wilen (Post 1063952)
Wrong thread. That has nothing to do with UAE expansions! (It is known problem, no one knows the reason). Donations won't make it any easier to post offtopic posts ::)

I was kinda hoping the loading times for Blizzard PPC would be comparable to Cyberstorm PPC using uaehf.device instead of the emulated IDE - clearly not :)

Quote:

Originally Posted by Toni Wilen (Post 1063953)
Possibly interesting screenshot attached..

Ahaha! That's awesome!

huepper 19 January 2016 22:48

Quote:

Originally Posted by Toni Wilen (Post 1063953)
Possibly interesting screenshot attached..

Really amazing!!! :great

ccapublic 19 January 2016 22:57

Quote:

Originally Posted by Toni Wilen (Post 1063953)
Possibly interesting screenshot attached..

you sure ? Noooo....

Considering how slow you're progressing, we should not expect sound, gfx & bsdsocket alpha before... end of next week I guess... ;)

mulder77 20 January 2016 13:29

Quote:

Originally Posted by Toni Wilen (Post 1063953)
Possibly interesting screenshot attached..

Any chance to have a new test version any time soon? ;)


All times are GMT +2. The time now is 13:26.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, vBulletin Solutions Inc.

Page generated in 0.06183 seconds with 11 queries