Minimig (Amiga) core discussion

https://github.com/MiSTer-devel/Main_MiSTer/wiki

Moderators: Mug UK, Zorro 2, Greenious, spiny, Sorgelig, Moderator Team

BBond007
Captain Atari
Captain Atari
Posts: 391
Joined: Wed Feb 28, 2018 3:23 am

Re: Minimig (Amiga) core discussion

Postby BBond007 » Tue Apr 30, 2019 4:43 pm

R4MS wrote:Do you mean http://aminet.net/package/driver/media/IDEfix97 ? What is the purpose of this driver in MiSTer? Does it increase transfer rates?


Like Sorgelig said, Its primary purpose is to allow 4 IDE devices when used with a special IDE splitter board, however (on a real Amiga anyway) it also increases transfer rates 200-300k/sec...
You do not have the required permissions to view the files attached to this post.

R4MS
Atari maniac
Atari maniac
Posts: 78
Joined: Sun Jan 13, 2019 6:54 pm

Re: Minimig (Amiga) core discussion

Postby R4MS » Tue Apr 30, 2019 5:34 pm

Until last Friday, I've been OCS only on real hardware. But my newly acquired A1200 indeed also has one of those IDE splitter boards. I'll have a look at it. That's a nice way to get to know my new Amiga.

Can somebody provide a reproducible crash scenario? In particular it would be interesting, which IDE devices were enabled. Moreover, at aminet there is the IDEFIX driver, a turbo patch and a patch from icomp for use with CF cards. It would be important to know which combination of those produce the crash.

blacklistedcard
Retro freak
Retro freak
Posts: 14
Joined: Thu Dec 20, 2018 1:06 am

Re: Minimig (Amiga) core discussion

Postby blacklistedcard » Wed May 01, 2019 5:47 am

I was wondering what people are running video resolution wise. I'm only getting 640x400 high res laced and lower resolutions offered by the DisplayMode program. Anybody running at 1280×720 or 1920×1080 with more than 256 colours on the mister?

R4MS
Atari maniac
Atari maniac
Posts: 78
Joined: Sun Jan 13, 2019 6:54 pm

Re: Minimig (Amiga) core discussion

Postby R4MS » Wed May 01, 2019 9:58 am

With Kickstart and Workbench 3.1 I get 1444x564 in Super-High Res Laced. More than 256 are probably not possible with AGA chipset, I only use 16 to get decent speed. To get maximal resolution, you have to select 2M chipram, (at least) 2M fastram and AGA chipset in OSD and use the latest ROMS/Workbench.

blacklistedcard
Retro freak
Retro freak
Posts: 14
Joined: Thu Dec 20, 2018 1:06 am

Re: Minimig (Amiga) core discussion

Postby blacklistedcard » Wed May 01, 2019 1:02 pm

R4MS wrote:With Kickstart and Workbench 3.1 I get 1444x564 in Super-High Res Laced. More than 256 are probably not possible with AGA chipset, I only use 16 to get decent speed. To get maximal resolution, you have to select 2M chipram, (at least) 2M fastram and AGA chipset in OSD and use the latest ROMS/Workbench.


Thanks. Let me go back and see what my settings are. I just see a bunch of people with WorkBench with like 8K looking resolution and colors and wonder how are they getting that.

R4MS
Atari maniac
Atari maniac
Posts: 78
Joined: Sun Jan 13, 2019 6:54 pm

Re: Minimig (Amiga) core discussion

Postby R4MS » Wed May 01, 2019 1:10 pm

There are RTG graphics cards for real Amigas, you probably mean those. They are also available in WinUAE. Minimig currently only recreates original Amiga hardware up to an A1200 with fastram and freezer module.

fpgaarcade
Obsessive compulsive Atari behavior
Obsessive compulsive Atari behavior
Posts: 102
Joined: Thu Sep 20, 2007 10:06 pm
Location: Sweden

Re: Minimig (Amiga) core discussion

Postby fpgaarcade » Wed May 01, 2019 1:47 pm

mahen wrote:Hi ! In the first post of this page (viewtopic.php?f=117&t=32761&start=475) Mike mentions the M68K soft 020 core "which is a tidy up of" TG68K and supposedly fixes quite a few issues (like the Pinball Illusions crash I guess). Would anyone be willing to implement this ?

M68K haven't been released. He gave test version of source to some people with condition to not release the sources till he finish his improvements. Minimig is open source core, so such condition is unacceptable. Thus i wait for release M68K first and only then it will be integrated (if it will be compatible).
So, ask Mike to release the sources.[/quote]

I know what the problem is with the 000 mode but I haven't had time to fix it yet, it's quite complex.
I'm busy porting the whole world to Replay1/2 (and tied up with work).
I plan to blitz this early next week.
I've also got trapV to fix as well, but the core is passing all the other regression tests.
/Mike

fpgaarcade
Obsessive compulsive Atari behavior
Obsessive compulsive Atari behavior
Posts: 102
Joined: Thu Sep 20, 2007 10:06 pm
Location: Sweden

Re: Minimig (Amiga) core discussion

Postby fpgaarcade » Wed May 01, 2019 1:51 pm

It's a drop in replacement for the TG68K as it's mostly the same core, so it should be no problem to run in the minimig design.
The superscaler version is completely different.
I'm still experimenting with extending the M68K to a true 32 bit bus, but making any changes to this design are rather fraught.
/Mike

R4MS
Atari maniac
Atari maniac
Posts: 78
Joined: Sun Jan 13, 2019 6:54 pm

Re: Minimig (Amiga) core discussion

Postby R4MS » Wed May 01, 2019 4:07 pm

My last bugfix had itself a small bug, which messes things up, if you use a lot of fast ram. I fixed that and also I did a little work on the cartridge module. The cartridge address space used to be always visible, also when hrtmon was disabled. Now it starts shows up, when the freeze is first requested, effectively hiding it completely when hrtmon is off. That's also how hrtmon works in WinUAE. Interestingly the RAM bank is not write protected... but since it's also that way in WinUAE and making it read-only by brute force breaks hrtmon, I think it should stay that way for now.

The pull request will be up today or tomorrow. I first have to find out, why my IO-board does not produce VGA output. Well, my first SMD soldering job...

Sorgelig
Ultimate Atarian
Ultimate Atarian
Posts: 5237
Joined: Mon Dec 14, 2015 10:51 am
Location: Russia/Taiwan

Re: Minimig (Amiga) core discussion

Postby Sorgelig » Wed May 01, 2019 4:39 pm

R4MS wrote:My last bugfix had itself a small bug, which messes things up, if you use a lot of fast ram. I fixed that and also I did a little work on the cartridge module. The cartridge address space used to be always visible, also when hrtmon was disabled. Now it starts shows up, when the freeze is first requested, effectively hiding it completely when hrtmon is off. That's also how hrtmon works in WinUAE. Interestingly the RAM bank is not write protected... but since it's also that way in WinUAE and making it read-only by brute force breaks hrtmon, I think it should stay that way for now.

The pull request will be up today or tomorrow. I first have to find out, why my IO-board does not produce VGA output. Well, my first SMD soldering job...

I've just about to release the update. Ok, i won't announce it now. Will wait for your fixes then.
What exactly do you mean by VGA output? If you mean VGA with standard VGA frequency then it doesn't. It's raw video from the core. So your monitor simply may not accept the Amiga video signal.

R4MS
Atari maniac
Atari maniac
Posts: 78
Joined: Sun Jan 13, 2019 6:54 pm

Re: Minimig (Amiga) core discussion

Postby R4MS » Wed May 01, 2019 8:26 pm

OK. Pull request is sent.

My next project is to add the option to boot from Kickstart disks. The code for mirroring a 256K Kickstart is already in the commit, but still disabled. It will take some advanced plumbing to make it accessible via OSD. For one, the A1000 bootrom images floating around are not of the same size, so we have to use some other measures to prevent users from loading the wrong rom file.

My IO-Board works! Both my monitors want composite HSync turned on. One of them actually is sort of able to sync 15kHz video, but gets the color wrong. The images of MiSTer and my A1000 look almost identical.

@sorgelig: Now that I have VGA video, I would like to create a "light" revision of the core without HDMI for faster synthesis. The straight forward method would be just to remove all hdmi* files from the revision. Is this the right way, or is there a better one?

BBond007
Captain Atari
Captain Atari
Posts: 391
Joined: Wed Feb 28, 2018 3:23 am

Re: Minimig (Amiga) core discussion

Postby BBond007 » Wed May 01, 2019 10:59 pm

R4MS wrote:Until last Friday, I've been OCS only on real hardware. But my newly acquired A1200 indeed also has one of those IDE splitter boards. I'll have a look at it. That's a nice way to get to know my new Amiga.


Congratulations on your A1200. It is my favorite Amiga system :)
It could be a regular IDE splitter board, however the 4xEIDE is another popular option that comes with different software - "EIDE'99"

R4MS wrote:Can somebody provide a reproducible crash scenario? In particular it would be interesting, which IDE devices were enabled. Moreover, at aminet there is the IDEFIX driver, a turbo patch and a patch from icomp for use with CF cards. It would be important to know which combination of those produce the crash.


With whdload I notice it more with Sierra games which preload a lot of data. For example King's Quest V and Space Quest IV.

Whdload starts to load the game but the whdload window never goes away. The mouse continues to work and windows can be moved around but the window contents are not properly redrawing and the system is hung.

Another situation where I encounter it is when transferring large files.

For example:
I have a System and Work hard disk - SYSWORK.HDF
A large HAM video hard disk formatted with PFS3. HAMVID.HDF
A 1GB transfer hard disk formatted with FFS which I mount on loop-back and FTP files into - TRANS.HDF

Ideally I would like to boot off of SYSWORK.HDF also have both HAMVID.HDF and TRANS.HDF available to copy files. Obviously having IDEFix active is necessary to access 3 HDFs simultaneously.

The problem is while copying files (via CLI or GUI) When using IDEFix I get the system hang.

I am using just the standard install of IDEFix and have never installed any patches.

What I end up doing is booting off a Workbench.ADF with HAMVID.HDF and TRANS.HDF mounted and copy the files that way.

Thanks :)
You do not have the required permissions to view the files attached to this post.

BBond007
Captain Atari
Captain Atari
Posts: 391
Joined: Wed Feb 28, 2018 3:23 am

Re: Minimig (Amiga) core discussion

Postby BBond007 » Wed May 01, 2019 11:14 pm

blacklistedcard wrote:Thanks. Let me go back and see what my settings are. I just see a bunch of people with WorkBench with like 8K looking resolution and colors and wonder how are they getting that.


R4MS wrote:There are RTG graphics cards for real Amigas, you probably mean those. They are also available in WinUAE. Minimig currently only recreates original Amiga hardware up to an A1200 with fastram and freezer module.


Probably using WinUAE or other UAE based emulator.

I'm not aware of any real Amiga graphics boards for classic Amigas supporting anywhere near that resolution.
Last edited by BBond007 on Wed May 01, 2019 11:29 pm, edited 3 times in total.

BBond007
Captain Atari
Captain Atari
Posts: 391
Joined: Wed Feb 28, 2018 3:23 am

Re: Minimig (Amiga) core discussion

Postby BBond007 » Wed May 01, 2019 11:15 pm

blacklistedcard wrote:I was wondering what people are running video resolution wise. I'm only getting 640x400 high res laced and lower resolutions offered by the DisplayMode program. Anybody running at 1280×720 or 1920×1080 with more than 256 colours on the mister?


With this monitor driver you can get the following in the DisplayMode program:

- HGFX - 1024x768
- HD720 - 1280x720

--> http://aminet.net/package/driver/moni/HighGFX40_6

I don't suggest running more than 16 colors because of performance reasons and the amount of chipram consumed.

You can still have a nice looking desktop.

I suggest this read --> http://www.mfilos.com/2012/01/guide-mak ... using.html

Sorgelig
Ultimate Atarian
Ultimate Atarian
Posts: 5237
Joined: Mon Dec 14, 2015 10:51 am
Location: Russia/Taiwan

Re: Minimig (Amiga) core discussion

Postby Sorgelig » Thu May 02, 2019 5:21 am

R4MS wrote:@sorgelig: Now that I have VGA video, I would like to create a "light" revision of the core without HDMI for faster synthesis. The straight forward method would be just to remove all hdmi* files from the revision. Is this the right way, or is there a better one?

With open source scaler the compilation time doesn't differ much with one without it. So it won't worth the hassle.
On some cores I create Quartus 13 project version. In most cases it takes almost twice less time to compile. So, for quick evaluation Q13 is good. Q13 uses beta timings so it may produce incorrect builds and it didn't work well with Minimig core.
But with timings fixes applied some time ago it may work better now. I need to try again.

Btw: in pull requests please try to avoid qpf/qsf files changes if you don't really add any new files to project.
Also, it's better if you will use Quartus 17.0.2. Later versions don't add anything for Cyclone V - just compilation time getting worse. And you will always get qsf/qpf patched with new version and potentially incompatible options.

Sorgelig
Ultimate Atarian
Ultimate Atarian
Posts: 5237
Joined: Mon Dec 14, 2015 10:51 am
Location: Russia/Taiwan

Re: Minimig (Amiga) core discussion

Postby Sorgelig » Thu May 02, 2019 7:10 am

I've pushed Q13 project files. Q13 takes ~8m30 to compile on my computer. While Q17 needs ~13m40.

Q13: you need exactly Quartus 13.1 update 4 version in order to be able to compile.

Sorgelig
Ultimate Atarian
Ultimate Atarian
Posts: 5237
Joined: Mon Dec 14, 2015 10:51 am
Location: Russia/Taiwan

Re: Minimig (Amiga) core discussion

Postby Sorgelig » Thu May 02, 2019 8:05 am

With DualShock 4 you can play Walker game on a single gamepad emulating joystick and mouse at the same time. A whole new experience!
P.S.: found that Walker has blinking 1-pixel column at the left. It wasn't there sometime ago... Will check it later.

R4MS
Atari maniac
Atari maniac
Posts: 78
Joined: Sun Jan 13, 2019 6:54 pm

Re: Minimig (Amiga) core discussion

Postby R4MS » Thu May 02, 2019 12:08 pm

BBond007 wrote:I am using just the standard install of IDEFix and have never installed any patches.


Can you post (or PM) your version of IDEFix, in case you did not directly download it from aminet. Are your HDFs in the single partition format (used also by WinUAE)?

One workaround would be to put two of your devices as partitions on a single (partitioned) image (Well, at least in theory... haven't tried it myself). In linux you might have to use "partprobe" on your loopback device to expose partitions (which will be named loop0p1, etc).

BBond007
Captain Atari
Captain Atari
Posts: 391
Joined: Wed Feb 28, 2018 3:23 am

Re: Minimig (Amiga) core discussion

Postby BBond007 » Fri May 03, 2019 4:15 am

R4MS wrote:Can you post (or PM) your version of IDEFix, in case you did not directly download it from aminet. Are your HDFs in the single partition format (used also by WinUAE)?


Done... I'm pretty sure it same as Aminet. Both 413,511 bytes.

All my HDFs are RDB (prepared with HDInstTools) style and not the UAE style. I typically avoid FFS fileSys for large partitions and use PFS3.

I do believe the HD image remowilliams (and others) are referring to is the UAE style (and FFS fileSys) if its the image set I'm thinking it is.

That may suggest that partition type, bootblock and fileSys are not factors?

R4MS wrote:One workaround would be to put two of your devices as partitions on a single (partitioned) image (Well, at least in theory... haven't tried it myself).


I do do that, for example my SYSWORK.HDF is a 512MB/3.5GB "System:" and "Work:".

So far I have been making my HDFs 4GB in size because I know there are some considerations as far as Amiga once you exceed 4GB. Also the HDF files I create will still work with my Minimig 1.1 if I don't exceed 4GB.

That is one reason I like the EIDE'99 software (and hardware) more than IDEFix97 (On a real Amiga) because it can split a large HD into multiple 4GB drives.

I could certainly put a small system on my HAMVID.HDF, but booting off the Workbench.ADF for the purpose of copying files works well enough.

I'm just discouraged IDEFix cause instability because having 4 HDFs is a great feature for "power users". It really helps out having two extra HDFs if you also want to mount CD-ROM ISO files.

R4MS wrote:In linux you might have to use "partprobe" on your loopback device to expose partitions (which will be named loop0p1, etc).


I will have to experiment. I'm not sure if I would be able to mount PFS3 partitions (such as my HAMVID.HDF) directly on loopback. I'm not sure I want to risk corruption. I have corrupted my TRANS.HDF a few times (usually by forgetting to properly umount it) which is no big deal. Umounting the 1GB HDF takes a really long time after uploading files, so I'm sure a 4GB file would not be any faster :)

Thanks for looking at this problem!

Don't forget: Only Amiga makes it possible! --> http://y2u.be/1C1_7d9bfGc

JimDrew
Atari Super Hero
Atari Super Hero
Posts: 828
Joined: Mon Nov 04, 2013 5:23 pm

Re: Minimig (Amiga) core discussion

Postby JimDrew » Sat May 04, 2019 7:30 am

Sorgelig wrote:I've pushed Q13 project files. Q13 takes ~8m30 to compile on my computer. While Q17 needs ~13m40.


Wow... and here I complain about 1.6 million lines of assembly code taking almost 5 seconds to assemble! :)
I am the flux ninja

BBond007
Captain Atari
Captain Atari
Posts: 391
Joined: Wed Feb 28, 2018 3:23 am

Re: Minimig (Amiga) core discussion

Postby BBond007 » Sun May 05, 2019 12:59 am

JimDrew wrote:
Sorgelig wrote:I've pushed Q13 project files. Q13 takes ~8m30 to compile on my computer. While Q17 needs ~13m40.


Wow... and here I complain about 1.6 million lines of assembly code taking almost 5 seconds to assemble! :)


Q17 needs 21m44 to compile it on my AMD FX 8320. Obviously not the best architecture on which to run Quartus...

R4MS
Atari maniac
Atari maniac
Posts: 78
Joined: Sun Jan 13, 2019 6:54 pm

Re: Minimig (Amiga) core discussion

Postby R4MS » Mon May 06, 2019 10:29 pm

Now Minimig is able to boot with the A1000 bootrom! Required a minor fix of the routing of reset signals.

The Amiga has two slightly different reset triggers. One external (keyboard reset) and one from the cpu (RESET instruction). The first one resets everything and has to go through minimig_syscontrol.v, which keep it active for some cycles. The cpu resets everything but itself (and also holds the signal active for a while) and therefore should not be routed through minimig_syscontrol.v.

Since this bugfix probably has no application besides running certain boot roms (using the RESET instruction), I'll finish the A1000 implementation and release it in one piece.

R4MS
Atari maniac
Atari maniac
Posts: 78
Joined: Sun Jan 13, 2019 6:54 pm

Re: Minimig (Amiga) core discussion

Postby R4MS » Tue May 07, 2019 12:17 am

New Feature: Boot like an A1000!

That was a little quicker than I expected. The current implementation hat two problems tough:

1. I tied the function to the "OCS-A1000" option, which is maybe not too surprising, but limits this option to 256K Kickstarts (like on an unmodded A1000)
2. You need to prepare an unencrypted A1000 boot rom and pad it to 256K or 512K to be acceptable for Main_Mister. But just booting a 256K Kickstart also works like before.

So, for now, we trade the option to use OCS-A1000 with 512K Kickstarts for an obscure way to boot from an A1000 boot rom.

I anyway planed to look at Main_Mister for some adjustments, but this will take a while.

What`s the opinion here? Should we release this version or wait for an adapted OSD/Main_Mister?

Sorgelig
Ultimate Atarian
Ultimate Atarian
Posts: 5237
Joined: Mon Dec 14, 2015 10:51 am
Location: Russia/Taiwan

Re: Minimig (Amiga) core discussion

Postby Sorgelig » Tue May 07, 2019 11:42 am

I think A1000 is not an urgent feature, so it's better to pilish it from both FPGA and HPS sides so no workarounds will be required.

R4MS
Atari maniac
Atari maniac
Posts: 78
Joined: Sun Jan 13, 2019 6:54 pm

Re: Minimig (Amiga) core discussion

Postby R4MS » Tue May 07, 2019 9:53 pm

OK. Just had a look at Main_Mister. A minimal version would be to define a new command, say OSD_CMD_ROM, in osd.h and send the type of rom loaded directly from UploadKickstart() in minimig_config.cpp to the FPGA. That way, changes would mostly be confined to minimig specific code. Would that be an acceptable approach?


Return to “MiSTer”

Who is online

Users browsing this forum: InfecteD and 4 guests