Homebrew GBARunner2

[Gericom]

Huh. That seems a bit arcane and maybe even flashcard-dependent.
Do you have any clues about what causes the flashcard to decide if a particular .nds is homebrew or retail? It doubt it's something as simple as the presence of a NitroFS.

Wood seems to look at the game id and the arm7 load address, I'm not sure about other loaders. And indeed, if those commands are supported at all is really flashcard dependent. A fact is though that accessing romfs the Nintendo way is often way faster than the libnitrofs way of doing it, especially with large nds files.

 

[metroid maniac]

Wood seems to look at the game id and the arm7 load address, I'm not sure about other loaders. And indeed, if those commands are supported at all is really flashcard dependent. A fact is though that accessing romfs the Nintendo way is often way faster than the libnitrofs way of doing it, especially with large nds files.

Makes sense, since homebrew prefers to use dldi instead.
How much is known about current compatibility issues? What are the major performance bottlenecks? What are the potential solutions?
Thanks for answering all my questions, by the way. I find the project very interesting.

 

[migles]

It's a complete hardware debugger. It connects to a pc over usb. You can then upload a rom straight from your pc and run and debug it (both arm9 and arm7). You can break it, set breakpoints, peek in and modify the memory and registers. It also supports up to 2 hardware breakpoints (read and/or write in address range) and you can print debug messages. With the communication library you can also communicate with the rom from a program running on your pc for quick testing.

If you like I could take some screenshots of the debugging software later this week.

This is-nitro has no video or wifi option. Instead of wifi it has an ethernet port with the word WIRELESS next to it xD. It seems to be non-standard ethernet as my pc acted as if no cable was inserted, so more research is needed into that. I am probably gonna have a look with a logic analyzer. I want to try to convert the signals back to wifi using a fpga and a raspberry pi.

It may happen that it only opens the interface when there is WiFi activity?
I remember reading that the console didn't even have the WiFi drivers.. They are located in the cartridges (that's why almost every single cart has the WiFi config thing) but then again the console has pico chat... But thats other protocol of communicating... (if I am right)

@Dodain47 may I suggest you to use pastebin.com? Create a single pastebin.com link and then edit with the new titles you test.. You created a mess in the thread..
OE maybe in github, or a Google docs Excell? that way you can even easily re use and edit when a new gba runner version comes out and there is more compatibility

 

[Deleted User]

Thanks for the suggestion @migles, but I don´t really understand pastebin or the Google docs Exell...

 

[Gericom]

Makes sense, since homebrew prefers to use dldi instead.
How much is known about current compatibility issues? What are the major performance bottlenecks? What are the potential solutions?
Thanks for answering all my questions, by the way. I find the project very interesting.

The major performance bottleneck is the fat chain. If the fat chain is not cached in some way (for direct lookup), you have to walk all the way along it for non-sequencial accesses.

It may happen that it only opens the interface when there is WiFi activity?
I remember reading that the console didn't even have the WiFi drivers.. They are located in the cartridges (that's why almost every single cart has the WiFi config thing) but then again the console has pico chat... But thats other protocol of communicating... (if I am right)

@Dodain47 may I suggest you to use pastebin.com? Create a single pastebin.com link and then edit with the new titles you test.. You created a mess in the thread..
OE maybe in github, or a Google docs Excell? that way you can even easily re use and edit when a new gba runner version comes out and there is more compatibility

Well, some kind of, but I have just done some tests with my logic analyzer, and it is outputting data when I run a multiboot parent. I'm not sure why my pc wouldn't even detect it, but I'm almost sure I must be able to decode this data using my fpga. The logic analyzer (which is actually the same fpga) has no differential settings, so the capture is not very good, but the fpga has differential input buffers, so I should be able to use that.

As for the compatibility list, we already have one on the gbatemp wiki. I agree it clutters the topic a little. @Dodain47 Maybe you should not post like everything in this topic, but only if you find special cases. I still really appreciate your testing though, don't get me wrong.

 

[Deleted User]

Ah, yes of course! Now I get it. I first thought @migles meant the compatibility wiki. Sorry about the mess. I will keep it in mind.

 

[metroid maniac]

The major performance bottleneck is the fat chain. If the fat chain is not cached in some way (for direct lookup), you have to walk all the way along it for non-sequencial accesses.

This is the physical DLDI FAT filesystem, not some sort of filesystem embedded in the GBA ROM? I'm guessing that GBARunner2 wouldn't need to be aware of such a thing.
I know that nds-bootstrap maintains a cache of some part of the FAT. Maybe the same problem was encountered and solved there.

 

[Gericom]

This is the physical DLDI FAT filesystem, not some sort of filesystem embedded in the GBA ROM? I'm guessing that GBARunner2 wouldn't need to be aware of such a thing.
I know that nds-bootstrap maintains a cache of some part of the FAT. Maybe the same problem was encountered and solved there.

The nitrofs implementation of libnds simply opens the nds file on your sd card with libfat. I suggested indeed to preload the fat chain, which is what I do in gbarunner2 too. Still it could probably be faster the Nintendo way if flashcards would implement some kind of cache in hardware.

 

[Mr Objection]

Do you have any compiled version that I could test? With the last changes I mean. And I want to say that I love your work, you are incredible.

 

[Gericom]

Do you have any compiled version that I could test? With the last changes I mean. And I want to say that I love your work, you are incredible.

I'm gonna merge the file explorer branch to master this weekend and I'll upload the latest compiled version here then.

 

[SLiV3R]

Very cool project! Good luck with it

 

[Pomegrenade]

So will GBARunner work on the original R4 for us old DS homebrew developers, or only on specific models, such as the R4i Gold?

 

[Gericom]

So will GBARunner work on the original R4 for us old DS homebrew developers, or only on specific models, such as the R4i Gold?

Yes, it works fine on the original r4. I have one myself

 

[Pomegrenade]

Yes, it works fine on the original r4. I have one myself

Cool, and wow I tried your YouTubeDS and "WholyLonelyPonyEatingMacorroni" it's impressive, do you know when it will get a GUI update (like seeing thumbnails)

 

[Gericom]

Good news, I've made some improvements using the is-nitro-emulator.

Big mutha truckers and street racing syndicate boot now. They had code in vram appearently at some place I didn't handle correctly yet. This might fix some other games too.

The file explorer branch has been merged into master now too.

Cool, and wow I tried your YouTubeDS and "WholyLonelyPonyEatingMacorroni" it's impressive, do you know when it will get a GUI update (like seeing thumbnails)

Maybe when I have time to do that.

 

[Deleted User]

Thanks for the update @Gericom!

I have tested some Games with the updated Master and compared them with the prievious Version.

Banjo Kazooie has a bit more Slowdown. The Audio has a bit more Noise. Circle of the Moon has a little less Slowdown. DKC 1 has a little less Slowdown. I noticed more Noise during the Intro. DKC 2 has a little less Slowdown. Much more Gfx Glitches during the First Waterlevel. DKC 3 has more Slowdown and much more Gfx Glitches during the First Level. Mario and Luigi Superstar Saga has a bit more Slowdown. The Audio has a bit more Noise and stuttering. Wario Land 4 has some Slowdown now.

3D: V-Rally 3, Doom 2, Driver 2, Driver 3 are not working. Ford Racing 3 is almost Full Speed now and has less Gfx Glitches. Doom runs a bit faster. NFS Most Wanted is working again. The Buildings have no Textures.

You will find the latest Version of the master below.

 

[Gericom]

I committed some bugfixes today.

Fixed some bugs, fixed is-nitro break(points)
- Disable FIQ in the abort handler
- Pass control to the debugger for is-nitro breakpoint opcodes
- Fixed the vcount comparison value in the displaystat register to match the values read from the vcount register
- Fixed dmaing from the edge of what was loaded in ram of the rom

Fixed games: F-Zero Maximum Velocity (fixes corrupted menu graphic and track invisible), Herbie Fully Loaded. Might impact other games too

Maybe more mode-7 games that were broken previously are fixed now. Maybe @Dodain47 wants to test a couple of them.

This is the latest commit ba222be:

 

[Deleted User]

I committed some bugfixes today.

Maybe more mode-7 games that were broken previously are fixed now. Maybe @Dodain47 wants to test a couple of them.


This is the latest commit ba222be:

Whoa. A major Improvement. All three F-Zero are working. Maximum Velocity is Full Speed now. Climax and GP Legend almost with many Gfx Glitches. I will test more Games and post the Results on the wiki.

Edit: Advance Guardian Heroes (E) and Astro Boy Omega Factor (E) are both having much Slowdown in this Version and are Unplayable.

 

[Gericom]

...
Edit: Advance Guardian Heroes (E) and Astro Boy Omega Factor (E) are both having much Slowdown in this Version and are Unplayable.

Are you sure it's this version, and not already in the version before that?

 

[Deleted User]

Are you sure it's this version, and not already in the version before that?

It goes back to the old File Explorer Version Commit f3d2ef4 (2017/02/18) I have posted.