Some posts don't appear in the thread for a while. Scrolling back, previously unseen posts pop up between old posts.

Whiplash random polygons fixed:
https://nirvtek.com/downloads/Speedy3D.202405 … 1.Alpha.001.zip
MD5: 11f42c68e33a35710f6f7e9ee0d8990c

In action:
https://youtu.be/oRMVeXXlgDM

Through some work between myself and Pavel who also creates ICR2 tracks (in this thread: https://icr2.net/forum/showthread.php?tid=139 … =13778#pid13778) we are finding there is some limit in the *number* of mips around 50-60 per track. Seems like a limit imposed by the game and not hardware?

Also to follow up on a point from earlier, I don't think the issue was to do with the nesting of 3dos within 3dos. I could replicate the issue without using any of those nested 3dos.

checkpoint20 wrote on 2024-05-31, 23:35:

Through some work between myself and Pavel who also creates ICR2 tracks (in this thread: https://icr2.net/forum/showthread.php?tid=139 … =13778#pid13778) we are finding there is some limit in the *number* of mips around 50-60 per track. Seems like a limit imposed by the game and not hardware?

Also to follow up on a point from earlier, I don't think the issue was to do with the nesting of 3dos within 3dos. I could replicate the issue without using any of those nested 3dos.

I think the it relates to how the game internally manages memory., but yes, game would be limiting it and not the Verité.

Could it be the size of the mips? Smaller sizes allow more of them?

I think I've worked out why V2000 boards have visual anomalies in ICR2. The V2000 BIOS has a VRAM pointer (relative to the PhysPt of VRAM base), which is different to the V1000. On a V1000 the address is beyond the framebuffer. On a V2000 it's near the start of the framebuffer.

The actual address will probably be different. The example I've attached is for the 8 MB AGP SuperGrace BIOS. I also tested against a V2200 vanilla (no name) and it has the same values.

The block in question begins with 0x82,0x19,0x00Rendition ReferenceBoard ... .

The bit that needs to be modified comes after PN 1000ALL??ZX?<here>.

On a V2000 BIOS it's E4000000.
On a V1000 it's E8E1FF40.
By modifying these bytes I can use a V2000 BIOS with Dosbox.

I wouldn't recommend flashing a V2000 to test this because that address is probably used for other things which will probably break.

The modded BIOSes can be used with Dosbox-Rendition/RReady, but won't actually make a difference to gameplay.

sharangad wrote on 2024-06-01, 01:54:

Could it be the size of the mips? Smaller sizes allow more of them?

I suspect it is not just the size of mips, because there are tracks that appear to reach the limit even with a smaller total size of textures, but higher number of mips. For Detroit I had separate mips for many smaller things. I have noticed on the later Papyrus games, they often put multiple textures into one mip (page01, page02, page03.mip) perhaps to avoid having too many mips.

Eurofighter 2000+ menus functional (gameplay still broken):
https://nirvtek.com/downloads/Speedy3D.202406 … 1.Alpha.001.zip
MD5: c6cf2e2a96a03dd21ceb4cc807a29f27

To run this
1) Eurofighter 2000+ has to be installed,
2) From the GFX+ app set detail to hight
2) Launch the game from the command line. Launching from GFX+ will not work.

Audio stutter in the menu can sometimes occur, alt-tabbing to dosbox and then back again will fix it.

checkpoint20 wrote on 2024-05-31, 23:35:

Through some work between myself and Pavel who also creates ICR2 tracks (in this thread: https://icr2.net/forum/showthread.php?tid=139 … =13778#pid13778) we are finding there is some limit in the *number* of mips around 50-60 per track. Seems like a limit imposed by the game and not hardware?

Also to follow up on a point from earlier, I don't think the issue was to do with the nesting of 3dos within 3dos. I could replicate the issue without using any of those nested 3dos.

3d0 size +mip size +pmp size (total level size). The level geometry ends up being stored in system ram as well.

not sure if this is of any use but:

if you have 4bit cars and start elkhart track, switch view from fpp to tpp with f10, and drive till you drive under briggs & stratton sign
than you will be able to exit and load laguna track (gonna have solid colors instead of textures though)

a reversed situation here:

if you have 4bit cars and start ohio track and drive till you drive under nissan sign (passing camel sign a bit earlier is enough too)
than you will NOT be able to load miami track

robertmo3 wrote on 2024-06-03, 06:42:

not sure if this is of any use but:

if you have 4bit cars and start elkhart track, switch view from fpp to tpp with f10, and drive till you drive under briggs & stratton sign
than you will be able to exit and load laguna track (gonna have solid colors instead of textures though)

Nice, modded tracks!

above tracks are taken from this post:
https://icr2.net/forum/showthread.php?tid=1419

EF2000, a hot mess:
https://youtu.be/pZEoty04PpM

This game has random command buffer corruption.

robertmo3 wrote on 2024-06-03, 09:27:

a reversed situation here:

if you have 4bit cars and start ohio track and drive till you drive under nissan sign (passing camel sign a bit earlier is enough too)
than you will NOT be able to load miami track

If you run the Elkhart hq and Laguna Seca hq tracks and then try the custom Nashville track it'll fail to load. This is a problem with the game. Memory management for textures is entirely under application control for Speedy3D.

Nashville :
https://icr2.net/forum/showthread.php?tid=135 … =11743#pid11743

https://nirvtek.com/downloads/Speedy3D.202406 … 4.Alpha.001.zip
MD5: 2190e87db0e92b8d2b0442ec2a9b99ed

Decent II cutscenes and menus work. This game must be run in FIFO mode at the moment. With DMA the exe crashes long before initialising video.

No video because the game appears to blit to the front buffer and OBS can't capture that, unless someone can tell me how. I don't have dedicated capture hardware.

Extremely alpha, other apps may be partially broken in subtle ways. Whiplash, I've noticed has a few regressions.

This isn't going on the first page. A bit too alpha.

Current state of Descent II:

Not sure if anyone want to play with an alpha build, though I did manage to fix the underline character in under 'W' in "Whiplash" in the rwhip demos. This also affects the bottoms of some characters in the player rankings list.

It was caused by VL_Square/VL_Rectangle (V_FIFO_SQUARE/V_FIFO_RECTANGLE) being called without enabling texture clamping (non-repeating texturemode).

Stable build (only for Speedy3D apps):
https://nirvtek.com/downloads/Speedy3D.202406 … 7.Alpha.001.zip
MD5: 0fa7b3449f2e2b2f6f4c1d7e26dbc6e0

- Descent II still not working, but elements of gameplay are now visible.
- Whiplash, working well.
- All other working dos apps should work as before.

There's also a new dosbox-rendition.conf option:
[rendition]
cacheddma = false

when set to true it does the same thing as threadsafe, but uses a more logical method to do so. It will allow [ALT+X] mode to function correctly with ICR2 on machines with discrete GPUs. On slower machines it could potentially be slower than threadsafe=true. It's best to see which one works best, at least until I figure out what's wrong with DMA. With enough CPU and GPU grunt neither of these settings should be needed.

Tomb Raider *might*produce an out of texture memory error with the last build. It's been fixed and will be released in another build later this afternoon.

Rebel Moon chroma keying and HUD works perfectly now:
https://youtu.be/-cWgGqMfzIg

Nascar 2/Nascar 1999 runs fine when RReady caps the framerate to 30 fps (Unforunately the last few builds have a broken maxfps settings in rendition.cfg). The graphical issues haven't been fixed yet.

The next build due later today will fix all these things. Regression testing first.

Build:

https://nirvtek.com/downloads/Speedy3D.202406 … 9.Alpha.001.zip
MD5: de7becd46dfcc935c0d415c47fbb5ccf

- Nascar 2 and Nascar 1999 hard coded framerate cap. Cockpit view messed up.
- Rebel Moon chroma keying changes. HUD should render correctly.
- Tomb Raider "Out of memory" OpenGL error fixed.