HW tests
Posted: Tue Dec 01, 2020 6:25 amFew ideas for some hardware tests ..
Would help in understanding some of the low level details. Even if these seems potentially stupid would eliminate some potential possibilities as to how things are working
Test 1:
3D and 2d tilegen sync test. The tilegen if it is single buffered should be 1 frame ahead of the 3d. Many games like vf3 seem to write a dummy viewport with 2 polys attached.
My idea is to draw one of these polys and set the background to a set colour, then maybe next frame alternate the background colour and the poly colour
This is some dummy polys vf3 always seems to draw. These are sometimes actually visible (end credits)
https://i.imgur.com/zbQp7OJ.png
May want to make them larger but that should work for a test. Could swap like red to green and swap the background colour.
Test 2:
What happens if the 3d data is updated by 0x88 is never called. Does swap buffers just never happen? So it displays the last frame? Get a black frame?
What happens if 0x88 is called then it writes more data?
Sega rally 2 can fall apart if the clock speed is too slow because we just draw every frame, even if the 3d hasn't finished updating. The same is happening in vf3 in the loading sections. It's creating all kinds of crazy models and textures.
Would help in understanding some of the low level details. Even if these seems potentially stupid would eliminate some potential possibilities as to how things are working
Test 1:
3D and 2d tilegen sync test. The tilegen if it is single buffered should be 1 frame ahead of the 3d. Many games like vf3 seem to write a dummy viewport with 2 polys attached.
My idea is to draw one of these polys and set the background to a set colour, then maybe next frame alternate the background colour and the poly colour
This is some dummy polys vf3 always seems to draw. These are sometimes actually visible (end credits)
https://i.imgur.com/zbQp7OJ.png
May want to make them larger but that should work for a test. Could swap like red to green and swap the background colour.
Test 2:
What happens if the 3d data is updated by 0x88 is never called. Does swap buffers just never happen? So it displays the last frame? Get a black frame?
What happens if 0x88 is called then it writes more data?
Sega rally 2 can fall apart if the clock speed is too slow because we just draw every frame, even if the 3d hasn't finished updating. The same is happening in vf3 in the loading sections. It's creating all kinds of crazy models and textures.
In the meantime, if you wanted to tackle writing the test programs, that would be really helpful. I have a gcc cross compiler and libmodel3 should provide enough functionality (e.g., DMA, interrupts, etc.) to make this easily doable in C and testable in Supermodel today.