themachine02 / virtual3d Goto Github PK

All the absolute value from file for indexing and offseting should be defined and not hardcoded.

Example :

ld ix, (iy+0) add ix, bc ld hl, (iy+10) ld (ix+4), l ld (ix+5), h ld a, (ix+6) lea hl, ix+0

To

ld ix, (iy+VX_INDEX0) add ix, bc ld hl, (iy+VX_INDEX_UV) ld (ix+VX_VERTEX_GPR0), l ld (ix+VX_VERTEX_GPR1), h ld a, (ix+VX_VERTEX_UNIFORM) lea hl, ix+0

This should probably moved out of it.

Because the geometry shader directly write to temporary buffer for sorting, the only way to skip sort is to copy data from temporary buffer to actual geometry queue. This is definitly bad for performance.
A better way should be to have a differential output way for geometry shader which should then decide to do sorting or not.

EDIT : several implementations are possible. Either the geometry shader select the queue in which data should go, and then must take care of the queue (maybe will helper function ?) or the switch is a bit more internal and geometry shader only declare the sorting/nosorting status.

Clipping method soemthing produce un-usefull triangle. A better approach could be found to avoid these which are expensive in later stage.

It appears as if the script (build.bat) still uses spasm instead of fasmg.
(Additionaly) Is there a Plan to implement a C wrapper? Id be willing to help out, but Id need a quick headsup regarding public? asm defines.

There is a random crash leading to reset with NMI trigger somewhere. Most likely in texture code or cliping code. This need to be investigated.
Based on early test, there may be a use after free from previous frame.
1 - try to clean up to code in init
2 - poison memory for an easy debug framework

A convenient way to load shader from external source should be included. It has to resolve linking issues, and copy issues.

Due to change in flash exploit, the library crash on OS > 5.5. Fix it by changing the exploit.

Hi!

I don't really code ASM, to be honest I'm just playing with my calculator right now.
I tried the build of the demo but it end up with a white screen and the calculator isn't responding the only way of getting out of it is by resetting.
You can reproduce the bug (if it's one) with the lastest firmware of TI-83 CE and installing by putting everyfile in the Archive and only TESTLVL.8xp in RAM memory and run the program from CarbiJR using Artifice.

I know this project is very old and didn't receive much update, I still hope to find help also I'm wondering if there is a way to debug what's happening on the calculator while plugged on computer.

Thank in advance.