Threaded Mode | Linear Mode

komame · 09-05-2024, 08:08 PM

(09-05-2024 05:11 PM)jfelten Wrote: I just installed the firmware, and I quite like the new bug fixes, but there are still a few issues which I would like to be addressed, as well as several feature requests, particularly regarding Python:

1. The hpprime drawing commands won't accept anything larger than 0x7fffffff as a color parameter, which makes is impossible to draw anything above half transparency (technically hpprime.eval() is possible but it's really slow). Anything above that threshold for example `hpprime.pixon(0, 0, 0, 0xff000000)`, outputs "OverflowError: overflow converting long int to machine word".

2. The garbage collector doesn't seem to work on `hpprime.eval("ROTATE(...)")`, which can result in a crash due to memory overflow after repeated calls.

3. It would be really nice to have an equivalent to GETPIX in Python, as `hpprime.eval("GETPIX_P(...)")` is very slow.

4. Some other PPL commands I've love to see in Python are TRIANGLE (especially the 3D renderer) and TEXTSIZE, since again, calling them through eval() yields poor performance.

5. There are issues with semi-transparent pixels being coped with full opacity with the blit commands.

6. Would it be possible to add texture mapping to TRIANGLE()?
[...]

When it comes to graphic commands, we still lack INVERT, optionally FILLPOLY (although it would be good to have some sort of flood fill) and definitely need to add font size to TEXTOUT.

Currently, executing the PPL command using hpprime.eval() operates slower than direct execution in PPL because each call to hpprime.eval() involves parsing and compilation (tokenization) before the PPL engine executes the code. It seems to me that, similar to the compile method in Python which allows saving a compiled piece of code to a variable for execution, a similar method could be implemented for PPL in the hpprime module. This way, the compilation could be performed once at the program’s start, allowing this compiled code to be executed multiple times without the need for reparsing and recompiling each time. This should significantly speed up the operation of any PPL functions called from Python.