poly

@polyproxy.bsky.social

programmer (he/him), reverse engineer.

Received the SpacemiT K3 #RISCV board today, and already running Crysis. I need to optimize the emulator over the next few months, but this is way better performance than the K1 already.

Bild

I'm very thankful for GCC's consistency when it comes to codegen, really. Which is why I'm quite surprised on the rare cases where this consistency disappears and codegen changes considerably. 🧵

I really hate how GT5 has great tunes from the previous games (notably Prologue) that are just disabled everywhere by default for no apparent reason. Thankfully you can change that, unlike GTS where you're stuck with the Transmission mode playlist and no way to individually play each song. Why?

I'm not very knowledgeable about modern SIMD, so I wonder if the PS2's SIMD capabilities (including parallel & VU instructions) still hold up to this day.

Last week, a major milestone was achieved--a matching decompilation of one of the IOP modules. These are little programs sent to the IOP (I/O Processor) responsible for interacting with peripherals (DVD, pad, etc). It is, to my knowledge, the first ever matching decompilation for an IOP module.

poly@polyproxy.bsky.social · 2y ago

Over a year ago, I started a decompilation of PaRappa the Rapper 2 for the PlayStation 2, in the hopes of improving the state of modding and make it easier. It's progressing at a quick pace (~45% as of writing) - you can check it out here: github.com/polybiusprox... #パラッパラッパー #parappatherapper

Despite always reaching its target frame rate of 60 FPS, not all of PTR2's code is well optimized. This example shows how a small oversight can waste CPU time on double-precision FP arithmetic, something that could've been fixed by suffixing a few 'f's. 🙂

BildBild