Oh I see! Basically "don't try to be smarter than GCC"? :D So does this mean that a "fast_sqrt" function taking 9 cycles should be written as "return x*1.0f/__builtin_sqrt(x)" (or in 2 different steps maybe) and let gcc take care of it? (And should it be written in KOS ...

Thanks for confirming, I didn't know the latency of each operand! (Although it's odd, since Moop's source indicates that their code was already 3x faster in practice than fsqrt? What am I missing?)

Hi everyone, Somebody mentioned this lib on Twitter and it is very cool! (I'm not sure if its author is still around? Also let me know if I should have created a new topic instead of replying here.) If I'm not mistaken, the MATH_Fast_Sqrt function can be optimized further using a small trick, which ...

Ah, thank you so much! I'll see if it's something I want to pursue further ; might be time to explore another route, from the sounds of it.

That looks like it is a file that should have been part of the FrotzDC code, not part of KOS (the fact that it is in #include'd with quotes and not angle brackets is kinda the giveaway there). Sorry, I just rephrased my error message cause it's in French, but it is indeed in angle brackets : src/dc...

Hello everyone! I'm very new to development for the Dreamcast, so apologies if my question is trivial :) I'm trying to compile FrotzDC (https://github.com/c99koder/FrotzDC/), which hasn't been touched in 15 years... I installed KOS successfully, and the compilation of FrotzDC stops when compiling &q...