kohugaly – Uploader Profile



Uploads:

Questions:

Answers:

  • Useful! (2014-10-11, 18:45): Yes, I know about that, but I also never used it. Every time I needed to shuffle it was always just contents of one register. By the way how do you dig up this stuff :-D the Cytos tools you posted on FS forum for example. I'm working on a topic about assembler in FS…
  • Very useful! (2014-11-11, 21:38): Thanks man! It was one of those ”why I haven’t came up with this earlier”-type of modules. It came to my mind when I was preparing schematics for the optimizations articles. I used similar approach to create multi-voice chorus effect (back in SM days, later revised when I learned ruby to manipulate midi a little).…
  • Very useful! (2014-11-11, 22:41): Yeah, I tried to make a test schematic, that has several (dozens) of synced modules. Since I wanted to have two groups (for A/B comparisons) they were hell to set up. After several rage-quits my multi-voice chorus came to my mind because I remember how it smashed CPU hard when many voices were setup… and…
  • other opcode (2014-11-19, 11:01): wow…. they really do… perhaps I should setup a section for ”semi-implemented” opcodes.
  • other opcode (2014-11-19, 11:27): cvtpd2ps works too and addpd and subpd doo too. mulpd doesn’t. This will require further testing.
  • other opcode (2014-11-19, 19:57): To be honest, I always thought the double precision operations were preformed via FPU instructions (fadd; fsub; fmul; frndint;) which might still be true for multiplication. perhaps there’s a way to declare double floats? and load them into FPU stack via ”fld”?
  • other opcode (2014-11-20, 15:19): Or precomputed in stage 0. In fact we can ”declare” doubles by converting singles to doubles in stage 0 like so float DF0p5=0.5; stage0; cvtps2pd xmm0,DF0p5; movaps DF0p5,xmm0; … To properly declare constants with full precision (like for pi), we need to sum 2 or even 3 single precision values, so DLL in that case…

Back to Top