Re: NASM 2.00 is Here!!!

"Wolfgang Kern" <spamtrap@xxxxxxxxxx> écrivait
news:fip7ud$hkl$1@xxxxxxxxxxxxxxxxxxxxx:

KESYS also got the opportunity for the user to define a 16 bit float
with one byte integer and one byte exponent, but really rare used ...

I think this 'byte-floats' are just fix-point interpreted values.
I've seen a few examples where factors below '1' (ie:sin/cos) were
used. And as a graphic screen will hardly exceed a 16-bit range, most
vector calculation can be reduced to a couple of MUL/SHIFT pairs.

SIN look-up-tables used for cheap jump&run games may contain just byte
values either signed or unsigned, representing 1/128 or 1/256 steps.
Also seen a fast, short but dirty hack with a one byte range for a
whole 2*pi circle [01h=1.40; 40h=90; 80h=180; c0h= 270 ffh=358.59
/degrees]

KESYS native graphics uses only one 16 bit signed SIN/COS-LUT to draw
circles,cake-lines,lines,rectangles and bitmaps in any 2D/3D aspect
and the images look exact enough even if displayed on a 3200*2400
screen. But I still hunt for GPU-docs to once speed this up a bit :)

OK, but you don't say how you *use* this. If you implemented
it for data-space saving, you have, so to say, to "unfold"
the Data before passing to the FPU. Is it like this?


Betov.

< http://rosasm.org >


.