You never stop learning, I just explored how I can force a compiler into producing the sbb op on x86. It’s quite difficult actually, most of them produce for example cmp + seta + neg for the term -(a > b). In fact the only deterministic behaviour is with a very special ternery term (a [...]
Havn’t found this anywhere, so I had to do it myself. The title says it all, so here it is without drama:
// mm7 = 0×8000 0×8000 0×8000 0×8000
// mm3 = 0×7FFF 0xFFFF 0×7FFF 0xFFFF
// mm5 = 0×7FFF 0×7FFF 0xFFFF 0xFFFF
movq mm6, mm7
pand mm7, mm5 /* b ? 0 | 0×8000 */
pand mm6, mm3 /* a ? 0 | 0×8000 */
pxor mm5, [...]
Ei, once in a while you ask yourself about the (in)sanity of the X86-architecture. Well everybody knows, the ops offered by MMX upto SSE58942 are actually choosen by dicing 42 out of 255 possible once. Well I just don’t go too deep into emotional depression here. Just my list of missing and missguided ops:
pmulbw
pmulsdq
pmulhd/pmulld
paddsd/psubsd/paddusd/psubusd
pmaddbw
pmadddq
pmaddubw/pmadduwd/pmaddudq
psadwd (horizontal [...]
While I was implementing several blending functions like the ‘over’-operator, I again was confronted with divisions by constants at range-boundaries. Like 0xFF for ‘normalizing’ unsigned char multiplications, and the equivalent 0xFFFF and 0xFFFFFFFF for unsigned short and unsigned long respectively. Most of us are familiar with Blinn’s approximation, which alone gave him a star in [...]
| M | T | W | T | F | S | S |
|---|---|---|---|---|---|---|
| « Aug | ||||||
| 1 | 2 | 3 | 4 | 5 | ||
| 6 | 7 | 8 | 9 | 10 | 11 | 12 |
| 13 | 14 | 15 | 16 | 17 | 18 | 19 |
| 20 | 21 | 22 | 23 | 24 | 25 | 26 |
| 27 | 28 | 29 | ||||