R: [Algorithms] non-branching zbuf code tricks?

[Giovanni B. <ba...@pr...>]

> > > >zbuf[x] = z < zbuf[x] ? z : zbuf[x];

[....]

> No!  There is a significant difference.  In my code, I gave
> the compile permission to evaluate both 'z' and 'zbuf[x]',
> in the '?:' version, you are telling the compiler that it's
> only ALLOWED to evaluate one or the other.  Hence, in the '?:'
> case, the compiler is logically required to generate a branch
> but in my version, it *could* generate a SETxx sequence to
> evaluate the conditional.

Thinking well about it, in the expression above you are already forcing the
compiler to evaluate both 'z' and 'zbuf[x]' just to executation the compare.
So, the expression above *could* be assembled into a CMOV, just like GCC
does on MIPS with MOVN, as Christer showed us.

---
Giovanni Bajo
Lead Programmer

Protonic Interactive
www.protonic.net

- Black holes are generated when God divides by zero -