>>>>> "Martin" == Martin Cracauer <[EMAIL PROTECTED]> writes:
Martin> Looking over the spec I also find that there is no way to detect
Martin> overflow. Not that big a deal, but still breaking some of the Lisp
Martin> spirit.
MMX can do basic arithmetic operations on 64-bit integers? It's
probably not a problem that overflow isn't detected. At least on
sparc, we never depend on overflow indicators. If the result won't
fit in the desired register, the vop isn't used.
Martin> I am also not quite sure it is that much faster than using the x86
Martin> instructions to do gcc-like 64 bit integer arthmetic on two 32 bits
Martin> registers. Granted, this way trashes registers like hell, but
Martin> depending on your code and the processor in use the trashing of
Martin> floating point state may be as bad or worse.
This, I think, would be easier since the complex floats use pairs of
registers to hold them. Might be a problem on the register-starved
x86 though. Probably also a problem on sparc since there are only a
few non-descriptor registers to hold these 32-bit numbers.
Ray
