Re:Explaining how the MASM 2-pass assembler works
by zh3 on Sunday September 12, 2021 @02:14PM
(#28503436)
Using a 'JMP SHORT' was pretty much reflexive; a good reason to do it was it not being in range was sort of a code smell back in the day.
But chips were simpler back then, and a programmer could often leave compiler code in the dust with small bits of assembler (esp. where Fortran was involved).
Nowadays, X86 has become so hard to reason about from a performance point of view that for critical code I just assume that GCC will emit something acceptable given a bit of sensibly-written C, and that to get a big boost beyond that is going to involves GPUs or custom silicon (depending on the application).
But chips were simpler back then, and a programmer could often leave compiler code in the dust with small bits of assembler (esp. where Fortran was involved).
Nowadays, X86 has become so hard to reason about from a performance point of view that for critical code I just assume that GCC will emit something acceptable given a bit of sensibly-written C, and that to get a big boost beyond that is going to involves GPUs or custom silicon (depending on the application).