Joshua Isom wrote:
I guess this is one place where being CISC really is better than being
RISC.
It really depends on the hardare you are running. E.g.
add I0, I1, 20
translates either to something like:
lea %ecx, 20(%edx) # not yet but in case ..
or
ori r11, r31, 20 # r31 is 0
add r3, r4, r11 # a 2nd oris is needed for constants > 0xffff
If x86 were not the leading instruction set, I'd toss the opcode
variants with constants almost all.
... But how much improvement with outputting to a pbc first?
If you are gonna running the code multiple times, you save a few cycles
compilation time. JIT code is still created at runtime.
... But a
couple notes, there's no --help-optimize like --help-debug, and as far
as I know,
perldoc docs/runnning.pod - but these optimizations aren't settled nor
finished. You can ignore them for now.
there's no way to disable optimizations completely, e.g. this
pir
.sub main :main
$N0 = pow 2.0, 5.0
.end
Is always converted to this.
main:
set N0, 32
end
This isn't an optimization. The 'pow' opcode is just run at compiletime.
Which can lead to misleading test results for when pow's actually broken.
If 'pow' is broken then it's borken. The only case where that would
matter is, when you are compiling a PBC on a machine with a broken 'pow'
and ship that PBC.
leo
