Re: How to eliminate redundant constant move instructions

On Nov 2, 6:35 am, glen herrmannsfeldt <gah@ugcs.caltech.edu> wrote:
> That is what register renaming is for. Usually using more than
> the architecturally specified number of registers, the CPU
> internally remaps the registers such that it can keep one value
> in a register while an instruction is being executed out of order.


For this specific case, I previous intention was optimizing codes


rx <- 0
...
use rx
...
ry <- 0
use ry


into


rx <- 0
...
use rx
...
use rx


In this manner, I guess the register renaming won't help,
since the transformation introduces true dependency, right?




> Now, how to choose the > weights when one doesn't know the specific
target processor? > That is a good question.


How could this happen if middle end does not know the specific
processor? What I can image is there should be a way in which
back end provides information.


Thanks

Return to the comp.compilers page.
Search the comp.compilers archives again.