Displaying 2 results from an estimated 2 matches for "collectfixupsandinitialformulae".
2015 May 09
4
[LLVMdev] [LSR] hoisting loop invariants in reverse order
...+ 3) * c;
SLSR would optimize them to
int arg0 = a * b * c;
int arg1 = arg0 + c;
int arg2 = arg1 + c;
int arg3 = arg2 + c;
// 2 muls and 3 adds/subs. 1 add/sub less than with the reversed order
I have a proof-of-concept patch (
http://reviews.llvm.org/differential/diff/25402/) that has
CollectFixupsAndInitialFormulae to sort initial formulae in a dominance
order (i.e. if A.getUser() dominates B.getUser(), then we put A before B).
It breaks some tests that are too sensitive to order changes; besides that,
I don't see significant issues.
Jingyue
-------------- next part --------------
An HTML attachment was...
2015 May 18
2
[LLVMdev] [LSR] hoisting loop invariants in reverse order
...t arg1 = arg0 + c;
> > int arg2 = arg1 + c;
> > int arg3 = arg2 + c;
> > // 2 muls and 3 adds/subs. 1 add/sub less than with the reversed order
> >
> > I have a proof-of-concept patch
> > (http://reviews.llvm.org/differential/diff/25402/) that has
> > CollectFixupsAndInitialFormulae to sort initial formulae in a dominance
> > order (i.e. if A.getUser() dominates B.getUser(), then we put A before
> B).
>
> Is there a reason to not just put it back in the original order?
> IE is dominance order better?
>
-------------- next part --------------
An HTML attach...