On Dec 19, 2007, at 5:15 PM, Chris Lattner wrote:
On Dec 19, 2007, at 5:10 PM, Chris Lattner wrote:
Finally, is "m" really required to be a constant? This xform seems
valid for any "m" and "A", and more general is good.
Incidentally, in the spirit of generalizing this, it seems like these
are also valid forms of the same thing:
(icmp lt (sub a, b), 1)
(icmp sgt (sub a, b), -1)
Instcombine will canonicalize (sle x, 0) -> (slt x, 1) and (sge x,
0) -> (sgt x, -1), so these extra cases are worth handling.
I just noticed that the CFE is fond of producing these. I agree
they're worth having.
--
Christopher Lamb
_______________________________________________
llvm-commits mailing list
llvm-commits@cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
