[llvm-commits] [llvm] r45230 - /llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp

[Chris Lattner]

On Dec 19, 2007, at 5:03 PM, Evan Cheng wrote:

> Sorry, the check in comment and some of the code is wrong. What it
> should say is "is all other uses are address and iv reuse for those
> are possible, then ignore the PHI use." In that case, an extra
> multiplication is needed (for PHI), but it eliminates an iv because
> the multiplication is folded into other uses. It turns out to be a
> good trade.
>
> I'll fix this in a bit.
>
Ok!

-Chris