Yes it is precise, if Y is constant and 1/Y is exact. In this case,
this transformation is conducted without "-fast-math".
What I meant is if 1/Y is not exact. I measured the results before.
Among all fast-math transformation,
reciprocal is most annoying one because it is very imprecise. I was
arguing at that time that reciprocal should be
ranked as the most relax xfrom. It seems the numerical guru Steve kinda
agree with me.
On 8/8/13 10:32 AM, Joerg Sonnenberger wrote:> On Thu, Aug 08, 2013 at 10:13:07AM -0700, Shuxin Yang wrote:
>> You should be very careful, reciprocal is very very very imprecise
>> transformation. Sometimes you will see big different
>> with and without this xform.
> What about constant %a, %b in this case? At least for power-of-two
> arguments, wouldn't it still be precise?
>
> Joerg
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev