[LLVMdev] Floating-point range checks

[Robison, Arch]

I'll submit my SimplifyFCmpInst improvement to llvm-commits after the tests finish running.

So it doesn't get lost, attached is what I had in mind for a basic FPRange.h .  The name "FPRange" isn't really quite right since it really represents a lattice point.  It's quite coarse-grain, but cheap to compute and hits on the basic properties of IEEE arithmetic.  The public interface hides the coarseness, so it could be internally extended to be more precise without affecting clients.

- Arch

-----Original Message-----
From: Hal Finkel [mailto:hfinkel at anl.gov] 
Sent: Tuesday, January 13, 2015 1:43 PM
To: Pete Cooper
Cc: Philip Reames; llvmdev at cs.uiuc.edu; Robison, Arch
Subject: Re: [LLVMdev] Floating-point range checks

...

I don't want your work on an FPRange class to get lost, however, because we do need something like that generally speaking.

 -Hal


-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: FPRange.cpp
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20150114/1275dcde/attachment.ksh>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: FPRange.h
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20150114/1275dcde/attachment.h>