[PATCH] Use invariants in LazyValueInfo

[Philip Reames]

Worth noting is that your "small addition to JumpThreading" is going to 
be a general maintenance burden imposed by your approach.  We'll have to 
make other such "small additions" throughout the optimizer when 
recognizing control flow patterns.

We should probably think about some form of placement canonicalization 
for them to reduce this burden.

Philip

On 07/14/2014 11:46 PM, hfinkel at anl.gov wrote:
> Hi chandlerc,
>
> This patch teaches LazyValueInfo to use the invariant intrinsic. Like with the known-bits patch, this requires feeding a "context" instruction pointer through many functions. Aside from a little refactoring to reuse the logic that turns predicates into constant ranges in LVI, the only new code is that which can 'merge' the range from an invariant into that otherwise computed. There is also a small addition to JumpThreading so that it can have LVI use invariants in the same block as the comparison feeding a conditional branch.
>
> With this patch, we can now simplify this as expected:
> int foo(int a) {
>    __builtin_assume(a > 5);
>    if (a > 3) {
>      bar();
>      return 1;
>    }
>    return 0;
> }
>
> http://reviews.llvm.org/D4511
>
> Files:
>    include/llvm/Analysis/LazyValueInfo.h
>    lib/Analysis/LazyValueInfo.cpp
>    lib/Transforms/Scalar/CorrelatedValuePropagation.cpp
>    lib/Transforms/Scalar/JumpThreading.cpp
>    test/Transforms/JumpThreading/invariants.ll
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140716/27530d43/attachment.html>