[llvm-commits] [llvm] r145906 - in /llvm/trunk: lib/Transforms/Scalar/LoopStrengthReduce.cpp test/Transforms/LoopStrengthReduce/2011-12-04-loserreg.ll

[Andrew Trick]

On Mar 9, 2012, at 8:09 AM, NAKAMURA Takumi <geek4civic at gmail.com> wrote:

> 2011/12/6 Andrew Trick <atrick at apple.com>:
>> Author: atrick
>> Date: Mon Dec  5 21:13:31 2011
>> New Revision: 145906
>> 
>> URL: http://llvm.org/viewvc/llvm-project?rev=145906&view=rev
>> Log:
>> LSR: prune undesirable formulae early.
>> 
>> It's always good to prune early, but formulae that are unsatisfactory
>> in their own right need to be removed before running any other pruning
>> heuristics. We easily avoid generating such formulae, but we need them
>> as an intermediate basis for forming other good formulae.
> 
>> Added: llvm/trunk/test/Transforms/LoopStrengthReduce/2011-12-04-loserreg.ll
>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/LoopStrengthReduce/2011-12-04-loserreg.ll?rev=145906&view=auto
> 
>> +; RUN: llc < %s | FileCheck %s
>> +;
>> +; Test LSR's ability to prune formulae that refer to nonexistant
>> +; AddRecs in other loops.
>> +;
>> +; Unable to reduce this case further because it requires LSR to exceed
>> +; ComplexityLimit.
>> +;
>> +; We really just want to ensure that LSR can process this loop without
>> +; finding an unsatisfactory solution and bailing out. I've added
>> +; dummyout, an obvious candidate for postinc replacement so we can
>> +; verify that LSR removes it.
>> +
>> +target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
>> +target triple = "x86_64-apple-darwin"
> 
> Andy, this depends on TARGETS=x86.

Thanks. I was recently wondering if we had any intention of supporting make check without TARGETS=x86. I see 'target triple = "x86..."' scattered thoughout the common test cases.

What's your suggestion? 

-Andy