Chris Lattner 1c7efba2bd When converting phi nodes into select instructions, we shouldn't promote PHI ...

nodes unless we KNOW that we are able to promote all of them.  In this case
promoting the phi to a select is silly because we will always have to do the
call conditionally.  As such, select promotion is actually a pessimization.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@16972 91177308-0d34-0410-b5e6-96231b3b80d8

2004-10-14 05:12:50 +00:00

..

2002-05-05-EmptyBlockMerge.ll

Fix a test that was "broken" by new optimizations. The transformation we

2004-05-26 23:51:29 +00:00

2002-05-21-PHIElimination.ll

Renamed as' => llvm-as', dis' => llvm-dis', link' => llvm-link'.

2003-09-16 15:29:54 +00:00

2002-06-24-PHINode.ll

Renamed as' => llvm-as', dis' => llvm-dis', link' => llvm-link'.

2003-09-16 15:29:54 +00:00

2002-09-24-PHIAssertion.ll

Renamed as' => llvm-as', dis' => llvm-dis', link' => llvm-link'.

2003-09-16 15:29:54 +00:00

2003-03-07-DominateProblem.ll

Renamed as' => llvm-as', dis' => llvm-dis', link' => llvm-link'.

2003-09-16 15:29:54 +00:00

2003-08-05-InvokeCrash.ll

Renamed as' => llvm-as', dis' => llvm-dis', link' => llvm-link'.

2003-09-16 15:29:54 +00:00

2003-08-05-MishandleInvoke.ll

Renamed as' => llvm-as', dis' => llvm-dis', link' => llvm-link'.

2003-09-16 15:29:54 +00:00

2003-08-17-BranchFold.ll

Renamed as' => llvm-as', dis' => llvm-dis', link' => llvm-link'.

2003-09-16 15:29:54 +00:00

2003-08-17-BranchFoldOrdering.ll

Renamed as' => llvm-as', dis' => llvm-dis', link' => llvm-link'.

2003-09-16 15:29:54 +00:00

2003-08-17-FoldSwitch.ll

Renamed as' => llvm-as', dis' => llvm-dis', link' => llvm-link'.

2003-09-16 15:29:54 +00:00

basictest.ll

Renamed as' => llvm-as', dis' => llvm-dis', link' => llvm-link'.

2003-09-16 15:29:54 +00:00

branch-cond-merge.ll

New testcase:

2004-05-01 23:35:21 +00:00

branch-cond-prop.ll

New testcase for case that simplifycfg should catch

2004-05-01 22:36:24 +00:00

branch-fold-test.ll

Renamed as' => llvm-as', dis' => llvm-dis', link' => llvm-link'.

2003-09-16 15:29:54 +00:00

BrUnwind.ll

test that unconditional branches to unwinds are always eliminated

2004-07-20 01:17:19 +00:00

EqualPHIEdgeBlockMerge.ll

Renamed as' => llvm-as', dis' => llvm-dis', link' => llvm-link'.

2003-09-16 15:29:54 +00:00

InvokeEliminate.ll

Fix testcase. For some reason this was failing only with a release build and

2004-07-01 04:09:14 +00:00

PhiBlockMerge2.ll

Renamed as' => llvm-as', dis' => llvm-dis', link' => llvm-link'.

2003-09-16 15:29:54 +00:00

PhiBlockMerge.ll

Renamed as' => llvm-as', dis' => llvm-dis', link' => llvm-link'.

2003-09-16 15:29:54 +00:00

PhiEliminate2.ll

Simplify cfg now eliminates all of the branches in this testcase

2004-04-09 23:50:29 +00:00

PhiEliminate.ll

Oops, actually USE the previously computed value

2004-03-30 19:45:39 +00:00

PhiNoEliminate.ll

When converting phi nodes into select instructions, we shouldn't promote PHI

2004-10-14 05:12:50 +00:00

return-merge.ll

New testcase

2004-04-02 18:12:49 +00:00

switch_create.ll

The simplifycfg pass should be able to turn stuff like:

2004-02-24 05:34:44 +00:00

switch_formation.ll

Add a testcase for switch instruction induction

2004-02-28 21:30:39 +00:00

switch_switch_fold.ll

New testcase. Switch instructions that go to switch instructions should be

2004-02-24 07:21:09 +00:00

UncondBranchToReturn.ll

New testcase, details in the comments

2004-02-16 06:35:19 +00:00