[llvm-bugs] [Bug 31468] New: NewGVN doesn't merge unconditional branches.

via llvm-bugs llvm-bugs at lists.llvm.org
Mon Dec 26 05:07:13 PST 2016


https://llvm.org/bugs/show_bug.cgi?id=31468

            Bug ID: 31468
           Summary: NewGVN doesn't merge unconditional branches.
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: All
            Status: NEW
          Severity: normal
          Priority: P
         Component: Scalar Optimizations
          Assignee: davide at freebsd.org
          Reporter: davide at freebsd.org
                CC: dberlin at dberlin.org, llvm-bugs at lists.llvm.org
            Blocks: 30995
    Classification: Unclassified

The code living in GVN.cpp claims this is done to expose "more PRE
opportunities". Fair enough, but I'm not entirely sure this is relevant anymore
in NewGVN as we're writing a real PRE infrastructure. 

Fixing is relatively trivial (patch inline), but not sure if really needed.
Dan, what you think? If we don't want to change this code, I'll modify the
failing test and call it a day (basic.ll)

diff --git a/lib/Transforms/Scalar/NewGVN.cpp
b/lib/Transforms/Scalar/NewGVN.cpp
index 1f12d10..430385a 100644
--- a/lib/Transforms/Scalar/NewGVN.cpp
+++ b/lib/Transforms/Scalar/NewGVN.cpp
@@ -1378,6 +1378,20 @@ bool NewGVN::runGVN(Function &F, DominatorTree *_DT,
AssumptionCache *_AC,
   unsigned ICount = 0;
   SmallPtrSet<BasicBlock *, 16> VisitedBlocks;

+  // Do a sweep over all the basic blocks to merge unconditional branches.
+  for (Function::iterator FI = F.begin(), FE = F.end(); FI != FE;) {
+    BasicBlock *BB = &*FI++;
+
+    // In theory here we could pass MemSSA if/when MergeBlockIntoPredecessor
+    // will grow version which accepts the analysis.
+    bool removedBlock = MergeBlockIntoPredecessor(
+        BB, DT, nullptr /* LoopInfo */, nullptr /* MemDep */);
+
+    if (removedBlock)
+      NumGVNBlocksDeleted++;
+    Changed |= removedBlock;
+  }
+
   // Note: We want RPO traversal of the blocks, which is not quite the same as
   // dominator tree order, particularly with regard whether backedges get
   // visited first or second, given a block with multiple successors.
diff --git a/test/Transforms/NewGVN/basic.ll b/test/Transforms/NewGVN/basic.ll
index f6670aa..f68823a 100644
--- a/test/Transforms/NewGVN/basic.ll
+++ b/test/Transforms/NewGVN/basic.ll
@@ -1,4 +1,3 @@
-; XFAIL: *
 ; RUN: opt < %s -newgvn -S | FileCheck %s

 define i32 @main() {

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20161226/df77a16a/attachment.html>


More information about the llvm-bugs mailing list