[MergeICmps] Handle chains with several complex BCE basic blocks.

 - Fix condition for detecting that a complex basic block was the first in
   the chain.
 - Add tests.

This was caught by buildbots when submitting rL324319.

llvm-svn: 324341

diff --git a/llvm/lib/Transforms/Scalar/MergeICmps.cpp b/llvm/lib/Transforms/Scalar/MergeICmps.cpp
index f045b62..745d29d 100644 (file)
--- a/llvm/lib/Transforms/Scalar/MergeICmps.cpp
+++ b/llvm/lib/Transforms/Scalar/MergeICmps.cpp
@@ -278,7 +278,8 @@ BCECmpChain::BCECmpChain(const std::vector<BasicBlock *> &Blocks, PHINode &Phi)
   assert(!Blocks.empty() && "a chain should have at least one block");
   // Now look inside blocks to check for BCE comparisons.
   std::vector<BCECmpBlock> Comparisons;
-  for (BasicBlock *Block : Blocks) {
+  for (size_t BlockIdx = 0; BlockIdx < Blocks.size(); ++BlockIdx) {
+    BasicBlock *const Block = Blocks[BlockIdx];
     assert(Block && "invalid block");
     BCECmpBlock Comparison = visitCmpBlock(Phi.getIncomingValueForBlock(Block),
                                            Block, Phi.getParent());
@@ -289,7 +290,7 @@ BCECmpChain::BCECmpChain(const std::vector<BasicBlock *> &Blocks, PHINode &Phi)
     }
     if (Comparison.doesOtherWork()) {
       DEBUG(dbgs() << "block does extra work besides compare\n");
-      if (Comparisons.empty()) {  // First block.
+      if (BlockIdx == 0) {  // First block.
         // TODO(courbet): The first block can do other things, and we should
         // split them apart in a separate block before the comparison chain.
         // Right now we just discard it and make the chain shorter.
@@ -330,7 +331,7 @@ BCECmpChain::BCECmpChain(const std::vector<BasicBlock *> &Blocks, PHINode &Phi)
     DEBUG(dbgs() << "\n");
     Comparisons.push_back(Comparison);
   }
-  assert(!Comparisons.empty() && "chain with a single complex basic block");
+  assert(!Comparisons.empty() && "chain with no BCE basic blocks");
   EntryBlock_ = Comparisons[0].BB;
   Comparisons_ = std::move(Comparisons);
 #ifdef MERGEICMPS_DOT_ON

diff --git a/llvm/test/Transforms/MergeICmps/X86/two-complex-bb.ll b/llvm/test/Transforms/MergeICmps/X86/two-complex-bb.ll
new file mode 100644 (file)
index 0000000..945a2d0
--- /dev/null
+++ b/llvm/test/Transforms/MergeICmps/X86/two-complex-bb.ll
@@ -0,0 +1,58 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -mergeicmps -mtriple=x86_64-unknown-unknown -S | FileCheck %s --check-prefix=X86
+
+%"struct.std::pair" = type { i32, i32 }
+
+; This tests a function with two complex basic blocks.
+define zeroext i1 @twocomplexblocks(
+; X86-LABEL: @twocomplexblocks(
+; X86-NEXT:  entry:
+; X86-NEXT:    [[FIRST_I:%.*]] = getelementptr inbounds %"struct.std::pair", %"struct.std::pair"* [[A:%.*]], i64 0, i32 0
+; X86-NEXT:    [[TMP0:%.*]] = load i32, i32* [[FIRST_I]], align 4
+; X86-NEXT:    [[FIRST1_I:%.*]] = getelementptr inbounds %"struct.std::pair", %"struct.std::pair"* [[B:%.*]], i64 0, i32 0
+; X86-NEXT:    [[TMP1:%.*]] = load i32, i32* [[FIRST1_I]], align 4
+; X86-NEXT:    [[EXTRAWORK:%.*]] = add i32 [[TMP0]], [[TMP1]]
+; X86-NEXT:    [[CMP_I:%.*]] = icmp eq i32 [[TMP0]], [[TMP1]]
+; X86-NEXT:    br i1 [[CMP_I]], label [[LAND_RHS_I:%.*]], label [[OPEQ1_EXIT:%.*]]
+; X86:       land.rhs.i:
+; X86-NEXT:    [[SECOND_I:%.*]] = getelementptr inbounds %"struct.std::pair", %"struct.std::pair"* [[A]], i64 0, i32 1
+; X86-NEXT:    [[TMP2:%.*]] = load i32, i32* [[SECOND_I]], align 4
+; X86-NEXT:    [[SECOND2_I:%.*]] = getelementptr inbounds %"struct.std::pair", %"struct.std::pair"* [[B]], i64 0, i32 1
+; X86-NEXT:    [[TMP3:%.*]] = load i32, i32* [[SECOND2_I]], align 4
+; X86-NEXT:    [[EXTRAWORK2:%.*]] = add i32 [[TMP2]], [[TMP3]]
+; X86-NEXT:    [[CMP3_I:%.*]] = icmp eq i32 [[TMP2]], [[TMP3]]
+; X86-NEXT:    br label [[OPEQ1_EXIT]]
+; X86:       opeq1.exit:
+; X86-NEXT:    [[TMP4:%.*]] = phi i1 [ false, [[ENTRY:%.*]] ], [ [[CMP3_I]], [[LAND_RHS_I]] ]
+; X86-NEXT:    ret i1 [[TMP4]]
+;
+  %"struct.std::pair"* nocapture readonly dereferenceable(8) %a,
+  %"struct.std::pair"* nocapture readonly dereferenceable(8) %b) local_unnamed_addr #0 {
+entry:
+  ; This is a complex BCE Basic Block.
+  %first.i = getelementptr inbounds %"struct.std::pair", %"struct.std::pair"* %a, i64 0, i32 0
+  %0 = load i32, i32* %first.i, align 4
+  %first1.i = getelementptr inbounds %"struct.std::pair", %"struct.std::pair"* %b, i64 0, i32 0
+  %1 = load i32, i32* %first1.i, align 4
+  %extrawork = add i32 %0, %1
+  %cmp.i = icmp eq i32 %0, %1
+  br i1 %cmp.i, label %land.rhs.i, label %opeq1.exit
+
+land.rhs.i:
+  ; This is a complex BCE Basic Block.
+  %second.i = getelementptr inbounds %"struct.std::pair", %"struct.std::pair"* %a, i64 0, i32 1
+  %2 = load i32, i32* %second.i, align 4
+  %second2.i = getelementptr inbounds %"struct.std::pair", %"struct.std::pair"* %b, i64 0, i32 1
+  %3 = load i32, i32* %second2.i, align 4
+  %extrawork2 = add i32 %2, %3
+  %cmp3.i = icmp eq i32 %2, %3
+  br label %opeq1.exit
+
+opeq1.exit:
+  %4 = phi i1 [ false, %entry ], [ %cmp3.i, %land.rhs.i ]
+  ret i1 %4
+}
+
+
+
+