Be more aggressive about tail-merging small blocks
if those blocks consist entirely of common instructions;
merging will not add an extra branch in this case.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51006 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/lib/CodeGen/BranchFolding.cpp b/lib/CodeGen/BranchFolding.cpp
index a8fee1c..1807f98 100644
--- a/lib/CodeGen/BranchFolding.cpp
+++ b/lib/CodeGen/BranchFolding.cpp
@@ -495,7 +495,18 @@
CurMPIter->second,
I->second,
TrialBBI1, TrialBBI2);
- if (CommonTailLen >= minCommonTailLength) {
+ // If we will have to split a block, there should be at least
+ // minCommonTailLength instructions in common; if not, at worst
+ // we will be replacing a fallthrough into the common tail with a
+ // branch, which at worst breaks even with falling through into
+ // the duplicated common tail, so 1 instruction in common is enough.
+ // We will always pick a block we do not have to split as the common
+ // tail if there is one.
+ // (Empty blocks will get forwarded and need not be considered.)
+ if (CommonTailLen >= minCommonTailLength ||
+ (CommonTailLen > 0 &&
+ (TrialBBI1==CurMPIter->second->begin() ||
+ TrialBBI2==I->second->begin()))) {
if (CommonTailLen > maxCommonTailLength) {
SameTails.clear();
maxCommonTailLength = CommonTailLen;