Subzero: Add branch optimization.
1. Unconditional branch to the next basic block is removed.
2. For a conditional branch with a "false" edge to the next basic block, remove the unconditional branch to the fallthrough block.
3. For a conditional branch with a "true" edge to the next basic block, invert the condition and do like #2.
This is enabled only for O2, particularly because inverting the branch condition is a marginally risky operation.
This decreases the instruction count by about 5-6%.
Also, --stats prints a final tally to make it easier to post-process the output.
BUG= none
R=jvoung@chromium.org
Review URL: https://codereview.chromium.org/580903005
diff --git a/src/IceTargetLoweringX8632.cpp b/src/IceTargetLoweringX8632.cpp
index 464a2e8..31c11d4 100644
--- a/src/IceTargetLoweringX8632.cpp
+++ b/src/IceTargetLoweringX8632.cpp
@@ -395,6 +395,14 @@
T_genFrame.printElapsedUs(Context, "genFrame()");
Func->dump("After stack frame mapping");
+ // Branch optimization. This needs to be done just before code
+ // emission. In particular, no transformations that insert or
+ // reorder CfgNodes should be done after branch optimization. We go
+ // ahead and do it before nop insertion to reduce the amount of work
+ // needed for searching for opportunities.
+ Func->doBranchOpt();
+ Func->dump("After branch optimization");
+
// Nop insertion
if (shouldDoNopInsertion()) {
Func->doNopInsertion();
@@ -444,6 +452,13 @@
}
}
+bool TargetX8632::doBranchOpt(Inst *I, const CfgNode *NextNode) {
+ if (InstX8632Br *Br = llvm::dyn_cast<InstX8632Br>(I)) {
+ return Br->optimizeBranch(NextNode);
+ }
+ return false;
+}
+
IceString TargetX8632::RegNames[] = {
#define X(val, init, name, name16, name8, scratch, preserved, stackptr, \
frameptr, isI8, isInt, isFP) \
