Switch the select to branch transformation on by default.

The primitive conservative heuristic seems to give a slight overall
improvement while not regressing stuff. Make it available to wider
testing. If you notice any speed regressions (or significant code
size regressions) let me know!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@156258 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Benjamin Kramer 2012-05-06 14:25:16 +00:00
parent 58c1da84f2
commit 77c4ef8a47
3 changed files with 6 additions and 5 deletions

View File

@ -71,8 +71,9 @@ static cl::opt<bool> DisableDeleteDeadBlocks(
"disable-cgp-delete-dead-blocks", cl::Hidden, cl::init(false),
cl::desc("Disable deleting dead blocks in CodeGenPrepare"));
static cl::opt<bool> EnableSelectToBranch("enable-cgp-select2branch", cl::Hidden,
cl::desc("Enable select to branch conversion."));
static cl::opt<bool> DisableSelectToBranch(
"disable-cgp-select2branch", cl::Hidden, cl::init(false),
cl::desc("Disable select to branch conversion."));
namespace {
class CodeGenPrepare : public FunctionPass {
@ -1132,7 +1133,7 @@ static bool isFormingBranchFromSelectProfitable(SelectInst *SI) {
bool CodeGenPrepare::OptimizeSelectInst(SelectInst *SI) {
// If we have a SelectInst that will likely profit from branch prediction,
// turn it into a branch.
if (!EnableSelectToBranch || OptSize || !TLI->isPredictableSelectExpensive())
if (DisableSelectToBranch || OptSize || !TLI->isPredictableSelectExpensive())
return false;
if (!SI->getCondition()->getType()->isIntegerTy(1) ||

View File

@ -1,4 +1,4 @@
; RUN: llc -march=x86-64 -mcpu=core2 -enable-cgp-select2branch < %s | FileCheck %s
; RUN: llc -march=x86-64 -mcpu=core2 < %s | FileCheck %s
; cmp with single-use load, should not form cmov.
define i32 @test1(double %a, double* nocapture %b, i32 %x, i32 %y) {

View File

@ -1,4 +1,4 @@
; RUN: llc < %s -mtriple=x86_64-apple-darwin10 | FileCheck %s
; RUN: llc < %s -mtriple=x86_64-apple-darwin10 -disable-cgp-select2branch | FileCheck %s
target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128"
define i32 @test1(i32 %x, i32 %n, i32 %w, i32* %vp) nounwind readnone {