Do trivial CSE of dead BBs during codegen preparation.

Some BBs can become dead after codegen preparation. If we delete them here, it
could help enable tail-call optimizations later on.
<rdar://problem/10256573>


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@152002 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Bill Wendling 2012-03-04 10:46:01 +00:00
parent e4fd907e72
commit e3e394d982
5 changed files with 25 additions and 6 deletions

View File

@ -65,6 +65,11 @@ static cl::opt<bool> DisableBranchOpts(
"disable-cgp-branch-opts", cl::Hidden, cl::init(false),
cl::desc("Disable branch optimizations in CodeGenPrepare"));
// FIXME: Remove this abomination once all of the tests pass without it!
static cl::opt<bool> DisableDeleteDeadBlocks(
"disable-cgp-delete-dead-blocks", cl::Hidden, cl::init(false),
cl::desc("Disable deleting dead blocks in CodeGenPrepare"));
namespace {
class CodeGenPrepare : public FunctionPass {
/// TLI - Keep a pointer of a TargetLowering to consult for determining
@ -160,8 +165,22 @@ bool CodeGenPrepare::runOnFunction(Function &F) {
if (!DisableBranchOpts) {
MadeChange = false;
for (Function::iterator BB = F.begin(), E = F.end(); BB != E; ++BB)
SmallPtrSet<BasicBlock*, 8> WorkList;
for (Function::iterator BB = F.begin(), E = F.end(); BB != E; ++BB) {
SmallVector<BasicBlock*, 2> Successors(succ_begin(BB), succ_end(BB));
MadeChange |= ConstantFoldTerminator(BB, true);
if (!MadeChange) continue;
for (SmallVectorImpl<BasicBlock*>::iterator
II = Successors.begin(), IE = Successors.end(); II != IE; ++II)
if (pred_begin(*II) == pred_end(*II))
WorkList.insert(*II);
}
if (!DisableDeleteDeadBlocks)
for (SmallPtrSet<BasicBlock*, 8>::iterator
I = WorkList.begin(), E = WorkList.end(); I != E; ++I)
DeleteDeadBlock(*I);
if (MadeChange)
ModifiedDT = true;

View File

@ -1,4 +1,4 @@
; RUN: llc < %s -mtriple=thumbv7-apple-darwin10 -relocation-model=pic -disable-fp-elim -mcpu=cortex-a8 | FileCheck %s
; RUN: llc < %s -mtriple=thumbv7-apple-darwin10 -relocation-model=pic -disable-fp-elim -disable-cgp-delete-dead-blocks -mcpu=cortex-a8 | FileCheck %s
; Do not form Thumb2 ldrd / strd if the offset is not multiple of 4.
; rdar://9133587

View File

@ -4,7 +4,7 @@ target triple = "i386-apple-darwin8"
; Without CSE of libcalls, there are two calls in the output instead of one.
define i32 @u_f_nonbon(double %lambda) nounwind {
define double @u_f_nonbon(double %lambda) nounwind {
entry:
%tmp19.i.i = load double* null, align 4 ; <double> [#uses=2]
%tmp6.i = fcmp olt double %tmp19.i.i, 1.000000e+00 ; <i1> [#uses=1]
@ -26,5 +26,5 @@ bb502.loopexit.i: ; preds = %bb28.i
br i1 false, label %bb.nph53.i, label %bb508.i
bb508.i: ; preds = %bb502.loopexit.i, %entry
ret i32 1
ret double %tmp10.i4
}

View File

@ -1,4 +1,4 @@
; RUN: llc < %s -verify-regalloc | FileCheck %s
; RUN: llc < %s -disable-cgp-delete-dead-blocks -verify-regalloc | FileCheck %s
; PR11347
;
; This test case materializes the constant 1 in a register:

View File

@ -1,4 +1,4 @@
; RUN: llc < %s | FileCheck %s
; RUN: llc < %s -disable-cgp-delete-dead-blocks | 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-n8:16:32:64"
target triple = "x86_64-apple-darwin10.0.0"