[WebAssembly] Disable folding of GA+reg into load/store constant offsets

Summary:
If the register has a negative value then unsigned overflow will occur;
this case is sometimes even created intentionally by LSR. For now
disable GA+reg folding. Fixes PR29127

Differential Revision: https://reviews.llvm.org/D24053

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@280285 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Derek Schuff 2016-08-31 20:27:20 +00:00
parent ebc9efbb36
commit 0d4df2a624
3 changed files with 91 additions and 46 deletions

View File

@ -41,15 +41,13 @@ def or_is_add : PatFrag<(ops node:$lhs, node:$rhs), (or node:$lhs, node:$rhs),[{
}]>;
// GlobalAddresses are conceptually unsigned values, so we can also fold them
// into immediate values as long as their offsets are non-negative.
// into immediate values as long as the add is 'nuw'.
// TODO: We'd like to also match GA offsets but there are cases where the
// register can have a negative value. Find out what more we can do.
def regPlusGA : PatFrag<(ops node:$addr, node:$off),
(add node:$addr, node:$off),
[{
return N->getFlags()->hasNoUnsignedWrap() ||
(N->getOperand(1)->getOpcode() == WebAssemblyISD::Wrapper &&
isa<GlobalAddressSDNode>(N->getOperand(1)->getOperand(0)) &&
cast<GlobalAddressSDNode>(N->getOperand(1)->getOperand(0))
->getOffset() >= 0);
return N->getFlags()->hasNoUnsignedWrap();
}]>;
// We don't need a regPlusES because external symbols never have constant

View File

@ -28,13 +28,17 @@ define i32 @load_test0_noinbounds() {
ret i32 %t
}
; TODO: load_test1 - load_test8 are disabled because folding GA+reg is disabled
; (there are cases where the value in the reg can be negative).
; Likewise for stores.
; CHECK-LABEL: load_test1:
; CHECK-NEXT: param i32{{$}}
; CHECK-NEXT: result i32{{$}}
; CHECK-NEXT: i32.const $push0=, 2{{$}}
; CHECK-NEXT: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEXT: i32.load $push2=, g+40($pop1){{$}}
; CHECK-NEXT: return $pop2{{$}}
; CHECK-NEX T: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEX T: i32.load $push2=, g+40($pop1){{$}}
; CHECK-NEX T: return $pop2{{$}}
define i32 @load_test1(i32 %n) {
%add = add nsw i32 %n, 10
%arrayidx = getelementptr inbounds [0 x i32], [0 x i32]* @g, i32 0, i32 %add
@ -46,9 +50,9 @@ define i32 @load_test1(i32 %n) {
; CHECK-NEXT: param i32{{$}}
; CHECK-NEXT: result i32{{$}}
; CHECK-NEXT: i32.const $push0=, 2{{$}}
; CHECK-NEXT: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEXT: i32.load $push2=, g+40($pop1){{$}}
; CHECK-NEXT: return $pop2{{$}}
; CHECK-NEX T: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEX T: i32.load $push2=, g+40($pop1){{$}}
; CHECK-NEX T: return $pop2{{$}}
define i32 @load_test2(i32 %n) {
%add = add nsw i32 10, %n
%arrayidx = getelementptr inbounds [0 x i32], [0 x i32]* @g, i32 0, i32 %add
@ -60,9 +64,9 @@ define i32 @load_test2(i32 %n) {
; CHECK-NEXT: param i32{{$}}
; CHECK-NEXT: result i32{{$}}
; CHECK-NEXT: i32.const $push0=, 2{{$}}
; CHECK-NEXT: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEXT: i32.load $push2=, g+40($pop1){{$}}
; CHECK-NEXT: return $pop2{{$}}
; CHECK-NEX T: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEX T: i32.load $push2=, g+40($pop1){{$}}
; CHECK-NEX T: return $pop2{{$}}
define i32 @load_test3(i32 %n) {
%add.ptr = getelementptr inbounds [0 x i32], [0 x i32]* @g, i32 0, i32 %n
%add.ptr1 = getelementptr inbounds i32, i32* %add.ptr, i32 10
@ -74,9 +78,9 @@ define i32 @load_test3(i32 %n) {
; CHECK-NEXT: param i32{{$}}
; CHECK-NEXT: result i32{{$}}
; CHECK-NEXT: i32.const $push0=, 2{{$}}
; CHECK-NEXT: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEXT: i32.load $push2=, g+40($pop1){{$}}
; CHECK-NEXT: return $pop2{{$}}
; CHECK-NEX T: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEX T: i32.load $push2=, g+40($pop1){{$}}
; CHECK-NEX T: return $pop2{{$}}
define i32 @load_test4(i32 %n) {
%add.ptr = getelementptr inbounds i32, i32* getelementptr inbounds ([0 x i32], [0 x i32]* @g, i32 0, i32 10), i32 %n
%t = load i32, i32* %add.ptr, align 4
@ -87,9 +91,9 @@ define i32 @load_test4(i32 %n) {
; CHECK-NEXT: param i32{{$}}
; CHECK-NEXT: result i32{{$}}
; CHECK-NEXT: i32.const $push0=, 2{{$}}
; CHECK-NEXT: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEXT: i32.load $push2=, g+40($pop1){{$}}
; CHECK-NEXT: return $pop2{{$}}
; CHECK-NEX T: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEX T: i32.load $push2=, g+40($pop1){{$}}
; CHECK-NEX T: return $pop2{{$}}
define i32 @load_test5(i32 %n) {
%add.ptr = getelementptr inbounds i32, i32* getelementptr inbounds ([0 x i32], [0 x i32]* @g, i32 0, i32 10), i32 %n
%t = load i32, i32* %add.ptr, align 4
@ -100,9 +104,9 @@ define i32 @load_test5(i32 %n) {
; CHECK-NEXT: param i32{{$}}
; CHECK-NEXT: result i32{{$}}
; CHECK-NEXT: i32.const $push0=, 2{{$}}
; CHECK-NEXT: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEXT: i32.load $push2=, g+40($pop1){{$}}
; CHECK-NEXT: return $pop2{{$}}
; CHECK-NEX T: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEX T: i32.load $push2=, g+40($pop1){{$}}
; CHECK-NEX T: return $pop2{{$}}
define i32 @load_test6(i32 %n) {
%add = add nsw i32 %n, 10
%add.ptr = getelementptr inbounds [0 x i32], [0 x i32]* @g, i32 0, i32 %add
@ -114,9 +118,9 @@ define i32 @load_test6(i32 %n) {
; CHECK-NEXT: param i32{{$}}
; CHECK-NEXT: result i32{{$}}
; CHECK-NEXT: i32.const $push0=, 2{{$}}
; CHECK-NEXT: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEXT: i32.load $push2=, g+40($pop1){{$}}
; CHECK-NEXT: return $pop2{{$}}
; CHECK-NEX T: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEX T: i32.load $push2=, g+40($pop1){{$}}
; CHECK-NEX T: return $pop2{{$}}
define i32 @load_test7(i32 %n) {
%add.ptr = getelementptr inbounds [0 x i32], [0 x i32]* @g, i32 0, i32 %n
%add.ptr1 = getelementptr inbounds i32, i32* %add.ptr, i32 10
@ -128,9 +132,9 @@ define i32 @load_test7(i32 %n) {
; CHECK-NEXT: param i32{{$}}
; CHECK-NEXT: result i32{{$}}
; CHECK-NEXT: i32.const $push0=, 2{{$}}
; CHECK-NEXT: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEXT: i32.load $push2=, g+40($pop1){{$}}
; CHECK-NEXT: return $pop2{{$}}
; CHECK-NEX T: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEX T: i32.load $push2=, g+40($pop1){{$}}
; CHECK-NEX T: return $pop2{{$}}
define i32 @load_test8(i32 %n) {
%add = add nsw i32 10, %n
%add.ptr = getelementptr inbounds [0 x i32], [0 x i32]* @g, i32 0, i32 %add
@ -374,8 +378,8 @@ define void @store_test0_noinbounds(i32 %i) {
; CHECK-NEXT: param i32, i32{{$}}
; CHECK-NEXT: i32.const $push0=, 2{{$}}
; CHECK-NEXT: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEXT: i32.store $drop=, g+40($pop1), $1{{$}}
; CHECK-NEXT: return{{$}}
; CHECK-NEX T: i32.store $drop=, g+40($pop1), $1{{$}}
; CHECK-NEX T: return{{$}}
define void @store_test1(i32 %n, i32 %i) {
%add = add nsw i32 %n, 10
%arrayidx = getelementptr inbounds [0 x i32], [0 x i32]* @g, i32 0, i32 %add
@ -387,8 +391,8 @@ define void @store_test1(i32 %n, i32 %i) {
; CHECK-NEXT: param i32, i32{{$}}
; CHECK-NEXT: i32.const $push0=, 2{{$}}
; CHECK-NEXT: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEXT: i32.store $drop=, g+40($pop1), $1{{$}}
; CHECK-NEXT: return{{$}}
; CHECK-NEX T: i32.store $drop=, g+40($pop1), $1{{$}}
; CHECK-NEX T: return{{$}}
define void @store_test2(i32 %n, i32 %i) {
%add = add nsw i32 10, %n
%arrayidx = getelementptr inbounds [0 x i32], [0 x i32]* @g, i32 0, i32 %add
@ -400,8 +404,8 @@ define void @store_test2(i32 %n, i32 %i) {
; CHECK-NEXT: param i32, i32{{$}}
; CHECK-NEXT: i32.const $push0=, 2{{$}}
; CHECK-NEXT: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEXT: i32.store $drop=, g+40($pop1), $1{{$}}
; CHECK-NEXT: return{{$}}
; CHECK-NEX T: i32.store $drop=, g+40($pop1), $1{{$}}
; CHECK-NEX T: return{{$}}
define void @store_test3(i32 %n, i32 %i) {
%add.ptr = getelementptr inbounds [0 x i32], [0 x i32]* @g, i32 0, i32 %n
%add.ptr1 = getelementptr inbounds i32, i32* %add.ptr, i32 10
@ -413,8 +417,8 @@ define void @store_test3(i32 %n, i32 %i) {
; CHECK-NEXT: param i32, i32{{$}}
; CHECK-NEXT: i32.const $push0=, 2{{$}}
; CHECK-NEXT: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEXT: i32.store $drop=, g+40($pop1), $1{{$}}
; CHECK-NEXT: return{{$}}
; CHECK-NEX T: i32.store $drop=, g+40($pop1), $1{{$}}
; CHECK-NEX T: return{{$}}
define void @store_test4(i32 %n, i32 %i) {
%add.ptr = getelementptr inbounds i32, i32* getelementptr inbounds ([0 x i32], [0 x i32]* @g, i32 0, i32 10), i32 %n
store i32 %i, i32* %add.ptr, align 4
@ -425,8 +429,8 @@ define void @store_test4(i32 %n, i32 %i) {
; CHECK-NEXT: param i32, i32{{$}}
; CHECK-NEXT: i32.const $push0=, 2{{$}}
; CHECK-NEXT: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEXT: i32.store $drop=, g+40($pop1), $1{{$}}
; CHECK-NEXT: return{{$}}
; CHECK-NEX T: i32.store $drop=, g+40($pop1), $1{{$}}
; CHECK-NEX T: return{{$}}
define void @store_test5(i32 %n, i32 %i) {
%add.ptr = getelementptr inbounds i32, i32* getelementptr inbounds ([0 x i32], [0 x i32]* @g, i32 0, i32 10), i32 %n
store i32 %i, i32* %add.ptr, align 4
@ -437,8 +441,8 @@ define void @store_test5(i32 %n, i32 %i) {
; CHECK-NEXT: param i32, i32{{$}}
; CHECK-NEXT: i32.const $push0=, 2{{$}}
; CHECK-NEXT: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEXT: i32.store $drop=, g+40($pop1), $1{{$}}
; CHECK-NEXT: return{{$}}
; CHECK-NEX T: i32.store $drop=, g+40($pop1), $1{{$}}
; CHECK-NEX T: return{{$}}
define void @store_test6(i32 %n, i32 %i) {
%add = add nsw i32 %n, 10
%add.ptr = getelementptr inbounds [0 x i32], [0 x i32]* @g, i32 0, i32 %add
@ -450,8 +454,8 @@ define void @store_test6(i32 %n, i32 %i) {
; CHECK-NEXT: param i32, i32{{$}}
; CHECK-NEXT: i32.const $push0=, 2{{$}}
; CHECK-NEXT: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEXT: i32.store $drop=, g+40($pop1), $1{{$}}
; CHECK-NEXT: return{{$}}
; CHECK-NEX T: i32.store $drop=, g+40($pop1), $1{{$}}
; CHECK-NEX T: return{{$}}
define void @store_test7(i32 %n, i32 %i) {
%add.ptr = getelementptr inbounds [0 x i32], [0 x i32]* @g, i32 0, i32 %n
%add.ptr1 = getelementptr inbounds i32, i32* %add.ptr, i32 10
@ -463,8 +467,8 @@ define void @store_test7(i32 %n, i32 %i) {
; CHECK-NEXT: param i32, i32{{$}}
; CHECK-NEXT: i32.const $push0=, 2{{$}}
; CHECK-NEXT: i32.shl $push1=, $0, $pop0{{$}}
; CHECK-NEXT: i32.store $drop=, g+40($pop1), $1{{$}}
; CHECK-NEXT: return{{$}}
; CHECK-NEX T: i32.store $drop=, g+40($pop1), $1{{$}}
; CHECK-NEX T: return{{$}}
define void @store_test8(i32 %n, i32 %i) {
%add = add nsw i32 10, %n
%add.ptr = getelementptr inbounds [0 x i32], [0 x i32]* @g, i32 0, i32 %add

View File

@ -0,0 +1,43 @@
; RUN: llc < %s -asm-verbose=false -verify-machineinstrs -disable-wasm-fallthrough-return-opt | FileCheck %s
target datalayout = "e-m:e-p:32:32-i64:64-n32:64-S128"
target triple = "wasm32"
@args = hidden local_unnamed_addr global [32 x i32] zeroinitializer, align 16
; Function Attrs: norecurse nounwind
define hidden i32 @main() local_unnamed_addr #0 {
; If LSR stops selecting a negative base reg value, then this test will no
; longer be useful as written.
; CHECK: i32.const $0=, -128
entry:
br label %for.body
for.body: ; preds = %for.body, %entry
%i.04 = phi i32 [ 0, %entry ], [ %inc, %for.body ]
; The offset should not be folded into the store.
; CHECK: i32.const $push{{[0-9]+}}=, args+128
; CHECK: i32.add
; CHECK: i32.store $drop=, 0(
%arrayidx = getelementptr inbounds [32 x i32], [32 x i32]* @args, i32 0, i32 %i.04
store i32 1, i32* %arrayidx, align 4, !tbaa !1
%inc = add nuw nsw i32 %i.04, 1
%exitcond = icmp eq i32 %inc, 32
br i1 %exitcond, label %for.end, label %for.body, !llvm.loop !5
for.end: ; preds = %for.body
ret i32 0
}
attributes #0 = { norecurse nounwind "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "no-frame-pointer-elim"="false" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="generic" "unsafe-fp-math"="false" "use-soft-float"="false" }
!llvm.ident = !{!0}
!0 = !{!"clang version 4.0.0 (trunk 279056) (llvm/trunk 279074)"}
!1 = !{!2, !2, i64 0}
!2 = !{!"int", !3, i64 0}
!3 = !{!"omnipotent char", !4, i64 0}
!4 = !{!"Simple C/C++ TBAA"}
!5 = distinct !{!5, !6}
!6 = !{!"llvm.loop.unroll.disable"}