RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-25 12:12:47 +01:00
Code Issues Projects Releases Wiki Activity

[SCEV] Retain AddExpr flags when subtracting a foldable constant.

Browse Source 
Currently we drop wrapping flags for expressions like (A + C1)<flags> - C2.

But we can retain flags under certain conditions:

* Adding a smaller constant is NUW if the original AddExpr was NUW.

* Adding a constant with the same sign and small magnitude is NSW, if the
  original AddExpr was NSW.

This can improve results after using `SimplifyICmpOperands`, which may
subtract one in order to use stricter predicates, as is the case for
`isKnownPredicate`.

Reviewed By: efriedma

Differential Revision: https://reviews.llvm.org/D104319
This commit is contained in:
Florian Hahn 2021-06-21 19:11:11 +01:00
parent a307928fe5
commit e93c4c3336
3 changed files with 44 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

42
lib/Analysis/ScalarEvolution.cpp
View File

@ -2541,6 +2541,48 @@ const SCEV *ScalarEvolution::getAddExpr(SmallVectorImpl<const SCEV *> &Ops,
}
}
if (Ops.size() == 2) {
// Check if we have an expression of the form ((X + C1) - C2), where C1 and
// C2 can be folded in a way that allows retaining wrapping flags of (X +
// C1).
const SCEV *A = Ops[0];
const SCEV *B = Ops[1];
auto *AddExpr = dyn_cast<SCEVAddExpr>(B);
auto *C = dyn_cast<SCEVConstant>(A);
if (AddExpr && C && isa<SCEVConstant>(AddExpr->getOperand(0))) {
auto C1 = cast<SCEVConstant>(AddExpr->getOperand(0))->getAPInt();
auto C2 = C->getAPInt();
SCEV::NoWrapFlags PreservedFlags = SCEV::FlagAnyWrap;
APInt ConstAdd = C1 + C2;
auto AddFlags = AddExpr->getNoWrapFlags();
// Adding a smaller constant is NUW if the original AddExpr was NUW.
if (ScalarEvolution::maskFlags(AddFlags, SCEV::FlagNUW) ==
SCEV::FlagNUW &&
ConstAdd.ule(C1)) {
PreservedFlags =
ScalarEvolution::setFlags(PreservedFlags, SCEV::FlagNUW);
}
// Adding a constant with the same sign and small magnitude is NSW, if the
// original AddExpr was NSW.
if (ScalarEvolution::maskFlags(AddFlags, SCEV::FlagNSW) ==
SCEV::FlagNSW &&
C1.isSignBitSet() == ConstAdd.isSignBitSet() &&
ConstAdd.abs().ule(C1.abs())) {
PreservedFlags =
ScalarEvolution::setFlags(PreservedFlags, SCEV::FlagNSW);
}
if (PreservedFlags != SCEV::FlagAnyWrap) {
SmallVector<const SCEV *, 4> NewOps(AddExpr->op_begin(),
AddExpr->op_end());
NewOps[0] = getConstant(ConstAdd);
return getAddExpr(NewOps, PreservedFlags);
}
}
}
// Skip past any other cast SCEVs.
while (Idx < Ops.size() && Ops[Idx]->getSCEVType() < scAddExpr)
++Idx;

2
test/Analysis/LoopAccessAnalysis/reverse-memcheck-bounds.ll
View File

@ -16,7 +16,7 @@ target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128"
target triple = "aarch64--linux-gnueabi"
; CHECK: function 'f':
; CHECK: (Low: (20000 + %a) High: (60004 + %a))
; CHECK: (Low: (20000 + %a)<nuw> High: (60004 + %a))
@B = common global i32* null, align 8
@A = common global i32* null, align 8

5
test/Transforms/IndVarSimplify/eliminate-exit-no-dl.ll
View File

@ -12,15 +12,12 @@
define void @foo() {
; CHECK-LABEL: @foo(
; CHECK-NEXT: bb:
; CHECK-NEXT: [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 add (i64 ptrtoint ([0 x i8]* @global to i64), i64 500), i64 add (i64 ptrtoint ([0 x i8]* @global to i64), i64 1))
; CHECK-NEXT: [[TMP0:%.*]] = sub i64 add (i64 ptrtoint ([0 x i8]* @global to i64), i64 1), [[UMIN]]
; CHECK-NEXT: [[TMP1:%.*]] = icmp ne i64 [[TMP0]], 0
; CHECK-NEXT: br label [[BB3:%.*]]
; CHECK: bb3:
; CHECK-NEXT: [[TMP:%.*]] = phi i8* [ [[TMP4:%.*]], [[BB7:%.*]] ], [ getelementptr inbounds ([0 x i8], [0 x i8]* @global, i64 0, i64 2), [[BB:%.*]] ]
; CHECK-NEXT: [[TMP4]] = getelementptr inbounds i8, i8* [[TMP]], i64 -1
; CHECK-NEXT: [[TMP6:%.*]] = load i8, i8* [[TMP4]], align 1
; CHECK-NEXT: br i1 [[TMP1]], label [[BB7]], label [[BB11:%.*]]
; CHECK-NEXT: br i1 false, label [[BB7]], label [[BB11:%.*]]
; CHECK: bb7:
; CHECK-NEXT: [[TMP8:%.*]] = zext i8 [[TMP6]] to i64
; CHECK-NEXT: br i1 true, label [[BB11]], label [[BB3]]