replica
/
hanchenye-llvm-project
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[SCEV] Make getRange smarter around selects 

Have ScalarEvolution::getRange re-consider cases like "{C?A:B,+,C?P:Q}"
by factoring out "C" and computing RangeOf{A,+,P} union RangeOf({B,+,Q})
instead.

The latter can be easier to compute precisely in cases like
"{C?0:N,+,C?1:-1}" N is the backedge taken count of the loop; since in
such cases the latter form simplifies to [0,N+1) union [0,N+1).

llvm-svn: 262438
This commit is contained in:
Sanjoy Das 2016-03-02 00:57:54 +00:00
parent b765b633cb
commit bf73098472
3 changed files with 178 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
llvm/include/llvm/Analysis/ScalarEvolution.h
View File

@ -672,6 +672,13 @@ namespace llvm {
const SCEV *MaxBECount,
unsigned BitWidth);
/// Try to compute a range for the affine SCEVAddRecExpr {\p Start,+,\p
/// Stop} by "factoring out" a ternary expression from the add recurrence.
/// Helper called by \c getRange.
ConstantRange getRangeViaFactoring(const SCEV *Start, const SCEV *Stop,
const SCEV *MaxBECount,
unsigned BitWidth);
/// We know that there is no SCEV for the specified value. Analyze the
/// expression.
const SCEV *createSCEV(Value *V);

83
llvm/lib/Analysis/ScalarEvolution.cpp
View File

@ -4400,6 +4400,13 @@ ScalarEvolution::getRange(const SCEV *S,
if (!RangeFromAffine.isFullSet())
ConservativeResult =
ConservativeResult.intersectWith(RangeFromAffine);
auto RangeFromFactoring = getRangeViaFactoring(
AddRec->getStart(), AddRec->getStepRecurrence(*this), MaxBECount,
BitWidth);
if (!RangeFromFactoring.isFullSet())
ConservativeResult =
ConservativeResult.intersectWith(RangeFromFactoring);
}
}
@ -4504,6 +4511,82 @@ ConstantRange ScalarEvolution::getRangeForAffineAR(const SCEV *Start,
return Result;
}
ConstantRange ScalarEvolution::getRangeViaFactoring(const SCEV *Start,
const SCEV *Step,
const SCEV *MaxBECount,
unsigned BitWidth) {
APInt Offset(BitWidth, 0);
if (auto *SA = dyn_cast<SCEVAddExpr>(Start)) {
// Peel off a constant offset, if possible. In the future we could consider
// being smarter here and handle {Start+Step,+,Step} too.
if (SA->getNumOperands() != 2 || !isa<SCEVConstant>(SA->getOperand(0)))
return ConstantRange(BitWidth, /* isFullSet = */ true);
Offset = cast<SCEVConstant>(SA->getOperand(0))->getAPInt();
Start = SA->getOperand(1);
}
if (!isa<SCEVUnknown>(Start) || !isa<SCEVUnknown>(Step))
// We don't have anything new to contribute in this case.
return ConstantRange(BitWidth, /* isFullSet = */ true);
// RangeOf({C?A:B,+,C?P:Q}) == RangeOf(C?{A,+,P}:{B,+,Q})
// == RangeOf({A,+,P}) union RangeOf({B,+,Q})
struct SelectPattern {
Value *Condition = nullptr;
const APInt *TrueValue = nullptr;
const APInt *FalseValue = nullptr;
explicit SelectPattern(const SCEVUnknown *SU) {
using namespace llvm::PatternMatch;
if (!match(SU->getValue(),
m_Select(m_Value(Condition), m_APInt(TrueValue),
m_APInt(FalseValue)))) {
Condition = nullptr;
TrueValue = FalseValue = nullptr;
}
}
bool isRecognized() {
assert(((Condition && TrueValue && FalseValue) ||
(!Condition && !TrueValue && !FalseValue)) &&
"Invariant: either all three are non-null or all three are null");
return TrueValue != nullptr;
}
};
SelectPattern StartPattern(cast<SCEVUnknown>(Start));
if (!StartPattern.isRecognized())
return ConstantRange(BitWidth, /* isFullSet = */ true);
SelectPattern StepPattern(cast<SCEVUnknown>(Step));
if (!StepPattern.isRecognized())
return ConstantRange(BitWidth, /* isFullSet = */ true);
if (StartPattern.Condition != StepPattern.Condition) {
// We don't handle this case today; but we could, by considering four
// possibilities below instead of two. I'm not sure if there are cases where
// that will help over what getRange already does, though.
return ConstantRange(BitWidth, /* isFullSet = */ true);
}
// NB! Calling ScalarEvolution::getConstant is fine, but we should not try to
// construct arbitrary general SCEV expressions here. This function is called
// from deep in the call stack, and calling getSCEV (on a sext instruction,
// say) can end up caching a suboptimal value.
auto TrueRange = getRangeForAffineAR(
getConstant(*StartPattern.TrueValue + Offset),
getConstant(*StepPattern.TrueValue), MaxBECount, BitWidth);
auto FalseRange = getRangeForAffineAR(
getConstant(*StartPattern.FalseValue + Offset),
getConstant(*StepPattern.FalseValue), MaxBECount, BitWidth);
return TrueRange.unionWith(FalseRange);
}
SCEV::NoWrapFlags ScalarEvolution::getNoWrapFlagsFromUB(const Value *V) {
if (isa<ConstantExpr>(V)) return SCEV::FlagAnyWrap;
const BinaryOperator *BinOp = cast<BinaryOperator>(V);

88
llvm/test/Analysis/ScalarEvolution/increasing-and-decreasing-range.ll Normal file
View File

@ -0,0 +1,88 @@
; RUN: opt -analyze -scalar-evolution < %s | FileCheck %s
define void @f0(i1 %c) {
; CHECK-LABEL: Classifying expressions for: @f0
entry:
%start = select i1 %c, i32 127, i32 0
%step = select i1 %c, i32 -1, i32 1
br label %loop
loop:
%loop.iv = phi i32 [ 0, %entry ], [ %loop.iv.inc, %loop ]
%iv = phi i32 [ %start, %entry ], [ %iv.next, %loop ]
; CHECK: %iv = phi i32 [ %start, %entry ], [ %iv.next, %loop ]
; CHECK-NEXT: --> {%start,+,%step}<%loop> U: [0,128) S: [0,128)
%iv.next = add i32 %iv, %step
%loop.iv.inc = add i32 %loop.iv, 1
%be.cond = icmp ne i32 %loop.iv.inc, 128
br i1 %be.cond, label %loop, label %leave
leave:
ret void
}
define void @f1(i1 %c) {
; CHECK-LABEL: Classifying expressions for: @f1
entry:
%start = select i1 %c, i32 120, i32 0
%step = select i1 %c, i32 -8, i32 8
br label %loop
loop:
%loop.iv = phi i32 [ 0, %entry ], [ %loop.iv.inc, %loop ]
%iv = phi i32 [ %start, %entry ], [ %iv.next, %loop ]
; CHECK: %iv.1 = add i32 %iv, 1
; CHECK-NEXT: --> {(1 + %start)<nuw><nsw>,+,%step}<%loop> U: [1,122) S: [1,122)
; CHECK: %iv.2 = add i32 %iv, 2
; CHECK-NEXT: --> {(2 + %start)<nuw><nsw>,+,%step}<%loop> U: [2,123) S: [2,123)
; CHECK: %iv.3 = add i32 %iv, 3
; CHECK-NEXT: --> {(3 + %start)<nuw><nsw>,+,%step}<%loop> U: [3,124) S: [3,124)
; CHECK: %iv.4 = add i32 %iv, 4
; CHECK-NEXT: --> {(4 + %start)<nuw><nsw>,+,%step}<%loop> U: [4,125) S: [4,125)
; CHECK: %iv.5 = add i32 %iv, 5
; CHECK-NEXT: --> {(5 + %start)<nuw><nsw>,+,%step}<%loop> U: [5,126) S: [5,126)
; CHECK: %iv.6 = add i32 %iv, 6
; CHECK-NEXT: --> {(6 + %start)<nuw><nsw>,+,%step}<%loop> U: [6,127) S: [6,127)
; CHECK: %iv.7 = add i32 %iv, 7
; CHECK-NEXT: --> {(7 + %start)<nuw><nsw>,+,%step}<%loop> U: [7,128) S: [7,128)
%iv.1 = add i32 %iv, 1
%iv.2 = add i32 %iv, 2
%iv.3 = add i32 %iv, 3
%iv.4 = add i32 %iv, 4
%iv.5 = add i32 %iv, 5
%iv.6 = add i32 %iv, 6
%iv.7 = add i32 %iv, 7
; CHECK: %iv.m1 = sub i32 %iv, 1
; CHECK-NEXT: --> {(-1 + %start)<nsw>,+,%step}<%loop> U: [-1,120) S: [-1,120)
; CHECK: %iv.m2 = sub i32 %iv, 2
; CHECK-NEXT: --> {(-2 + %start)<nsw>,+,%step}<%loop> U: [-2,119) S: [-2,119)
; CHECK: %iv.m3 = sub i32 %iv, 3
; CHECK-NEXT: --> {(-3 + %start)<nsw>,+,%step}<%loop> U: [-3,118) S: [-3,118)
; CHECK: %iv.m4 = sub i32 %iv, 4
; CHECK-NEXT: --> {(-4 + %start)<nsw>,+,%step}<%loop> U: [-4,117) S: [-4,117)
; CHECK: %iv.m5 = sub i32 %iv, 5
; CHECK-NEXT: --> {(-5 + %start)<nsw>,+,%step}<%loop> U: [-5,116) S: [-5,116)
; CHECK: %iv.m6 = sub i32 %iv, 6
; CHECK-NEXT: --> {(-6 + %start)<nsw>,+,%step}<%loop> U: [-6,115) S: [-6,115)
; CHECK: %iv.m7 = sub i32 %iv, 7
; CHECK-NEXT: --> {(-7 + %start)<nsw>,+,%step}<%loop> U: [-7,114) S: [-7,114)
%iv.m1 = sub i32 %iv, 1
%iv.m2 = sub i32 %iv, 2
%iv.m3 = sub i32 %iv, 3
%iv.m4 = sub i32 %iv, 4
%iv.m5 = sub i32 %iv, 5
%iv.m6 = sub i32 %iv, 6
%iv.m7 = sub i32 %iv, 7
%iv.next = add i32 %iv, %step
%loop.iv.inc = add i32 %loop.iv, 1
%be.cond = icmp sgt i32 %loop.iv, 14
br i1 %be.cond, label %leave, label %loop
leave:
ret void
}