[llvm] d177988 - [SCEV] Retain AddExpr flags when subtracting a foldable constant.
Florian Hahn via llvm-commits
llvm-commits at lists.llvm.org
Tue Jun 22 04:08:52 PDT 2021
Author: Florian Hahn
Date: 2021-06-22T11:27:51+01:00
New Revision: d17798823ca2213af353e125f6f5233522dda7ce
URL: https://github.com/llvm/llvm-project/commit/d17798823ca2213af353e125f6f5233522dda7ce
DIFF: https://github.com/llvm/llvm-project/commit/d17798823ca2213af353e125f6f5233522dda7ce.diff
LOG: [SCEV] Retain AddExpr flags when subtracting a foldable constant.
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
Added:
Modified:
llvm/lib/Analysis/ScalarEvolution.cpp
llvm/test/Analysis/LoopAccessAnalysis/reverse-memcheck-bounds.ll
llvm/test/Transforms/IndVarSimplify/eliminate-exit-no-dl.ll
Removed:
################################################################################
diff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp
index f011020266c74..4af19e49fb6f7 100644
--- a/llvm/lib/Analysis/ScalarEvolution.cpp
+++ b/llvm/lib/Analysis/ScalarEvolution.cpp
@@ -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;
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/reverse-memcheck-bounds.ll b/llvm/test/Analysis/LoopAccessAnalysis/reverse-memcheck-bounds.ll
index be51167dab8a9..84ee043b5c264 100644
--- a/llvm/test/Analysis/LoopAccessAnalysis/reverse-memcheck-bounds.ll
+++ b/llvm/test/Analysis/LoopAccessAnalysis/reverse-memcheck-bounds.ll
@@ -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
diff --git a/llvm/test/Transforms/IndVarSimplify/eliminate-exit-no-dl.ll b/llvm/test/Transforms/IndVarSimplify/eliminate-exit-no-dl.ll
index 11c960e103f57..0abfb06bfe2b0 100644
--- a/llvm/test/Transforms/IndVarSimplify/eliminate-exit-no-dl.ll
+++ b/llvm/test/Transforms/IndVarSimplify/eliminate-exit-no-dl.ll
@@ -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]]
More information about the llvm-commits
mailing list