[ValueTracking][SelectionDAG] Rename ComputeMinSignedBits->ComputeMaxSignificantBits. NFC

[craig.topper]

Summary

This function returns an upper bound on the number of bits needed
to represent the signed value. Use "Max" to match similar functions
in KnownBits like countMaxActiveBits.

Rename APInt::getMinSignedBits->getSignificantBits. Keeping the old
name around to keep this patch size down. Will do a bulk rename as
follow up.

Rename KnownBits::countMaxSignedBits->countMaxSignificantBits.

[RKSimon]

LGTM - with one (very pedantic) minor

llvm/include/llvm/Analysis/ValueTracking.h

204–205

The use of minimum might be confusing - please can you tweak the description?

llvm/include/llvm/CodeGen/SelectionDAG.h

1835–1836

Tweak the description?

1841–1842

Tweak the description?

[craig.topper]

Tweak comments.

[craig.topper]

Use ComputeMaxSignificantBits instead.

[spatel]

LGTM - someday we might fix the capitalization of function names too. :)

llvm/include/llvm/Analysis/ValueTracking.h

214

That should be "APInt::getSignificantBits function" (no "Max")?

llvm/include/llvm/CodeGen/SelectionDAG.h

1838

That should be "APInt::getSignificantBits function" (no "Max")?

[foad]

Rename APInt::getMinSignedBits->getSignificantBits

Why drop "Signed" from the name? Now we have getActiveBits for the unsigned version of this functionality and getSignificantBits for the signed version, which I don't think is very intuitive.

[spatel]

In D116522#3219081, @foad wrote:

Rename APInt::getMinSignedBits->getSignificantBits

Why drop "Signed" from the name? Now we have getActiveBits for the unsigned version of this functionality and getSignificantBits for the signed version, which I don't think is very intuitive.

Good point. So we should standardize on either "active" or "significant". Go with "active" since it already exists and is shorter?
We could also standardize the verb in these names. Currently, it is "get", "count", or "compute".
If we unify those, it could be something like this across all of these APIs:
get[Max]ActiveBitsSigned()
get[Max]ActiveBitsUnsigned()

[craig.topper]

In D116522#3219566, @spatel wrote:

In D116522#3219081, @foad wrote:

Rename APInt::getMinSignedBits->getSignificantBits

Why drop "Signed" from the name? Now we have getActiveBits for the unsigned version of this functionality and getSignificantBits for the signed version, which I don't think is very intuitive.

Good point. So we should standardize on either "active" or "significant". Go with "active" since it already exists and is shorter?
We could also standardize the verb in these names. Currently, it is "get", "count", or "compute".
If we unify those, it could be something like this across all of these APIs:
get[Max]ActiveBitsSigned()
get[Max]ActiveBitsUnsigned()

I don't think we should apply the "Active" term to signed. In my head Active is kind of associated with bits that are ones. There's also APIntgetActiveWords that is implemented using getActiveBits to determine the last word that has ones in it.

llvm/include/llvm/Analysis/ValueTracking.h

 //===- llvm/Analysis/ValueTracking.h - Walk computations --------*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
constexpr unsigned MaxAnalysisRecursionDepth = 6;
   /// equal to each other, so we return 3. For vectors, return the number of
   /// sign bits for the vector element with the mininum number of known sign
   /// bits.
   unsigned ComputeNumSignBits(const Value *Op, const DataLayout &DL,
                               unsigned Depth = 0, AssumptionCache *AC = nullptr,
                               const Instruction *CxtI = nullptr,
                               const DominatorTree *DT = nullptr,
                               bool UseInstrInfo = true);
 
-  /// Get the minimum bit size for this Value \p Op as a signed integer.
-  /// i.e.  x == sext(trunc(x to MinSignedBits) to bitwidth(x)).
-  /// Similar to the APInt::getMinSignedBits function.
-  unsigned ComputeMinSignedBits(const Value *Op, const DataLayout &DL,
+  /// Get the upper bound on bit size for this Value \p Op as a signed integer.
+  /// i.e.  x == sext(trunc(x to MaxSignedBits) to bitwidth(x)).
+  /// Similar to the APInt::getMaxSignedBits function.
+  unsigned ComputeMaxSignedBits(const Value *Op, const DataLayout &DL,
                                 unsigned Depth = 0,
                                 AssumptionCache *AC = nullptr,
                                 const Instruction *CxtI = nullptr,
                                 const DominatorTree *DT = nullptr);
 
   /// This function computes the integer multiple of Base that equals V. If
   /// successful, it returns true and returns the multiple in Multiple. If
   /// unsuccessful, it returns false. Also, if V can be simplified to an
   /// integer, then the simplified V is returned in Val. Look through sext only
   /// if LookThroughSExt=true.
   bool ComputeMultiple(Value *V, unsigned Base, Value *&Multiple,
                        bool LookThroughSExt = false,
                        unsigned Depth = 0);
 

llvm/include/llvm/CodeGen/SelectionDAG.h

 //===- llvm/CodeGen/SelectionDAG.h - InstSelection DAG ----------*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
public:
   /// bit (itself), but other cases can give us information. For example,
   /// immediately after an "SRA X, 2", we know that the top 3 bits are all equal
   /// to each other, so we return 3. The DemandedElts argument allows
   /// us to only collect the minimum sign bits of the requested vector elements.
   /// Targets can implement the ComputeNumSignBitsForTarget method in the
   /// TargetLowering class to allow target nodes to be understood.
   unsigned ComputeNumSignBits(SDValue Op, const APInt &DemandedElts,
                               unsigned Depth = 0) const;
 
-  /// Get the minimum bit size for this Value \p Op as a signed integer.
-  /// i.e.  x == sext(trunc(x to MinSignedBits) to bitwidth(x)).
-  /// Similar to the APInt::getMinSignedBits function.
+  /// Get the upper bound on bit size for this Value \p Op as a signed integer.
+  /// i.e.  x == sext(trunc(x to MaxSignedBits) to bitwidth(x)).
+  /// Similar to the APInt::getMaxSignedBits function.
   /// Helper wrapper to ComputeNumSignBits.
-  unsigned ComputeMinSignedBits(SDValue Op, unsigned Depth = 0) const;
+  unsigned ComputeMaxSignedBits(SDValue Op, unsigned Depth = 0) const;
 
-  /// Get the minimum bit size for this Value \p Op as a signed integer.
-  /// i.e.  x == sext(trunc(x to MinSignedBits) to bitwidth(x)).
-  /// Similar to the APInt::getMinSignedBits function.
+  /// Get the upper bound on bit size for this Value \p Op as a signed integer.
+  /// i.e.  x == sext(trunc(x to MaxSignedBits) to bitwidth(x)).
+  /// Similar to the APInt::getMaxSignedBits function.
   /// Helper wrapper to ComputeNumSignBits.
-  unsigned ComputeMinSignedBits(SDValue Op, const APInt &DemandedElts,
+  unsigned ComputeMaxSignedBits(SDValue Op, const APInt &DemandedElts,
                                 unsigned Depth = 0) const;
 
   /// Return true if this function can prove that \p Op is never poison
   /// and, if \p PoisonOnly is false, does not have undef bits.
   bool isGuaranteedNotToBeUndefOrPoison(SDValue Op, bool PoisonOnly = false,
                                         unsigned Depth = 0) const;
 
   /// Return true if this function can prove that \p Op is never poison

llvm/include/llvm/Transforms/InstCombine/InstCombiner.h

 //===- InstCombiner.h - InstCombine implementation --------------*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 /// \file
bool MaskedValueIsZero(const Value *V, const APInt &Mask, unsigned Depth = 0,
     return llvm::MaskedValueIsZero(V, Mask, DL, Depth, &AC, CxtI, &DT);
   }
 
   unsigned ComputeNumSignBits(const Value *Op, unsigned Depth = 0,
                               const Instruction *CxtI = nullptr) const {
     return llvm::ComputeNumSignBits(Op, DL, Depth, &AC, CxtI, &DT);
   }
 
-  unsigned ComputeMinSignedBits(const Value *Op, unsigned Depth = 0,
+  unsigned ComputeMaxSignedBits(const Value *Op, unsigned Depth = 0,
                                 const Instruction *CxtI = nullptr) const {
-    return llvm::ComputeMinSignedBits(Op, DL, Depth, &AC, CxtI, &DT);
+    return llvm::ComputeMaxSignedBits(Op, DL, Depth, &AC, CxtI, &DT);
   }
 
   OverflowResult computeOverflowForUnsignedMul(const Value *LHS,
                                                const Value *RHS,
                                                const Instruction *CxtI) const {
     return llvm::computeOverflowForUnsignedMul(LHS, RHS, DL, &AC, CxtI, &DT);
   }
 

llvm/lib/Analysis/ValueTracking.cpp

 //===- ValueTracking.cpp - Walk computations to compute properties --------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //

 unsigned llvm::ComputeNumSignBits(const Value *V, const DataLayout &DL,
                                   unsigned Depth, AssumptionCache *AC,
                                   const Instruction *CxtI,
                                   const DominatorTree *DT, bool UseInstrInfo) {
   return ::ComputeNumSignBits(
       V, Depth, Query(DL, AC, safeCxtI(V, CxtI), DT, UseInstrInfo));
 }
 
-unsigned llvm::ComputeMinSignedBits(const Value *V, const DataLayout &DL,
+unsigned llvm::ComputeMaxSignedBits(const Value *V, const DataLayout &DL,
                                     unsigned Depth, AssumptionCache *AC,
                                     const Instruction *CxtI,
                                     const DominatorTree *DT) {
   unsigned SignBits = ComputeNumSignBits(V, DL, Depth, AC, CxtI, DT);
   return V->getType()->getScalarSizeInBits() - SignBits + 1;
 }
 
 static void computeKnownBitsAddSub(bool Add, const Value *Op0, const Value *Op1,

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

 //===- DAGCombiner.cpp - Implement a DAG node combiner --------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
SDValue DAGCombiner::visitSIGN_EXTEND_INREG(SDNode *N) {
   if (N0.isUndef())
     return DAG.getConstant(0, SDLoc(N), VT);
 
   // fold (sext_in_reg c1) -> c1
   if (DAG.isConstantIntBuildVectorOrConstantInt(N0))
     return DAG.getNode(ISD::SIGN_EXTEND_INREG, SDLoc(N), VT, N0, N1);
 
   // If the input is already sign extended, just drop the extension.
-  if (ExtVTBits >= DAG.ComputeMinSignedBits(N0))
+  if (ExtVTBits >= DAG.ComputeMaxSignedBits(N0))
     return N0;
 
   // fold (sext_in_reg (sext_in_reg x, VT2), VT1) -> (sext_in_reg x, minVT) pt2
   if (N0.getOpcode() == ISD::SIGN_EXTEND_INREG &&
       ExtVT.bitsLT(cast<VTSDNode>(N0.getOperand(1))->getVT()))
     return DAG.getNode(ISD::SIGN_EXTEND_INREG, SDLoc(N), VT, N0.getOperand(0),
                        N1);
 
   // fold (sext_in_reg (sext x)) -> (sext x)
   // fold (sext_in_reg (aext x)) -> (sext x)
   // if x is small enough or if we know that x has more than 1 sign bit and the
   // sign_extend_inreg is extending from one of them.
   if (N0.getOpcode() == ISD::SIGN_EXTEND || N0.getOpcode() == ISD::ANY_EXTEND) {
     SDValue N00 = N0.getOperand(0);
     unsigned N00Bits = N00.getScalarValueSizeInBits();
-    if ((N00Bits <= ExtVTBits || DAG.ComputeMinSignedBits(N00) <= ExtVTBits) &&
+    if ((N00Bits <= ExtVTBits || DAG.ComputeMaxSignedBits(N00) <= ExtVTBits) &&
         (!LegalOperations || TLI.isOperationLegal(ISD::SIGN_EXTEND, VT)))
       return DAG.getNode(ISD::SIGN_EXTEND, SDLoc(N), VT, N00);
   }
 
   // fold (sext_in_reg (*_extend_vector_inreg x)) -> (sext_vector_inreg x)
   // if x is small enough or if we know that x has more than 1 sign bit and the
   // sign_extend_inreg is extending from one of them.
   if (N0.getOpcode() == ISD::ANY_EXTEND_VECTOR_INREG ||
       N0.getOpcode() == ISD::SIGN_EXTEND_VECTOR_INREG ||
       N0.getOpcode() == ISD::ZERO_EXTEND_VECTOR_INREG) {
     SDValue N00 = N0.getOperand(0);
     unsigned N00Bits = N00.getScalarValueSizeInBits();
     unsigned DstElts = N0.getValueType().getVectorMinNumElements();
     unsigned SrcElts = N00.getValueType().getVectorMinNumElements();
     bool IsZext = N0.getOpcode() == ISD::ZERO_EXTEND_VECTOR_INREG;
     APInt DemandedSrcElts = APInt::getLowBitsSet(SrcElts, DstElts);
     if ((N00Bits == ExtVTBits ||
          (!IsZext && (N00Bits < ExtVTBits ||
-                      DAG.ComputeMinSignedBits(N00) <= ExtVTBits))) &&
+                      DAG.ComputeMaxSignedBits(N00) <= ExtVTBits))) &&
         (!LegalOperations ||
          TLI.isOperationLegal(ISD::SIGN_EXTEND_VECTOR_INREG, VT)))
       return DAG.getNode(ISD::SIGN_EXTEND_VECTOR_INREG, SDLoc(N), VT, N00);
   }
 
   // fold (sext_in_reg (zext x)) -> (sext x)
   // iff we are extending the source sign bit.
   if (N0.getOpcode() == ISD::ZERO_EXTEND) {

llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp

 //===----- LegalizeIntegerTypes.cpp - Legalization of integer types -------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
void DAGTypeLegalizer::PromoteSetCCOperands(SDValue &LHS, SDValue &RHS,
   }
 
   // Prefer to promote the comparison operand with zero extension.
 
   // If this is an equality comparison and the width of OpL/OpR excluding the
   // duplicated sign bits is no greater than the width of LHS/RHS, we can avoid
   // inserting a zext_inreg operation that we might not be able to remove.
   if (ISD::isIntEqualitySetCC(CCCode)) {
-    unsigned OpLEffectiveBits = DAG.ComputeMinSignedBits(OpL);
-    unsigned OpREffectiveBits = DAG.ComputeMinSignedBits(OpR);
+    unsigned OpLEffectiveBits = DAG.ComputeMaxSignedBits(OpL);
+    unsigned OpREffectiveBits = DAG.ComputeMaxSignedBits(OpR);
     if (OpLEffectiveBits <= LHS.getScalarValueSizeInBits() &&
         OpREffectiveBits <= RHS.getScalarValueSizeInBits()) {
       LHS = OpL;
       RHS = OpR;
       return;
     }
   }
 

llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp

 //===- SelectionDAG.cpp - Implement the SelectionDAG data structures ------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
unsigned SelectionDAG::ComputeNumSignBits(SDValue Op, const APInt &DemandedElts,
   }
 
   // Finally, if we can prove that the top bits of the result are 0's or 1's,
   // use this information.
   KnownBits Known = computeKnownBits(Op, DemandedElts, Depth);
   return std::max(FirstAnswer, Known.countMinSignBits());
 }
 
-unsigned SelectionDAG::ComputeMinSignedBits(SDValue Op, unsigned Depth) const {
+unsigned SelectionDAG::ComputeMaxSignedBits(SDValue Op, unsigned Depth) const {
   unsigned SignBits = ComputeNumSignBits(Op, Depth);
   return Op.getScalarValueSizeInBits() - SignBits + 1;
 }
 
-unsigned SelectionDAG::ComputeMinSignedBits(SDValue Op,
+unsigned SelectionDAG::ComputeMaxSignedBits(SDValue Op,
                                             const APInt &DemandedElts,
                                             unsigned Depth) const {
   unsigned SignBits = ComputeNumSignBits(Op, DemandedElts, Depth);
   return Op.getScalarValueSizeInBits() - SignBits + 1;
 }
 
 bool SelectionDAG::isGuaranteedNotToBeUndefOrPoison(SDValue Op, bool PoisonOnly,
                                                     unsigned Depth) const {

llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp

 //===-- TargetLowering.cpp - Implement the TargetLowering class -----------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
bool TargetLowering::SimplifyDemandedBits(
   case ISD::SIGN_EXTEND_INREG: {
     SDValue Op0 = Op.getOperand(0);
     EVT ExVT = cast<VTSDNode>(Op.getOperand(1))->getVT();
     unsigned ExVTBits = ExVT.getScalarSizeInBits();
 
     // If we only care about the highest bit, don't bother shifting right.
     if (DemandedBits.isSignMask()) {
       unsigned MinSignedBits =
-          TLO.DAG.ComputeMinSignedBits(Op0, DemandedElts, Depth + 1);
+          TLO.DAG.ComputeMaxSignedBits(Op0, DemandedElts, Depth + 1);
       bool AlreadySignExtended = ExVTBits >= MinSignedBits;
       // However if the input is already sign extended we expect the sign
       // extension to be dropped altogether later and do not simplify.
       if (!AlreadySignExtended) {
         // Compute the correct shift amount type, which must be getShiftAmountTy
         // for scalar types after legalization.
         EVT ShiftAmtTy = VT;
         if (TLO.LegalTypes() && !ShiftAmtTy.isVector())

llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp

 //===-- AMDGPUISelLowering.cpp - AMDGPU Common DAG lowering functions -----===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //

 
 unsigned AMDGPUTargetLowering::numBitsUnsigned(SDValue Op, SelectionDAG &DAG) {
   return DAG.computeKnownBits(Op).countMaxActiveBits();
 }
 
 unsigned AMDGPUTargetLowering::numBitsSigned(SDValue Op, SelectionDAG &DAG) {
   // In order for this to be a signed 24-bit value, bit 23, must
   // be a sign bit.
-  return DAG.ComputeMinSignedBits(Op);
+  return DAG.ComputeMaxSignedBits(Op);
 }
 
 AMDGPUTargetLowering::AMDGPUTargetLowering(const TargetMachine &TM,
                                            const AMDGPUSubtarget &STI)
     : TargetLowering(TM), Subtarget(&STI) {
   // Lower floating point store/load to integer store/load to reduce the number
   // of patterns in tablegen.
   setOperationAction(ISD::LOAD, MVT::f32, Promote);

llvm/lib/Target/X86/X86ISelLowering.cpp

 //===-- X86ISelLowering.cpp - X86 DAG Lowering Implementation -------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
static SDValue getPack(SelectionDAG &DAG, const X86Subtarget &Subtarget,
 
   // See if we already have sufficient leading bits for PACKSS/PACKUS.
   if (!PackHiHalf) {
     if (UsePackUS &&
         DAG.computeKnownBits(LHS).countMaxActiveBits() <= EltSizeInBits &&
         DAG.computeKnownBits(RHS).countMaxActiveBits() <= EltSizeInBits)
       return DAG.getNode(X86ISD::PACKUS, dl, VT, LHS, RHS);
 
-    if (DAG.ComputeMinSignedBits(LHS) <= EltSizeInBits &&
-        DAG.ComputeMinSignedBits(RHS) <= EltSizeInBits)
+    if (DAG.ComputeMaxSignedBits(LHS) <= EltSizeInBits &&
+        DAG.ComputeMaxSignedBits(RHS) <= EltSizeInBits)
       return DAG.getNode(X86ISD::PACKSS, dl, VT, LHS, RHS);
   }
 
   // Fallback to sign/zero extending the requested half and pack.
   SDValue Amt = DAG.getTargetConstant(EltSizeInBits, dl, MVT::i8);
   if (UsePackUS) {
     if (PackHiHalf) {
       LHS = DAG.getNode(X86ISD::VSRLI, dl, OpVT, LHS, Amt);
if (CmpVT == MVT::i16 && !Subtarget.isAtom() &&
     if ((COp0 && !COp0->getAPIntValue().isSignedIntN(8)) ||
         (COp1 && !COp1->getAPIntValue().isSignedIntN(8))) {
       unsigned ExtendOp =
           isX86CCSigned(X86CC) ? ISD::SIGN_EXTEND : ISD::ZERO_EXTEND;
       if (X86CC == X86::COND_E || X86CC == X86::COND_NE) {
         // For equality comparisons try to use SIGN_EXTEND if the input was
         // truncate from something with enough sign bits.
         if (Op0.getOpcode() == ISD::TRUNCATE) {
-          if (DAG.ComputeMinSignedBits(Op0.getOperand(0)) <= 16)
+          if (DAG.ComputeMaxSignedBits(Op0.getOperand(0)) <= 16)
             ExtendOp = ISD::SIGN_EXTEND;
         } else if (Op1.getOpcode() == ISD::TRUNCATE) {
-          if (DAG.ComputeMinSignedBits(Op1.getOperand(0)) <= 16)
+          if (DAG.ComputeMaxSignedBits(Op1.getOperand(0)) <= 16)
             ExtendOp = ISD::SIGN_EXTEND;
         }
       }
 
       CmpVT = MVT::i32;
       Op0 = DAG.getNode(ExtendOp, dl, CmpVT, Op0);
       Op1 = DAG.getNode(ExtendOp, dl, CmpVT, Op1);
     }
static SDValue combineMulToPMADDWD(SDNode *N, SelectionDAG &DAG,
   if (!Subtarget.hasSSE41() &&
       (N0.getOpcode() == ISD::SIGN_EXTEND &&
        N0.getOperand(0).getValueSizeInBits() > 128) &&
       (N1.getOpcode() == ISD::SIGN_EXTEND &&
        N1.getOperand(0).getValueSizeInBits() > 128))
     return SDValue();
 
   // Sign bits must extend down to the lowest i16.
-  if (DAG.ComputeMinSignedBits(N1) > 16 || DAG.ComputeMinSignedBits(N0) > 16)
+  if (DAG.ComputeMaxSignedBits(N1) > 16 || DAG.ComputeMaxSignedBits(N0) > 16)
     return SDValue();
 
   // At least one of the elements must be zero in the upper 17 bits, or can be
   // safely made zero without altering the final result.
   auto GetZeroableOp = [&](SDValue Op) {
     APInt Mask17 = APInt::getHighBitsSet(32, 17);
     if (DAG.MaskedValueIsZero(Op, Mask17))
       return Op;
static SDValue combinePMULH(SDValue Src, EVT VT, const SDLoc &DL,
 
   SDValue LHS = Src.getOperand(0).getOperand(0);
   SDValue RHS = Src.getOperand(0).getOperand(1);
 
   // Count leading sign/zero bits on both inputs - if there are enough then
   // truncation back to vXi16 will be cheap - either as a pack/shuffle
   // sequence or using AVX512 truncations. If the inputs are sext/zext then the
   // truncations may actually be free by peeking through to the ext source.
   auto IsSext = [&DAG](SDValue V) {
-    return DAG.ComputeMinSignedBits(V) <= 16;
+    return DAG.ComputeMaxSignedBits(V) <= 16;
   };
   auto IsZext = [&DAG](SDValue V) {
     return DAG.computeKnownBits(V).countMaxActiveBits() <= 16;
   };
 
   bool IsSigned = IsSext(LHS) && IsSext(RHS);
   bool IsUnsigned = IsZext(LHS) && IsZext(RHS);
   if (!IsSigned && !IsUnsigned)

llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp

 //===- InstCombineCompares.cpp --------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
static Instruction *processUGT_ADDCST_ADD(ICmpInst &I, Value *A, Value *B,
   // Check to see that CI1 is an all-ones value with NewWidth bits.
   if (CI1->getBitWidth() == NewWidth ||
       CI1->getValue() != APInt::getLowBitsSet(CI1->getBitWidth(), NewWidth))
     return nullptr;
 
   // This is only really a signed overflow check if the inputs have been
   // sign-extended; check for that condition. For example, if CI2 is 2^31 and
   // the operands of the add are 64 bits wide, we need at least 33 sign bits.
-  if (IC.ComputeMinSignedBits(A, 0, &I) > NewWidth ||
-      IC.ComputeMinSignedBits(B, 0, &I) > NewWidth)
+  if (IC.ComputeMaxSignedBits(A, 0, &I) > NewWidth ||
+      IC.ComputeMaxSignedBits(B, 0, &I) > NewWidth)
     return nullptr;
 
   // In order to replace the original add with a narrower
   // llvm.sadd.with.overflow, the only uses allowed are the add-with-constant
   // and truncates that discard the high bits of the add.  Verify that this is
   // the case.
   Instruction *OrigAdd = cast<Instruction>(AddWithCst->getOperand(0));
   for (User *U : OrigAdd->users()) {

llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp

 //===- InstCombineSelect.cpp ----------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
if (AddSub->getOpcode() == Instruction::Add)
     IntrinsicID = Intrinsic::sadd_sat;
   else if (AddSub->getOpcode() == Instruction::Sub)
     IntrinsicID = Intrinsic::ssub_sat;
   else
     return nullptr;
 
   // The two operands of the add/sub must be nsw-truncatable to the NewTy. This
   // is usually achieved via a sext from a smaller type.
-  if (ComputeMinSignedBits(AddSub->getOperand(0), 0, AddSub) > NewBitWidth ||
-      ComputeMinSignedBits(AddSub->getOperand(1), 0, AddSub) > NewBitWidth)
+  if (ComputeMaxSignedBits(AddSub->getOperand(0), 0, AddSub) > NewBitWidth ||
+      ComputeMaxSignedBits(AddSub->getOperand(1), 0, AddSub) > NewBitWidth)
     return nullptr;
 
   // Finally create and return the sat intrinsic, truncated to the new type
   Function *F = Intrinsic::getDeclaration(MinMax1.getModule(), IntrinsicID, NewTy);
   Value *AT = Builder.CreateTrunc(AddSub->getOperand(0), NewTy);
   Value *BT = Builder.CreateTrunc(AddSub->getOperand(1), NewTy);
   Value *Sat = Builder.CreateCall(F, {AT, BT});
   return CastInst::Create(Instruction::SExt, Sat, Ty);