AArch64: Use TargetConstant for intrinsic IDs #166661

arsenm · 2025-11-05T23:01:23Z

These should always use TargetConstant

arsenm · 2025-11-05T23:01:37Z

AArch64: Use TargetConstant for intrinsic IDs #166661 👈 (View in Graphite)
main

This stack of pull requests is managed by Graphite. Learn more about stacking.

llvmbot · 2025-11-05T23:01:58Z

@llvm/pr-subscribers-backend-aarch64

Author: Matt Arsenault (arsenm)

Changes

These should always use TargetConstant

Full diff: https://github.com/llvm/llvm-project/pull/166661.diff

1 Files Affected:

(modified) llvm/lib/Target/AArch64/AArch64ISelLowering.cpp (+47-41)

diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index 298746863d221..10464bb5e50a9 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -5544,9 +5544,10 @@ SDValue AArch64TargetLowering::LowerGET_ROUNDING(SDValue Op,
   SDLoc DL(Op);
 
   SDValue Chain = Op.getOperand(0);
-  SDValue FPCR_64 = DAG.getNode(
-      ISD::INTRINSIC_W_CHAIN, DL, {MVT::i64, MVT::Other},
-      {Chain, DAG.getConstant(Intrinsic::aarch64_get_fpcr, DL, MVT::i64)});
+  SDValue FPCR_64 =
+      DAG.getNode(ISD::INTRINSIC_W_CHAIN, DL, {MVT::i64, MVT::Other},
+                  {Chain, DAG.getTargetConstant(Intrinsic::aarch64_get_fpcr, DL,
+                                                MVT::i64)});
   Chain = FPCR_64.getValue(1);
   SDValue FPCR_32 = DAG.getNode(ISD::TRUNCATE, DL, MVT::i32, FPCR_64);
   SDValue FltRounds = DAG.getNode(ISD::ADD, DL, MVT::i32, FPCR_32,
@@ -5632,7 +5633,8 @@ SDValue AArch64TargetLowering::LowerSET_FPMODE(SDValue Op,
 
   // Set new value of FPCR.
   SDValue Ops2[] = {
-      Chain, DAG.getConstant(Intrinsic::aarch64_set_fpcr, DL, MVT::i64), FPCR};
+      Chain, DAG.getTargetConstant(Intrinsic::aarch64_set_fpcr, DL, MVT::i64),
+      FPCR};
   return DAG.getNode(ISD::INTRINSIC_VOID, DL, MVT::Other, Ops2);
 }
 
@@ -5655,9 +5657,9 @@ SDValue AArch64TargetLowering::LowerRESET_FPMODE(SDValue Op,
       DAG.getConstant(AArch64::ReservedFPControlBits, DL, MVT::i64));
 
   // Set new value of FPCR.
-  SDValue Ops2[] = {Chain,
-                    DAG.getConstant(Intrinsic::aarch64_set_fpcr, DL, MVT::i64),
-                    FPSCRMasked};
+  SDValue Ops2[] = {
+      Chain, DAG.getTargetConstant(Intrinsic::aarch64_set_fpcr, DL, MVT::i64),
+      FPSCRMasked};
   return DAG.getNode(ISD::INTRINSIC_VOID, DL, MVT::Other, Ops2);
 }
 
@@ -7289,17 +7291,19 @@ SDValue AArch64TargetLowering::LowerVECTOR_COMPRESS(SDValue Op,
 
   SDValue Compressed = DAG.getNode(
       ISD::INTRINSIC_WO_CHAIN, DL, Vec.getValueType(),
-      DAG.getConstant(Intrinsic::aarch64_sve_compact, DL, MVT::i64), Mask, Vec);
+      DAG.getTargetConstant(Intrinsic::aarch64_sve_compact, DL, MVT::i64), Mask,
+      Vec);
 
   // compact fills with 0s, so if our passthru is all 0s, do nothing here.
   if (HasPassthru && !ISD::isConstantSplatVectorAllZeros(Passthru.getNode())) {
     SDValue Offset = DAG.getNode(
         ISD::INTRINSIC_WO_CHAIN, DL, MVT::i64,
-        DAG.getConstant(Intrinsic::aarch64_sve_cntp, DL, MVT::i64), Mask, Mask);
+        DAG.getTargetConstant(Intrinsic::aarch64_sve_cntp, DL, MVT::i64), Mask,
+        Mask);
 
     SDValue IndexMask = DAG.getNode(
         ISD::INTRINSIC_WO_CHAIN, DL, MaskVT,
-        DAG.getConstant(Intrinsic::aarch64_sve_whilelo, DL, MVT::i64),
+        DAG.getTargetConstant(Intrinsic::aarch64_sve_whilelo, DL, MVT::i64),
         DAG.getConstant(0, DL, MVT::i64), Offset);
 
     Compressed =
@@ -7428,10 +7432,10 @@ static SDValue LowerFLDEXP(SDValue Op, SelectionDAG &DAG) {
                              DAG.getUNDEF(ExpVT), Exp, Zero);
   SDValue VPg = getPTrue(DAG, DL, XVT.changeVectorElementType(MVT::i1),
                          AArch64SVEPredPattern::all);
-  SDValue FScale =
-      DAG.getNode(ISD::INTRINSIC_WO_CHAIN, DL, XVT,
-                  DAG.getConstant(Intrinsic::aarch64_sve_fscale, DL, MVT::i64),
-                  VPg, VX, VExp);
+  SDValue FScale = DAG.getNode(
+      ISD::INTRINSIC_WO_CHAIN, DL, XVT,
+      DAG.getTargetConstant(Intrinsic::aarch64_sve_fscale, DL, MVT::i64), VPg,
+      VX, VExp);
   SDValue Final =
       DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, X.getValueType(), FScale, Zero);
   if (X.getValueType() != XScalarTy)
@@ -8094,7 +8098,7 @@ static SDValue emitRestoreZALazySave(SDValue Chain, SDLoc DL,
       TLI.getLibcallName(LC), TLI.getPointerTy(DAG.getDataLayout()));
   SDValue TPIDR2_EL0 = DAG.getNode(
       ISD::INTRINSIC_W_CHAIN, DL, MVT::i64, Chain,
-      DAG.getConstant(Intrinsic::aarch64_sme_get_tpidr2, DL, MVT::i32));
+      DAG.getTargetConstant(Intrinsic::aarch64_sme_get_tpidr2, DL, MVT::i32));
   // Copy the address of the TPIDR2 block into X0 before 'calling' the
   // RESTORE_ZA pseudo.
   SDValue Glue;
@@ -8109,7 +8113,7 @@ static SDValue emitRestoreZALazySave(SDValue Chain, SDLoc DL,
   // Finally reset the TPIDR2_EL0 register to 0.
   Chain = DAG.getNode(
       ISD::INTRINSIC_VOID, DL, MVT::Other, Chain,
-      DAG.getConstant(Intrinsic::aarch64_sme_set_tpidr2, DL, MVT::i32),
+      DAG.getTargetConstant(Intrinsic::aarch64_sme_set_tpidr2, DL, MVT::i32),
       DAG.getConstant(0, DL, MVT::i64));
   TPIDR2.Uses++;
   return Chain;
@@ -8704,7 +8708,7 @@ SDValue AArch64TargetLowering::LowerFormalArguments(
     if (Attrs.isNewZT0())
       Chain = DAG.getNode(
           ISD::INTRINSIC_VOID, DL, MVT::Other, Chain,
-          DAG.getConstant(Intrinsic::aarch64_sme_zero_zt, DL, MVT::i32),
+          DAG.getTargetConstant(Intrinsic::aarch64_sme_zero_zt, DL, MVT::i32),
           DAG.getTargetConstant(0, DL, MVT::i32));
   }
 
@@ -9517,7 +9521,7 @@ AArch64TargetLowering::LowerCall(CallLoweringInfo &CLI,
         DAG.getTargetLoweringInfo().getFrameIndexTy(DAG.getDataLayout()));
     Chain = DAG.getNode(
         ISD::INTRINSIC_VOID, DL, MVT::Other, Chain,
-        DAG.getConstant(Intrinsic::aarch64_sme_set_tpidr2, DL, MVT::i32),
+        DAG.getTargetConstant(Intrinsic::aarch64_sme_set_tpidr2, DL, MVT::i32),
         TPIDR2ObjAddr);
     OptimizationRemarkEmitter ORE(&MF.getFunction());
     ORE.emit([&]() {
@@ -13408,8 +13412,8 @@ SDValue ReconstructShuffleWithRuntimeMask(SDValue Op, SelectionDAG &DAG) {
 
   return DAG.getNode(
       ISD::INTRINSIC_WO_CHAIN, DL, VT,
-      DAG.getConstant(Intrinsic::aarch64_neon_tbl1, DL, MVT::i32), SourceVec,
-      MaskSourceVec);
+      DAG.getTargetConstant(Intrinsic::aarch64_neon_tbl1, DL, MVT::i32),
+      SourceVec, MaskSourceVec);
 }
 
 // Gather data to see if the operation can be modelled as a
@@ -14265,14 +14269,16 @@ static SDValue GenerateTBL(SDValue Op, ArrayRef<int> ShuffleMask,
       V1Cst = DAG.getNode(ISD::CONCAT_VECTORS, DL, MVT::v16i8, V1Cst, V1Cst);
     Shuffle = DAG.getNode(
         ISD::INTRINSIC_WO_CHAIN, DL, IndexVT,
-        DAG.getConstant(Intrinsic::aarch64_neon_tbl1, DL, MVT::i32), V1Cst,
+        DAG.getTargetConstant(Intrinsic::aarch64_neon_tbl1, DL, MVT::i32),
+        V1Cst,
         DAG.getBuildVector(IndexVT, DL, ArrayRef(TBLMask.data(), IndexLen)));
   } else {
     if (IndexLen == 8) {
       V1Cst = DAG.getNode(ISD::CONCAT_VECTORS, DL, MVT::v16i8, V1Cst, V2Cst);
       Shuffle = DAG.getNode(
           ISD::INTRINSIC_WO_CHAIN, DL, IndexVT,
-          DAG.getConstant(Intrinsic::aarch64_neon_tbl1, DL, MVT::i32), V1Cst,
+          DAG.getTargetConstant(Intrinsic::aarch64_neon_tbl1, DL, MVT::i32),
+          V1Cst,
           DAG.getBuildVector(IndexVT, DL, ArrayRef(TBLMask.data(), IndexLen)));
     } else {
       // FIXME: We cannot, for the moment, emit a TBL2 instruction because we
@@ -14283,8 +14289,8 @@ static SDValue GenerateTBL(SDValue Op, ArrayRef<int> ShuffleMask,
       //                   IndexLen));
       Shuffle = DAG.getNode(
           ISD::INTRINSIC_WO_CHAIN, DL, IndexVT,
-          DAG.getConstant(Intrinsic::aarch64_neon_tbl2, DL, MVT::i32), V1Cst,
-          V2Cst,
+          DAG.getTargetConstant(Intrinsic::aarch64_neon_tbl2, DL, MVT::i32),
+          V1Cst, V2Cst,
           DAG.getBuildVector(IndexVT, DL, ArrayRef(TBLMask.data(), IndexLen)));
     }
   }
@@ -16437,10 +16443,10 @@ SDValue AArch64TargetLowering::LowerVectorSRA_SRL_SHL(SDValue Op,
     if (isVShiftLImm(Op.getOperand(1), VT, false, Cnt) && Cnt < EltSize)
       return DAG.getNode(AArch64ISD::VSHL, DL, VT, Op.getOperand(0),
                          DAG.getTargetConstant(Cnt, DL, MVT::i32));
-    return DAG.getNode(ISD::INTRINSIC_WO_CHAIN, DL, VT,
-                       DAG.getConstant(Intrinsic::aarch64_neon_ushl, DL,
-                                       MVT::i32),
-                       Op.getOperand(0), Op.getOperand(1));
+    return DAG.getNode(
+        ISD::INTRINSIC_WO_CHAIN, DL, VT,
+        DAG.getTargetConstant(Intrinsic::aarch64_neon_ushl, DL, MVT::i32),
+        Op.getOperand(0), Op.getOperand(1));
   case ISD::SRA:
   case ISD::SRL:
     if (VT.isScalableVector() &&
@@ -20160,7 +20166,7 @@ static SDValue performFpToIntCombine(SDNode *N, SelectionDAG &DAG,
                                       : Intrinsic::aarch64_neon_vcvtfp2fxu;
   SDValue FixConv =
       DAG.getNode(ISD::INTRINSIC_WO_CHAIN, DL, ResTy,
-                  DAG.getConstant(IntrinsicOpcode, DL, MVT::i32),
+                  DAG.getTargetConstant(IntrinsicOpcode, DL, MVT::i32),
                   Op->getOperand(0), DAG.getTargetConstant(C, DL, MVT::i32));
   // We can handle smaller integers by generating an extra trunc.
   if (IntBits < FloatBits)
@@ -27416,8 +27422,8 @@ static SDValue combineSVEPrefetchVecBaseImmOff(SDNode *N, SelectionDAG &DAG,
   // ...and remap the intrinsic `aarch64_sve_prf<T>_gather_scalar_offset` to
   // `aarch64_sve_prfb_gather_uxtw_index`.
   SDLoc DL(N);
-  Ops[1] = DAG.getConstant(Intrinsic::aarch64_sve_prfb_gather_uxtw_index, DL,
-                           MVT::i64);
+  Ops[1] = DAG.getTargetConstant(Intrinsic::aarch64_sve_prfb_gather_uxtw_index,
+                                 DL, MVT::i64);
 
   return DAG.getNode(N->getOpcode(), DL, DAG.getVTList(MVT::Other), Ops);
 }
@@ -31251,10 +31257,10 @@ static SDValue GenerateFixedLengthSVETBL(SDValue Op, SDValue Op1, SDValue Op2,
 
   SDValue Shuffle;
   if (IsSingleOp)
-    Shuffle =
-        DAG.getNode(ISD::INTRINSIC_WO_CHAIN, DL, ContainerVT,
-                    DAG.getConstant(Intrinsic::aarch64_sve_tbl, DL, MVT::i32),
-                    Op1, SVEMask);
+    Shuffle = DAG.getNode(
+        ISD::INTRINSIC_WO_CHAIN, DL, ContainerVT,
+        DAG.getTargetConstant(Intrinsic::aarch64_sve_tbl, DL, MVT::i32), Op1,
+        SVEMask);
   else if (Subtarget.hasSVE2()) {
     if (!MinMaxEqual) {
       unsigned MinNumElts = AArch64::SVEBitsPerBlock / BitsPerElt;
@@ -31273,10 +31279,10 @@ static SDValue GenerateFixedLengthSVETBL(SDValue Op, SDValue Op1, SDValue Op2,
       SVEMask = convertToScalableVector(
           DAG, getContainerForFixedLengthVector(DAG, MaskType), UpdatedVecMask);
     }
-    Shuffle =
-        DAG.getNode(ISD::INTRINSIC_WO_CHAIN, DL, ContainerVT,
-                    DAG.getConstant(Intrinsic::aarch64_sve_tbl2, DL, MVT::i32),
-                    Op1, Op2, SVEMask);
+    Shuffle = DAG.getNode(
+        ISD::INTRINSIC_WO_CHAIN, DL, ContainerVT,
+        DAG.getTargetConstant(Intrinsic::aarch64_sve_tbl2, DL, MVT::i32), Op1,
+        Op2, SVEMask);
   }
   Shuffle = convertFromScalableVector(DAG, VT, Shuffle);
   return DAG.getNode(ISD::BITCAST, DL, Op.getValueType(), Shuffle);
@@ -31436,8 +31442,8 @@ SDValue AArch64TargetLowering::LowerFixedLengthVECTOR_SHUFFLEToSVE(
       unsigned SegmentElts = VT.getVectorNumElements() / Segments;
       if (std::optional<unsigned> Lane =
               isDUPQMask(ShuffleMask, Segments, SegmentElts)) {
-        SDValue IID =
-            DAG.getConstant(Intrinsic::aarch64_sve_dup_laneq, DL, MVT::i64);
+        SDValue IID = DAG.getTargetConstant(Intrinsic::aarch64_sve_dup_laneq,
+                                            DL, MVT::i64);
         return convertFromScalableVector(
             DAG, VT,
             DAG.getNode(ISD::INTRINSIC_WO_CHAIN, DL, ContainerVT,

AArch64: Use TargetConstant for intrinsic IDs

6bdf7e0

These should always use TargetConstant

arsenm added the backend:AArch64 label Nov 5, 2025 — with Graphite App

arsenm marked this pull request as ready for review November 5, 2025 23:01

arsenm merged commit bbf62dc into main Nov 11, 2025
14 checks passed

arsenm deleted the users/arsenm/aarch64/use-targetconstant-intrinsic-id branch November 11, 2025 22:51

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

AArch64: Use TargetConstant for intrinsic IDs #166661

AArch64: Use TargetConstant for intrinsic IDs #166661

arsenm commented Nov 5, 2025

Uh oh!

arsenm commented Nov 5, 2025

Uh oh!

llvmbot commented Nov 5, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

AArch64: Use TargetConstant for intrinsic IDs #166661

AArch64: Use TargetConstant for intrinsic IDs #166661

Conversation

arsenm commented Nov 5, 2025

Uh oh!

arsenm commented Nov 5, 2025

Uh oh!

llvmbot commented Nov 5, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants