[LegalizeTypes][VP] Add integer promotions support for VP_TRUNCATE

[Jimerlife]

Summary

Add promoteRes and promoteOp for VP_TRUNCATE.

I couldn't write tests about VP_TRUNCATE when TargetLowering::TypeWidenVector in RISCV, so I reserved a TODO.

[craig.topper]

llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp

1358

Shouldn't we have if and else instead of two ifs?

1392

Use an if and two returns instead of ?: operator.

2098

Use an if and two returns.

[Jimerlife]

Address comments

[Jimerlife]

llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp

1358

Done. Use if and else instead of two ifs

1392

Done.

2098

Done

[craig.topper]

LGTM

llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp

#endif
   case ISD::SHL:
   case ISD::VP_SHL:      Res = PromoteIntRes_SHL(N); break;
   case ISD::SIGN_EXTEND_INREG:
                          Res = PromoteIntRes_SIGN_EXTEND_INREG(N); break;
   case ISD::SRA:
   case ISD::VP_ASHR:     Res = PromoteIntRes_SRA(N); break;
   case ISD::SRL:
   case ISD::VP_LSHR:     Res = PromoteIntRes_SRL(N); break;
+  case ISD::VP_TRUNCATE:
   case ISD::TRUNCATE:    Res = PromoteIntRes_TRUNCATE(N); break;
   case ISD::UNDEF:       Res = PromoteIntRes_UNDEF(N); break;
   case ISD::VAARG:       Res = PromoteIntRes_VAARG(N); break;
   case ISD::VSCALE:      Res = PromoteIntRes_VSCALE(N); break;
 
   case ISD::EXTRACT_SUBVECTOR:
                          Res = PromoteIntRes_EXTRACT_SUBVECTOR(N); break;
   case ISD::INSERT_SUBVECTOR:
SDValue DAGTypeLegalizer::PromoteIntRes_FunnelShift(SDNode *N) {
 
   return DAG.getNode(Opcode, DL, VT, Hi, Lo, Amt);
 }
 
 SDValue DAGTypeLegalizer::PromoteIntRes_TRUNCATE(SDNode *N) {
   EVT NVT = TLI.getTypeToTransformTo(*DAG.getContext(), N->getValueType(0));
   SDValue Res;
   SDValue InOp = N->getOperand(0);
+  unsigned Opcode = N->getOpcode();
   SDLoc dl(N);
 
   switch (getTypeAction(InOp.getValueType())) {
   default: llvm_unreachable("Unknown type action!");
   case TargetLowering::TypeLegal:
   case TargetLowering::TypeExpandInteger:
     Res = InOp;
     break;
case TargetLowering::TypeSplitVector: {
     assert(isPowerOf2_32(NumElts.getKnownMinValue()) &&
            "Promoted vector type must be a power of two");
 
     SDValue EOp1, EOp2;
     GetSplitVector(InOp, EOp1, EOp2);
 
     EVT HalfNVT = EVT::getVectorVT(*DAG.getContext(), NVT.getScalarType(),
                                    NumElts.divideCoefficientBy(2));
+    if (Opcode == ISD::TRUNCATE) {
       EOp1 = DAG.getNode(ISD::TRUNCATE, dl, HalfNVT, EOp1);
       EOp2 = DAG.getNode(ISD::TRUNCATE, dl, HalfNVT, EOp2);
+    }
 
+    if (Opcode == ISD::VP_TRUNCATE) {
+      SDValue MaskLo, MaskHi, EVLLo, EVLHi;
+      std::tie(MaskLo, MaskHi) = SplitMask(N->getOperand(1));
+      std::tie(EVLLo, EVLHi) =
+          DAG.SplitEVL(N->getOperand(2), N->getValueType(0), dl);
+      EOp1 = DAG.getNode(ISD::VP_TRUNCATE, dl, HalfNVT, EOp1, MaskLo, EVLLo);
+      EOp2 = DAG.getNode(ISD::VP_TRUNCATE, dl, HalfNVT, EOp2, MaskHi, EVLHi);
+    }
     return DAG.getNode(ISD::CONCAT_VECTORS, dl, NVT, EOp1, EOp2);
   }
+  // TODO: VP_TRUNCATE need to handle if necessary for some targets.
   case TargetLowering::TypeWidenVector: {
     SDValue WideInOp = GetWidenedVector(InOp);
 
     // Truncate widened InOp.
     unsigned NumElem = WideInOp.getValueType().getVectorNumElements();
     EVT TruncVT = EVT::getVectorVT(*DAG.getContext(),
                                    N->getValueType(0).getScalarType(), NumElem);
     SDValue WideTrunc = DAG.getNode(ISD::TRUNCATE, dl, TruncVT, WideInOp);
 
     // Zero extend so that the elements are of same type as those of NVT
     EVT ExtVT = EVT::getVectorVT(*DAG.getContext(), NVT.getVectorElementType(),
                                  NumElem);
     SDValue WideExt = DAG.getNode(ISD::ZERO_EXTEND, dl, ExtVT, WideTrunc);
 
     // Extract the low NVT subvector.
     SDValue ZeroIdx = DAG.getVectorIdxConstant(0, dl);
     return DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, NVT, WideExt, ZeroIdx);
   }
   }
 
   // Truncate to NVT instead of VT
-  return DAG.getNode(ISD::TRUNCATE, dl, NVT, Res);
+  return Opcode == ISD::VP_TRUNCATE
+             ? DAG.getNode(Opcode, dl, NVT, Res, N->getOperand(1),
+                           N->getOperand(2))
+             : DAG.getNode(Opcode, dl, NVT, Res);
 }
 
 SDValue DAGTypeLegalizer::PromoteIntRes_UADDSUBO(SDNode *N, unsigned ResNo) {
   if (ResNo == 1)
     return PromoteIntRes_Overflow(N);
 
   // The operation overflowed iff the result in the larger type is not the
   // zero extension of its truncation to the original type.
bool DAGTypeLegalizer::PromoteIntegerOperand(SDNode *N, unsigned OpNo) {
   case ISD::MSTORE:       Res = PromoteIntOp_MSTORE(cast<MaskedStoreSDNode>(N),
                                                     OpNo); break;
   case ISD::MLOAD:        Res = PromoteIntOp_MLOAD(cast<MaskedLoadSDNode>(N),
                                                     OpNo); break;
   case ISD::MGATHER:  Res = PromoteIntOp_MGATHER(cast<MaskedGatherSDNode>(N),
                                                  OpNo); break;
   case ISD::MSCATTER: Res = PromoteIntOp_MSCATTER(cast<MaskedScatterSDNode>(N),
                                                   OpNo); break;
+  case ISD::VP_TRUNCATE:
   case ISD::TRUNCATE:     Res = PromoteIntOp_TRUNCATE(N); break;
   case ISD::FP16_TO_FP:
   case ISD::UINT_TO_FP:   Res = PromoteIntOp_UINT_TO_FP(N); break;
   case ISD::STRICT_UINT_TO_FP:  Res = PromoteIntOp_STRICT_UINT_TO_FP(N); break;
   case ISD::ZERO_EXTEND:  Res = PromoteIntOp_ZERO_EXTEND(N); break;
   case ISD::EXTRACT_SUBVECTOR: Res = PromoteIntOp_EXTRACT_SUBVECTOR(N); break;
   case ISD::INSERT_SUBVECTOR: Res = PromoteIntOp_INSERT_SUBVECTOR(N); break;
 
SDValue DAGTypeLegalizer::PromoteIntOp_MSCATTER(MaskedScatterSDNode *N,
 
   return DAG.getMaskedScatter(DAG.getVTList(MVT::Other), N->getMemoryVT(),
                               SDLoc(N), NewOps, N->getMemOperand(),
                               N->getIndexType(), TruncateStore);
 }
 
 SDValue DAGTypeLegalizer::PromoteIntOp_TRUNCATE(SDNode *N) {
   SDValue Op = GetPromotedInteger(N->getOperand(0));
-  return DAG.getNode(ISD::TRUNCATE, SDLoc(N), N->getValueType(0), Op);
+  return N->getOpcode() == ISD::VP_TRUNCATE
+             ? DAG.getNode(ISD::VP_TRUNCATE, SDLoc(N), N->getValueType(0), Op,
+                           N->getOperand(1), N->getOperand(2))
+             : DAG.getNode(ISD::TRUNCATE, SDLoc(N), N->getValueType(0), Op);
 }
 
 SDValue DAGTypeLegalizer::PromoteIntOp_UINT_TO_FP(SDNode *N) {
   return SDValue(DAG.UpdateNodeOperands(N,
                                 ZExtPromotedInteger(N->getOperand(0))), 0);
 }
 
 SDValue DAGTypeLegalizer::PromoteIntOp_STRICT_UINT_TO_FP(SDNode *N) {

llvm/test/CodeGen/RISCV/rvv/fixed-vector-trunc-vp.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -mattr=+v -riscv-v-vector-bits-min=128 -verify-machineinstrs < %s | FileCheck %s
 ; RUN: llc -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-min=128 -verify-machineinstrs < %s | FileCheck %s
 
+declare <2 x i7> @llvm.vp.trunc.nxv2i7.nxv2i16(<2 x i16>, <2 x i1>, i32)
+
+define <2 x i7> @vtrunc_nxv2i7_nxv2i16(<2 x i16> %a, <2 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vtrunc_nxv2i7_nxv2i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e8, mf8, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v8, v8, v0.t
+; CHECK-NEXT:    ret
+  %v = call <2 x i7> @llvm.vp.trunc.nxv2i7.nxv2i16(<2 x i16> %a, <2 x i1> %m, i32 %vl)
+  ret <2 x i7> %v
+}
+
+declare <2 x i8> @llvm.vp.trunc.nxv2i8.nxv2i15(<2 x i15>, <2 x i1>, i32)
+
+define <2 x i8> @vtrunc_nxv2i8_nxv2i15(<2 x i15> %a, <2 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vtrunc_nxv2i8_nxv2i15:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e8, mf8, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v8, v8, v0.t
+; CHECK-NEXT:    ret
+  %v = call <2 x i8> @llvm.vp.trunc.nxv2i8.nxv2i15(<2 x i15> %a, <2 x i1> %m, i32 %vl)
+  ret <2 x i8> %v
+}
+
 declare <2 x i8> @llvm.vp.trunc.nxv2i8.nxv2i16(<2 x i16>, <2 x i1>, i32)
 
 define <2 x i8> @vtrunc_nxv2i8_nxv2i16(<2 x i16> %a, <2 x i1> %m, i32 zeroext %vl) {
 ; CHECK-LABEL: vtrunc_nxv2i8_nxv2i16:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vsetvli zero, a0, e8, mf8, ta, mu
 ; CHECK-NEXT:    vncvt.x.x.w v8, v8, v0.t
 ; CHECK-NEXT:    ret
   %v = call <2 x i8> @llvm.vp.trunc.nxv2i8.nxv2i16(<2 x i16> %a, <2 x i1> %m, i32 %vl)
   ret <2 x i8> %v
 }
 
 define <2 x i8> @vtrunc_nxv2i8_nxv2i16_unmasked(<2 x i16> %a, i32 zeroext %vl) {
 ; CHECK-LABEL: vtrunc_nxv2i8_nxv2i16_unmasked:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vsetvli zero, a0, e8, mf8, ta, mu
 ; CHECK-NEXT:    vncvt.x.x.w v8, v8
 ; CHECK-NEXT:    ret
   %v = call <2 x i8> @llvm.vp.trunc.nxv2i8.nxv2i16(<2 x i16> %a, <2 x i1> shufflevector (<2 x i1> insertelement (<2 x i1> undef, i1 true, i32 0), <2 x i1> undef, <2 x i32> zeroinitializer), i32 %vl)
   ret <2 x i8> %v
 }
 
+declare <128 x i7> @llvm.vp.trunc.nxv128i7.nxv128i16(<128 x i16>, <128 x i1>, i32)
+
+define <128 x i7> @vtrunc_nxv128i7_nxv128i16(<128 x i16> %a, <128 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vtrunc_nxv128i7_nxv128i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi sp, sp, -16
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    csrr a1, vlenb
+; CHECK-NEXT:    slli a1, a1, 3
+; CHECK-NEXT:    sub sp, sp, a1
+; CHECK-NEXT:    vmv1r.v v24, v0
+; CHECK-NEXT:    addi a1, sp, 16
+; CHECK-NEXT:    vs8r.v v8, (a1) # Unknown-size Folded Spill
+; CHECK-NEXT:    li a1, 0
+; CHECK-NEXT:    vsetivli zero, 8, e8, m1, ta, mu
+; CHECK-NEXT:    addi a2, a0, -64
+; CHECK-NEXT:    vslidedown.vi v0, v0, 8
+; CHECK-NEXT:    bltu a0, a2, .LBB4_2
+; CHECK-NEXT:  # %bb.1:
+; CHECK-NEXT:    mv a1, a2
+; CHECK-NEXT:  .LBB4_2:
+; CHECK-NEXT:    vsetvli zero, a1, e8, m4, ta, mu
+; CHECK-NEXT:    li a1, 64
+; CHECK-NEXT:    vncvt.x.x.w v8, v16, v0.t
+; CHECK-NEXT:    bltu a0, a1, .LBB4_4
+; CHECK-NEXT:  # %bb.3:
+; CHECK-NEXT:    li a0, 64
+; CHECK-NEXT:  .LBB4_4:
+; CHECK-NEXT:    vsetvli zero, a0, e8, m4, ta, mu
+; CHECK-NEXT:    vmv1r.v v0, v24
+; CHECK-NEXT:    addi a0, sp, 16
+; CHECK-NEXT:    vl8re8.v v24, (a0) # Unknown-size Folded Reload
+; CHECK-NEXT:    vncvt.x.x.w v16, v24, v0.t
+; CHECK-NEXT:    li a0, 128
+; CHECK-NEXT:    vsetvli zero, a0, e8, m8, tu, mu
+; CHECK-NEXT:    vslideup.vx v16, v8, a1
+; CHECK-NEXT:    vmv8r.v v8, v16
+; CHECK-NEXT:    csrr a0, vlenb
+; CHECK-NEXT:    slli a0, a0, 3
+; CHECK-NEXT:    add sp, sp, a0
+; CHECK-NEXT:    addi sp, sp, 16
+; CHECK-NEXT:    ret
+  %v = call <128 x i7> @llvm.vp.trunc.nxv128i7.nxv128i16(<128 x i16> %a, <128 x i1> %m, i32 %vl)
+  ret <128 x i7> %v
+}
+
 declare <2 x i8> @llvm.vp.trunc.nxv2i8.nxv2i32(<2 x i32>, <2 x i1>, i32)
 
 define <2 x i8> @vtrunc_nxv2i8_nxv2i32(<2 x i32> %a, <2 x i1> %m, i32 zeroext %vl) {
 ; CHECK-LABEL: vtrunc_nxv2i8_nxv2i32:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vsetvli zero, a0, e16, mf4, ta, mu
 ; CHECK-NEXT:    vncvt.x.x.w v8, v8, v0.t
 ; CHECK-NEXT:    vsetvli zero, zero, e8, mf8, ta, mu

llvm/test/CodeGen/RISCV/rvv/vtrunc-vp.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -mattr=+v -verify-machineinstrs < %s | FileCheck %s
 ; RUN: llc -mtriple=riscv64 -mattr=+v -verify-machineinstrs < %s | FileCheck %s
 
+declare <vscale x 2 x i7> @llvm.vp.trunc.nxv2i7.nxv2i16(<vscale x 2 x i16>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x i7> @vtrunc_nxv2i7_nxv2i16(<vscale x 2 x i16> %a, <vscale x 2 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vtrunc_nxv2i7_nxv2i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e8, mf4, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v8, v8, v0.t
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i7> @llvm.vp.trunc.nxv2i7.nxv2i16(<vscale x 2 x i16> %a, <vscale x 2 x i1> %m, i32 %vl)
+  ret <vscale x 2 x i7> %v
+}
+
+declare <vscale x 2 x i8> @llvm.vp.trunc.nxv2i8.nxv2i15(<vscale x 2 x i15>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x i8> @vtrunc_nxv2i8_nxv2i15(<vscale x 2 x i15> %a, <vscale x 2 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vtrunc_nxv2i8_nxv2i15:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e8, mf4, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v8, v8, v0.t
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i8> @llvm.vp.trunc.nxv2i8.nxv2i15(<vscale x 2 x i15> %a, <vscale x 2 x i1> %m, i32 %vl)
+  ret <vscale x 2 x i8> %v
+}
+
 declare <vscale x 2 x i8> @llvm.vp.trunc.nxv2i8.nxv2i16(<vscale x 2 x i16>, <vscale x 2 x i1>, i32)
 
 define <vscale x 2 x i8> @vtrunc_nxv2i8_nxv2i16(<vscale x 2 x i16> %a, <vscale x 2 x i1> %m, i32 zeroext %vl) {
 ; CHECK-LABEL: vtrunc_nxv2i8_nxv2i16:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vsetvli zero, a0, e8, mf4, ta, mu
 ; CHECK-NEXT:    vncvt.x.x.w v8, v8, v0.t
 ; CHECK-NEXT:    ret

 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
 ; CHECK-NEXT:    vncvt.x.x.w v10, v8
 ; CHECK-NEXT:    vmv.v.v v8, v10
 ; CHECK-NEXT:    ret
   %v = call <vscale x 2 x i32> @llvm.vp.trunc.nxv2i64.nxv2i32(<vscale x 2 x i64> %a, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %vl)
   ret <vscale x 2 x i32> %v
 }
+
+declare <vscale x 32 x i7> @llvm.vp.trunc.nxv32i7.nxv32i32(<vscale x 32 x i32>, <vscale x 32 x i1>, i32)
+
+define <vscale x 32 x i7> @vtrunc_nxv32i7_nxv32i32(<vscale x 32 x i32> %a, <vscale x 32 x i1> %m, i32 zeroext %vl) {
+; CHECK-LABEL: vtrunc_nxv32i7_nxv32i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vmv1r.v v24, v0
+; CHECK-NEXT:    li a2, 0
+; CHECK-NEXT:    csrr a1, vlenb
+; CHECK-NEXT:    srli a4, a1, 2
+; CHECK-NEXT:    vsetvli a3, zero, e8, mf2, ta, mu
+; CHECK-NEXT:    slli a1, a1, 1
+; CHECK-NEXT:    sub a3, a0, a1
+; CHECK-NEXT:    vslidedown.vx v0, v0, a4
+; CHECK-NEXT:    bltu a0, a3, .LBB14_2
+; CHECK-NEXT:  # %bb.1:
+; CHECK-NEXT:    mv a2, a3
+; CHECK-NEXT:  .LBB14_2:
+; CHECK-NEXT:    vsetvli zero, a2, e16, m4, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v28, v16, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e8, m2, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v18, v28, v0.t
+; CHECK-NEXT:    bltu a0, a1, .LBB14_4
+; CHECK-NEXT:  # %bb.3:
+; CHECK-NEXT:    mv a0, a1
+; CHECK-NEXT:  .LBB14_4:
+; CHECK-NEXT:    vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT:    vmv1r.v v0, v24
+; CHECK-NEXT:    vncvt.x.x.w v20, v8, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e8, m2, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v16, v20, v0.t
+; CHECK-NEXT:    vmv4r.v v8, v16
+; CHECK-NEXT:    ret
+  %v = call <vscale x 32 x i7> @llvm.vp.trunc.nxv32i7.nxv32i32(<vscale x 32 x i32> %a, <vscale x 32 x i1> %m, i32 %vl)
+  ret <vscale x 32 x i7> %v
+}