[RISCV] Add RV64F codegen support

[asb]

Summary

This requires a little extra work due tothe fact i32 is not a legal type. When call lowering happens post-legalisation (e.g. when an intrinsic was inserted during legalisation). A bitcast from f32
to i32 can't be introduced. This is similar to the challenges with RV32D. To handle this, we introduce target-specific DAG nodes that perform bitcast+sext for f32->i64 and trunc+bitcast for i64->f32.

[asb]

The presence of no-op any_extend can cause assertions during the DAG combiner, and there's no guarantee that the no-op any_extend will be visited first. I'll update this to avoid the issue, probably setting a target combiner for any_extend to legalise any_extend + bitcast directly to BitcastAndSextF32ToI64.

[efriedma]

lib/CodeGen/SelectionDAG/DAGCombiner.cpp

8959 ↗

(On Diff #169532)

Not sure how you're ending up in a situation where this matters; getNode() will normally fold away an ANY_EXTEND where the operand and result have the same type.

lib/Target/RISCV/RISCVISelDAGToDAG.cpp

161 ↗

(On Diff #169532)

The point of this is to try to generate a "narrower" FP_TO_SINT where possible? Does this transform actually improve performance?

lib/Target/RISCV/RISCVInstrInfoF.td

354 ↗

(On Diff #169532)

Maybe better to implement this as a DAGCombine?

[asb]

Updated to use target DAG combines for the int/float conversions.

[rogfer01]

lib/Target/RISCV/RISCVInstrInfoF.td

352 ↗

(On Diff #171079)

Typo → because

[asb]

Refresh patch and fix typo in comment. This patch does not update float-intrinsics.ll as new promotion code needs to be added for FPOWI. I will post a patch to do this now.

[asb]

Submitting some old in-line comments that were accidentally left unsubmitted.

Patch still applies cleanly.

lib/CodeGen/SelectionDAG/DAGCombiner.cpp

8959 ↗

(On Diff #169532)

This fold is insufficient anyway. The issue is that anyextend is left hanging around after one of its children is custom legalised. Just handling everything in a target dag combine avoids this and other problems.

lib/Target/RISCV/RISCVISelDAGToDAG.cpp

161 ↗

(On Diff #169532)

The alternative is fcvt.l.s + sign-extension of the result. Just emitting fcvt.w.s is fewer instructions.

[asb]

Rebased on top of D56264 and D53230. Although this patch could be written to apply without those changes, given all patches modify the dag combines it's simpler to keep them linear.

I'd really appreciate a review. From my perspective, this is ready to land.

[asb]

Rebased (no meaningful changes).

I would really, really appreciate a review. I believe this is ready to land and overall a pretty straight-forward patch. The introduction of target-specific nodes to avoid bitcasts to illegal types is used across other backends in LLVM, as well as for RV32D and is pretty straight-forward.

[rogfer01]

Hi Alex,

then change looks sensible to me but I found there is some sort of issue with the new DAG nodes. Consider the following example (minimised from CodeGen/Generic/2008-02-25-NegateZero.ll.

define void @test(float* %p) {
entry:
	%tmp98 = load float, float* %p, align 4
	call void (i32, ...) @foo( i32 0, float %tmp98 ) nounwind 
	ret void
}

declare void @foo(i32, ...)

$ ./bin/llc -mtriple riscv64 < t.ll

	.text
	.file	"<stdin>"
SoftenFloatOperand Op #0: t21: i64 = RISCVISD::BitcastAndSextF32ToI64 t5

Do not know how to soften this operator's operand!
UNREACHABLE executed at /home/rferrer/fio/upstream/llvm-src/lib/CodeGen/SelectionDAG/LegalizeFloatTypes.cpp:753!
Stack dump:
0.	Program arguments: ./bin/llc -mtriple riscv64 
1.	Running pass 'Function Pass Manager' on module '<stdin>'.
2.	Running pass 'RISCV DAG->DAG Pattern Instruction Selection' on function '@test'

I think this happens because f32 is not legal when no f is present, so the type legalizer tries to "soften" it but it doesn't know how to do this.

[rogfer01]

Ah, I think what happens is that we should not emit such patterns without the presence of the F extension: these patterns are ultimately selected as instructions in F. Does this make sense?

[asb]

In D53235#1356067, @rogfer01 wrote:

Ah, I think what happens is that we should not emit such patterns without the presence of the F extension: these patterns are ultimately selected as instructions in F. Does this make sense?

Thanks, the code in PerformDAGCombine is incorrectly assuming it will only see f32 if the F extension is enabled. But of course, that's only the case if you're in a post-legalisation combine call. I'll fix and add a test case.

[efriedma]

lib/Target/RISCV/RISCVISelDAGToDAG.cpp

161 ↗

(On Diff #169532)

I still don't follow; AssertSext doesn't lower to any instructions. So the alternative is just the one fcvt.l.s. (The "trick" here is that fptosi returns poison if the result would be out of range.)

[asb]

Updated to address all outstanding feedback. The combines are now properly guarded and a test case verified this. I've removed the logic that prefers fcvt.w[u].s and added exhaustive tests for sext/zext/aext of the results.

Looks good?

[asb]

lib/Target/RISCV/RISCVISelDAGToDAG.cpp

161 ↗

(On Diff #169532)

Thanks for pointing out that fptosi with an out-of-range input is poison. The same is true for fptoui. I've removed this logic.

[lewis-revill]

lib/Target/RISCV/RISCVISelLowering.cpp

636 ↗

(On Diff #182851)

Typo 'redundant'

test/CodeGen/RISCV/float-arith.ll

148 ↗

(On Diff #182851)

Can RV64 employ the same technique as RV32 here, and for other similar DAG combines? I'm guessing cost-wise it would be desirable to do so.

[jrtc27]

test/CodeGen/RISCV/float-arith.ll

148 ↗

(On Diff #182851)

It should work, since fmv.x.w sign-extends:

For RV64, the higher 32 bits of the destination register are filled with copies of the floating-point number’s sign bit.

and the lui will also sign-extend, so xor will flip all the higher bits too.

[asb]

Changes since last time:

• Use custom legalisation rather than DAG combines to produce the f32 <-> i64 nodes
• Add target DAG combines that perform the equivalent of the (bitcast (fneg/fabs x)) conversion that the DAGCombiner does
• Rename target-specific SelectionDAG nodes to match the RISC-V fmv instruction naming.

test/CodeGen/RISCV/float-arith.ll

148 ↗

(On Diff #182851)

I've split out float-bitmanip-dagcombines.ll and double-bitmanip-dagcombines.ll to better demonstrate the opportunities here. One RV32F, target-independent code provides the combine for us. As we need to use RISCVISD::FMV_X_W_RV64 (as bitcasting to i32 wouldn't be legal), we need to provide an equivalent combine in the RISC-V backend. I've updated the patch to do so.

[asb]

Attached the wrong version of the patch, which had a simple issue with the dagcombine. Right version now attached.

[asb]

Clearly everyone is too polite to point out that I attached the RV64D patch in the last update rather than RV64F. Now addressed.

Reviews would be appreciated.

[efriedma]

lib/Target/RISCV/RISCVISelLowering.cpp

666 ↗

(On Diff #184254)

This combine isn't consistent with the ComputeNumSignBitsForTargetNode change: if value is supposed to be sign-extended, you need to insert a SIGN_EXTEND_INREG

[asb]

Hi Eli. Thanks, as always, for the careful review. I've opted not to to introduce a SEXT_IN_REG in this case, and instead define a target-specific node that produces an anyext result (and a pattern that ensures FMV_X_W is selected for (sexti32 (riscv_fmv_x_anyextw_rv64 ..)) meaning I can replace with an ANY_EXTEND. Otherwise there are cases where we end up with an unwanted sext.w in cases where redundant GPR->FPR->GPR moves have been eliminated.

For this reason I think this is the more useful semantic, and I believe it's safe based on my understanding of the legalisation+widening expectations but I'd really welcome a second opinion.

[Comment was edited for clarity after submission].

[efriedma]

LGTM

Making FMV_X_ANYEXTW_RV64 any-extend the result seems fine.

lib/Target/RISCV/RISCVInstrInfoF.td

372 ↗

(On Diff #184533)

This is supposed to be assertzexti32, as opposed to zexti32, right? I doubt it has any practical impact because an "AND" would be dead, but better to be clear.

[asb]

lib/Target/RISCV/RISCVInstrInfoF.td

372 ↗

(On Diff #184533)

Yes, the AND would be dead but as you suggest it's better to stick with assertzexti32 so people don't need to evaluate whether the AND pattern is dead and unnecessary when reading the code.

I'll adjust when committing - thanks.

llvm/trunk/lib/Target/RISCV/RISCVISelLowering.h

enum NodeType : unsigned {
   // instructions.
   SLLW,
   SRAW,
   SRLW,
   // 32-bit operations from RV64M that can't be simply matched with a pattern
   // at instruction selection time.
   DIVW,
   DIVUW,
-  REMUW
+  REMUW,
+  // FPR32<->GPR transfer operations for RV64. Needed as an i32<->f32 bitcast
+  // is not legal on RV64. FMV_W_X_RV64 matches the semantics of the FMV.W.X.
+  // FMV_X_ANYEXTW_RV64 is similar to FMV.X.W but has an any-extended result.
+  // This is a more convenient semantic for producing dagcombines that remove
+  // unnecessary GPR->FPR->GPR moves.
+  FMV_W_X_RV64,
+  FMV_X_ANYEXTW_RV64
 };
 }
 
 class RISCVTargetLowering : public TargetLowering {
   const RISCVSubtarget &Subtarget;
 
 public:
   explicit RISCVTargetLowering(const TargetMachine &TM,

llvm/trunk/lib/Target/RISCV/RISCVISelLowering.cpp

for (auto CC : FPCCToExtend)
       setCondCodeAction(CC, MVT::f32, Expand);
     setOperationAction(ISD::SELECT_CC, MVT::f32, Expand);
     setOperationAction(ISD::SELECT, MVT::f32, Custom);
     setOperationAction(ISD::BR_CC, MVT::f32, Expand);
     for (auto Op : FPOpToExtend)
       setOperationAction(Op, MVT::f32, Expand);
   }
 
+  if (Subtarget.hasStdExtF() && Subtarget.is64Bit())
+    setOperationAction(ISD::BITCAST, MVT::i32, Custom);
+
   if (Subtarget.hasStdExtD()) {
     setOperationAction(ISD::FMINNUM, MVT::f64, Legal);
     setOperationAction(ISD::FMAXNUM, MVT::f64, Legal);
     for (auto CC : FPCCToExtend)
       setCondCodeAction(CC, MVT::f64, Expand);
     setOperationAction(ISD::SELECT_CC, MVT::f64, Expand);
     setOperationAction(ISD::SELECT, MVT::f64, Custom);
     setOperationAction(ISD::BR_CC, MVT::f64, Expand);
SDValue RISCVTargetLowering::LowerOperation(SDValue Op,
   case ISD::SELECT:
     return lowerSELECT(Op, DAG);
   case ISD::VASTART:
     return lowerVASTART(Op, DAG);
   case ISD::FRAMEADDR:
     return lowerFRAMEADDR(Op, DAG);
   case ISD::RETURNADDR:
     return lowerRETURNADDR(Op, DAG);
+  case ISD::BITCAST: {
+    assert(Subtarget.is64Bit() && Subtarget.hasStdExtF() &&
+           "Unexpected custom legalisation");
+    SDLoc DL(Op);
+    SDValue Op0 = Op.getOperand(0);
+    if (Op.getValueType() != MVT::f32 || Op0.getValueType() != MVT::i32)
+      return SDValue();
+    SDValue NewOp0 = DAG.getNode(ISD::ANY_EXTEND, DL, MVT::i64, Op0);
+    SDValue FPConv = DAG.getNode(RISCVISD::FMV_W_X_RV64, DL, MVT::f32, NewOp0);
+    return FPConv;
+  }
   }
 }
 
 SDValue RISCVTargetLowering::lowerGlobalAddress(SDValue Op,
                                                 SelectionDAG &DAG) const {
   SDLoc DL(Op);
   EVT Ty = Op.getValueType();
   GlobalAddressSDNode *N = cast<GlobalAddressSDNode>(Op);
void RISCVTargetLowering::ReplaceNodeResults(SDNode *N,
   case ISD::UREM:
     assert(N->getValueType(0) == MVT::i32 && Subtarget.is64Bit() &&
            Subtarget.hasStdExtM() && "Unexpected custom legalisation");
     if (N->getOperand(0).getOpcode() == ISD::Constant ||
         N->getOperand(1).getOpcode() == ISD::Constant)
       return;
     Results.push_back(customLegalizeToWOp(N, DAG));
     break;
+  case ISD::BITCAST: {
+    assert(N->getValueType(0) == MVT::i32 && Subtarget.is64Bit() &&
+           Subtarget.hasStdExtF() && "Unexpected custom legalisation");
+    SDLoc DL(N);
+    SDValue Op0 = N->getOperand(0);
+    if (Op0.getValueType() != MVT::f32)
+      return;
+    SDValue FPConv =
+        DAG.getNode(RISCVISD::FMV_X_ANYEXTW_RV64, DL, MVT::i64, Op0);
+    Results.push_back(DAG.getNode(ISD::TRUNCATE, DL, MVT::i32, FPConv));
+    break;
+  }
   }
 }
 
 SDValue RISCVTargetLowering::PerformDAGCombine(SDNode *N,
                                                DAGCombinerInfo &DCI) const {
   SelectionDAG &DAG = DCI.DAG;
 
   switch (N->getOpcode()) {
case RISCVISD::SRLW: {
     SDValue RHS = N->getOperand(1);
     APInt LHSMask = APInt::getLowBitsSet(LHS.getValueSizeInBits(), 32);
     APInt RHSMask = APInt::getLowBitsSet(RHS.getValueSizeInBits(), 5);
     if ((SimplifyDemandedBits(N->getOperand(0), LHSMask, DCI)) ||
         (SimplifyDemandedBits(N->getOperand(1), RHSMask, DCI)))
       return SDValue();
     break;
   }
+  case RISCVISD::FMV_X_ANYEXTW_RV64: {
+    SDLoc DL(N);
+    SDValue Op0 = N->getOperand(0);
+    // If the input to FMV_X_ANYEXTW_RV64 is just FMV_W_X_RV64 then the
+    // conversion is unnecessary and can be replaced with an ANY_EXTEND
+    // of the FMV_W_X_RV64 operand.
+    if (Op0->getOpcode() == RISCVISD::FMV_W_X_RV64) {
+      SDValue AExtOp =
+          DAG.getNode(ISD::ANY_EXTEND, DL, MVT::i64, Op0.getOperand(0));
+      return DCI.CombineTo(N, AExtOp);
+    }
+
+    // This is a target-specific version of a DAGCombine performed in
+    // DAGCombiner::visitBITCAST. It performs the equivalent of:
+    // fold (bitconvert (fneg x)) -> (xor (bitconvert x), signbit)
+    // fold (bitconvert (fabs x)) -> (and (bitconvert x), (not signbit))
+    if (!(Op0.getOpcode() == ISD::FNEG || Op0.getOpcode() == ISD::FABS) ||
+        !Op0.getNode()->hasOneUse())
+      break;
+    SDValue NewFMV = DAG.getNode(RISCVISD::FMV_X_ANYEXTW_RV64, DL, MVT::i64,
+                                 Op0.getOperand(0));
+    APInt SignBit = APInt::getSignMask(32).sext(64);
+    if (Op0.getOpcode() == ISD::FNEG) {
+      return DCI.CombineTo(N,
+                           DAG.getNode(ISD::XOR, DL, MVT::i64, NewFMV,
+                                       DAG.getConstant(SignBit, DL, MVT::i64)));
+    }
+    assert(Op0.getOpcode() == ISD::FABS);
+    return DCI.CombineTo(N,
+                         DAG.getNode(ISD::AND, DL, MVT::i64, NewFMV,
+                                     DAG.getConstant(~SignBit, DL, MVT::i64)));
+  }
   }
 
   return SDValue();
 }
 
 unsigned RISCVTargetLowering::ComputeNumSignBitsForTargetNode(
     SDValue Op, const APInt &DemandedElts, const SelectionDAG &DAG,
     unsigned Depth) const {

 // Implements the RISC-V calling convention. Returns true upon failure.
 static bool CC_RISCV(const DataLayout &DL, unsigned ValNo, MVT ValVT, MVT LocVT,
                      CCValAssign::LocInfo LocInfo, ISD::ArgFlagsTy ArgFlags,
                      CCState &State, bool IsFixed, bool IsRet, Type *OrigTy) {
   unsigned XLen = DL.getLargestLegalIntTypeSizeInBits();
   assert(XLen == 32 || XLen == 64);
   MVT XLenVT = XLen == 32 ? MVT::i32 : MVT::i64;
   if (ValVT == MVT::f32) {
-    LocVT = MVT::i32;
+    LocVT = XLenVT;
     LocInfo = CCValAssign::BCvt;
   }
 
   // Any return value split in to more than two values can't be returned
   // directly.
   if (IsRet && ValNo > 1)
     return true;
 

 static SDValue convertLocVTToValVT(SelectionDAG &DAG, SDValue Val,
                                    const CCValAssign &VA, const SDLoc &DL) {
   switch (VA.getLocInfo()) {
   default:
     llvm_unreachable("Unexpected CCValAssign::LocInfo");
   case CCValAssign::Full:
     break;
   case CCValAssign::BCvt:
+    if (VA.getLocVT() == MVT::i64 && VA.getValVT() == MVT::f32) {
+      Val = DAG.getNode(RISCVISD::FMV_W_X_RV64, DL, MVT::f32, Val);
+      break;
+    }
     Val = DAG.getNode(ISD::BITCAST, DL, VA.getValVT(), Val);
     break;
   }
   return Val;
 }
 
 // The caller is responsible for loading the full value if the argument is
 // passed with CCValAssign::Indirect.
static SDValue convertValVTToLocVT(SelectionDAG &DAG, SDValue Val,
   EVT LocVT = VA.getLocVT();
 
   switch (VA.getLocInfo()) {
   default:
     llvm_unreachable("Unexpected CCValAssign::LocInfo");
   case CCValAssign::Full:
     break;
   case CCValAssign::BCvt:
+    if (VA.getLocVT() == MVT::i64 && VA.getValVT() == MVT::f32) {
+      Val = DAG.getNode(RISCVISD::FMV_X_ANYEXTW_RV64, DL, MVT::i64, Val);
+      break;
+    }
     Val = DAG.getNode(ISD::BITCAST, DL, LocVT, Val);
     break;
   }
   return Val;
 }
 
 // The caller is responsible for loading the full value if the argument is
 // passed with CCValAssign::Indirect.
static SDValue unpackFromMemLoc(SelectionDAG &DAG, SDValue Chain,
   SDValue Val;
 
   ISD::LoadExtType ExtType;
   switch (VA.getLocInfo()) {
   default:
     llvm_unreachable("Unexpected CCValAssign::LocInfo");
   case CCValAssign::Full:
   case CCValAssign::Indirect:
+  case CCValAssign::BCvt:
     ExtType = ISD::NON_EXTLOAD;
     break;
   }
+  if (ValVT == MVT::f32)
+    LocVT = MVT::f32;
   Val = DAG.getExtLoad(
       ExtType, DL, LocVT, Chain, FIN,
       MachinePointerInfo::getFixedStack(DAG.getMachineFunction(), FI), ValVT);
   return Val;
 }
 
 static SDValue unpackF64OnRV32DSoftABI(SelectionDAG &DAG, SDValue Chain,
                                        const CCValAssign &VA, const SDLoc &DL) {
const char *RISCVTargetLowering::getTargetNodeName(unsigned Opcode) const {
   case RISCVISD::SRLW:
     return "RISCVISD::SRLW";
   case RISCVISD::DIVW:
     return "RISCVISD::DIVW";
   case RISCVISD::DIVUW:
     return "RISCVISD::DIVUW";
   case RISCVISD::REMUW:
     return "RISCVISD::REMUW";
+  case RISCVISD::FMV_W_X_RV64:
+    return "RISCVISD::FMV_W_X_RV64";
+  case RISCVISD::FMV_X_ANYEXTW_RV64:
+    return "RISCVISD::FMV_X_ANYEXTW_RV64";
   }
   return nullptr;
 }
 
 std::pair<unsigned, const TargetRegisterClass *>
 RISCVTargetLowering::getRegForInlineAsmConstraint(const TargetRegisterInfo *TRI,
                                                   StringRef Constraint,
                                                   MVT VT) const {

llvm/trunk/lib/Target/RISCV/RISCVInstrInfoF.td

 //===-- RISCVInstrInfoF.td - RISC-V 'F' instructions -------*- tablegen -*-===//
 //
 // 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
 //
 //===----------------------------------------------------------------------===//
 //
 // This file describes the RISC-V instructions from the standard 'F',
 // Single-Precision Floating-Point instruction set extension.
 //
 //===----------------------------------------------------------------------===//
 
 //===----------------------------------------------------------------------===//
+// RISC-V specific DAG Nodes.
+//===----------------------------------------------------------------------===//
+
+def SDT_RISCVFMV_W_X_RV64
+    : SDTypeProfile<1, 1, [SDTCisVT<0, f32>, SDTCisVT<1, i64>]>;
+def SDT_RISCVFMV_X_ANYEXTW_RV64
+    : SDTypeProfile<1, 1, [SDTCisVT<0, i64>, SDTCisVT<1, f32>]>;
+
+def riscv_fmv_w_x_rv64
+    : SDNode<"RISCVISD::FMV_W_X_RV64", SDT_RISCVFMV_W_X_RV64>;
+def riscv_fmv_x_anyextw_rv64
+    : SDNode<"RISCVISD::FMV_X_ANYEXTW_RV64", SDT_RISCVFMV_X_ANYEXTW_RV64>;
+
+//===----------------------------------------------------------------------===//
 // Operand and SDNode transformation definitions.
 //===----------------------------------------------------------------------===//
 
 // Floating-point rounding mode
 
 def FRMArg : AsmOperandClass {
   let Name = "FRMArg";
   let RenderMethod = "addFRMArgOperands";

 // float->[u]int. Round-to-zero must be used.
 def : Pat<(fp_to_sint FPR32:$rs1), (FCVT_W_S $rs1, 0b001)>;
 def : Pat<(fp_to_uint FPR32:$rs1), (FCVT_WU_S $rs1, 0b001)>;
 
 // [u]int->float. Match GCC and default to using dynamic rounding mode.
 def : Pat<(sint_to_fp GPR:$rs1), (FCVT_S_W $rs1, 0b111)>;
 def : Pat<(uint_to_fp GPR:$rs1), (FCVT_S_WU $rs1, 0b111)>;
 } // Predicates = [HasStdExtF, IsRV32]
+
+let Predicates = [HasStdExtF, IsRV32] in {
+// FP->[u]int. Round-to-zero must be used
+def : Pat<(fp_to_sint FPR32:$rs1), (FCVT_W_S $rs1, 0b001)>;
+def : Pat<(fp_to_uint FPR32:$rs1), (FCVT_WU_S $rs1, 0b001)>;
+
+// [u]int->fp. Match GCC and default to using dynamic rounding mode.
+def : Pat<(sint_to_fp GPR:$rs1), (FCVT_S_W $rs1, 0b111)>;
+def : Pat<(uint_to_fp GPR:$rs1), (FCVT_S_WU $rs1, 0b111)>;
+} // Predicates = [HasStdExtF, IsRV32]
+
+let Predicates = [HasStdExtF, IsRV64] in {
+def : Pat<(riscv_fmv_w_x_rv64 GPR:$src), (FMV_W_X GPR:$src)>;
+def : Pat<(riscv_fmv_x_anyextw_rv64 FPR32:$src), (FMV_X_W FPR32:$src)>;
+def : Pat<(sexti32 (riscv_fmv_x_anyextw_rv64 FPR32:$src)),
+          (FMV_X_W FPR32:$src)>;
+
+// FP->[u]int32 is mostly handled by the FP->[u]int64 patterns. This is safe
+// because fpto[u|s]i produces poison if the value can't fit into the target.
+// We match the single case below because fcvt.wu.s sign-extends its result so
+// is cheaper than fcvt.lu.s+sext.w.
+def : Pat<(sext_inreg (assertzexti32 (fp_to_uint FPR32:$rs1)), i32),
+          (FCVT_WU_S $rs1, 0b001)>;
+
+// FP->[u]int64
+def : Pat<(fp_to_sint FPR32:$rs1), (FCVT_L_S $rs1, 0b001)>;
+def : Pat<(fp_to_uint FPR32:$rs1), (FCVT_LU_S $rs1, 0b001)>;
+
+// [u]int->fp. Match GCC and default to using dynamic rounding mode.
+def : Pat<(sint_to_fp (sext_inreg GPR:$rs1, i32)), (FCVT_S_W $rs1, 0b111)>;
+def : Pat<(uint_to_fp (zexti32 GPR:$rs1)), (FCVT_S_WU $rs1, 0b111)>;
+def : Pat<(sint_to_fp GPR:$rs1), (FCVT_S_L $rs1, 0b111)>;
+def : Pat<(uint_to_fp GPR:$rs1), (FCVT_S_LU $rs1, 0b111)>;
+} // Predicates = [HasStdExtF, IsRV64]

llvm/trunk/test/CodeGen/RISCV/float-arith.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -mattr=+f -verify-machineinstrs < %s \
 ; RUN:   | FileCheck -check-prefix=RV32IF %s
+; RUN: llc -mtriple=riscv64 -mattr=+f -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefix=RV64IF %s
 
 ; These tests are each targeted at a particular RISC-V FPU instruction. Most
 ; other files in this folder exercise LLVM IR instructions that don't directly
 ; match a RISC-V instruction.
 
 define float @fadd_s(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fadd_s:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    fadd.s ft0, ft1, ft0
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fadd_s:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    fadd.s ft0, ft1, ft0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = fadd float %a, %b
   ret float %1
 }
 
 define float @fsub_s(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fsub_s:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    fsub.s ft0, ft1, ft0
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fsub_s:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    fsub.s ft0, ft1, ft0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = fsub float %a, %b
   ret float %1
 }
 
 define float @fmul_s(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fmul_s:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    fmul.s ft0, ft1, ft0
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fmul_s:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    fmul.s ft0, ft1, ft0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = fmul float %a, %b
   ret float %1
 }
 
 define float @fdiv_s(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fdiv_s:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    fdiv.s ft0, ft1, ft0
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fdiv_s:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    fdiv.s ft0, ft1, ft0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = fdiv float %a, %b
   ret float %1
 }
 
 declare float @llvm.sqrt.f32(float)
 
 define float @fsqrt_s(float %a) nounwind {
 ; RV32IF-LABEL: fsqrt_s:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    fsqrt.s ft0, ft0
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fsqrt_s:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fsqrt.s ft0, ft0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = call float @llvm.sqrt.f32(float %a)
   ret float %1
 }
 
 declare float @llvm.copysign.f32(float, float)
 
 define float @fsgnj_s(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fsgnj_s:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    fsgnj.s ft0, ft1, ft0
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fsgnj_s:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    fsgnj.s ft0, ft1, ft0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = call float @llvm.copysign.f32(float %a, float %b)
   ret float %1
 }
 
 ; This function performs extra work to ensure that
 ; DAGCombiner::visitBITCAST doesn't replace the fneg with an xor.
 define i32 @fneg_s(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fneg_s:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    fadd.s ft0, ft0, ft0
 ; RV32IF-NEXT:    fneg.s ft1, ft0
 ; RV32IF-NEXT:    feq.s a0, ft0, ft1
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fneg_s:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fadd.s ft0, ft0, ft0
+; RV64IF-NEXT:    fneg.s ft1, ft0
+; RV64IF-NEXT:    feq.s a0, ft0, ft1
+; RV64IF-NEXT:    ret
   %1 = fadd float %a, %a
   %2 = fneg float %1
   %3 = fcmp oeq float %1, %2
   %4 = zext i1 %3 to i32
   ret i32 %4
 }
 
 ; This function performs extra work to ensure that
 ; DAGCombiner::visitBITCAST doesn't replace the fneg with an xor.
 define float @fsgnjn_s(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fsgnjn_s:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    fadd.s ft0, ft1, ft0
 ; RV32IF-NEXT:    fsgnjn.s ft0, ft1, ft0
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fsgnjn_s:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    fadd.s ft0, ft1, ft0
+; RV64IF-NEXT:    fsgnjn.s ft0, ft1, ft0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = fadd float %a, %b
   %2 = fneg float %1
   %3 = call float @llvm.copysign.f32(float %a, float %2)
   ret float %3
 }
 
 declare float @llvm.fabs.f32(float)
 
 ; This function performs extra work to ensure that
 ; DAGCombiner::visitBITCAST doesn't replace the fabs with an and.
 define float @fabs_s(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fabs_s:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    fadd.s ft0, ft1, ft0
 ; RV32IF-NEXT:    fabs.s ft1, ft0
 ; RV32IF-NEXT:    fadd.s ft0, ft1, ft0
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fabs_s:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    fadd.s ft0, ft1, ft0
+; RV64IF-NEXT:    fabs.s ft1, ft0
+; RV64IF-NEXT:    fadd.s ft0, ft1, ft0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = fadd float %a, %b
   %2 = call float @llvm.fabs.f32(float %1)
   %3 = fadd float %2, %1
   ret float %3
 }
 
 declare float @llvm.minnum.f32(float, float)
 
 define float @fmin_s(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fmin_s:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    fmin.s ft0, ft1, ft0
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fmin_s:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    fmin.s ft0, ft1, ft0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = call float @llvm.minnum.f32(float %a, float %b)
   ret float %1
 }
 
 declare float @llvm.maxnum.f32(float, float)
 
 define float @fmax_s(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fmax_s:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    fmax.s ft0, ft1, ft0
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fmax_s:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    fmax.s ft0, ft1, ft0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = call float @llvm.maxnum.f32(float %a, float %b)
   ret float %1
 }
 
 define i32 @feq_s(float %a, float %b) nounwind {
 ; RV32IF-LABEL: feq_s:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    feq.s a0, ft1, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: feq_s:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    feq.s a0, ft1, ft0
+; RV64IF-NEXT:    ret
   %1 = fcmp oeq float %a, %b
   %2 = zext i1 %1 to i32
   ret i32 %2
 }
 
 define i32 @flt_s(float %a, float %b) nounwind {
 ; RV32IF-LABEL: flt_s:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    flt.s a0, ft1, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: flt_s:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    flt.s a0, ft1, ft0
+; RV64IF-NEXT:    ret
   %1 = fcmp olt float %a, %b
   %2 = zext i1 %1 to i32
   ret i32 %2
 }
 
 define i32 @fle_s(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fle_s:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    fle.s a0, ft1, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fle_s:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    fle.s a0, ft1, ft0
+; RV64IF-NEXT:    ret
   %1 = fcmp ole float %a, %b
   %2 = zext i1 %1 to i32
   ret i32 %2
 }
 
 declare float @llvm.fma.f32(float, float, float)
 
 define float @fmadd_s(float %a, float %b, float %c) nounwind {
 ; RV32IF-LABEL: fmadd_s:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a2
 ; RV32IF-NEXT:    fmv.w.x ft1, a1
 ; RV32IF-NEXT:    fmv.w.x ft2, a0
 ; RV32IF-NEXT:    fmadd.s ft0, ft2, ft1, ft0
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fmadd_s:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a2
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    fmv.w.x ft2, a0
+; RV64IF-NEXT:    fmadd.s ft0, ft2, ft1, ft0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = call float @llvm.fma.f32(float %a, float %b, float %c)
   ret float %1
 }
 
 define float @fmsub_s(float %a, float %b, float %c) nounwind {
 ; RV32IF-LABEL: fmsub_s:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a2
 ; RV32IF-NEXT:    lui a2, %hi(.LCPI15_0)
 ; RV32IF-NEXT:    addi a2, a2, %lo(.LCPI15_0)
 ; RV32IF-NEXT:    flw ft1, 0(a2)
 ; RV32IF-NEXT:    fadd.s ft0, ft0, ft1
 ; RV32IF-NEXT:    fmv.w.x ft1, a1
 ; RV32IF-NEXT:    fmv.w.x ft2, a0
 ; RV32IF-NEXT:    fmsub.s ft0, ft2, ft1, ft0
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fmsub_s:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a2
+; RV64IF-NEXT:    lui a2, %hi(.LCPI15_0)
+; RV64IF-NEXT:    addi a2, a2, %lo(.LCPI15_0)
+; RV64IF-NEXT:    flw ft1, 0(a2)
+; RV64IF-NEXT:    fadd.s ft0, ft0, ft1
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    fmv.w.x ft2, a0
+; RV64IF-NEXT:    fmsub.s ft0, ft2, ft1, ft0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %c_ = fadd float 0.0, %c ; avoid negation using xor
   %negc = fsub float -0.0, %c_
   %1 = call float @llvm.fma.f32(float %a, float %b, float %negc)
   ret float %1
 }
 
 define float @fnmadd_s(float %a, float %b, float %c) nounwind {
 ; RV32IF-LABEL: fnmadd_s:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a2
 ; RV32IF-NEXT:    lui a2, %hi(.LCPI16_0)
 ; RV32IF-NEXT:    addi a2, a2, %lo(.LCPI16_0)
 ; RV32IF-NEXT:    flw ft1, 0(a2)
 ; RV32IF-NEXT:    fadd.s ft0, ft0, ft1
 ; RV32IF-NEXT:    fmv.w.x ft2, a0
 ; RV32IF-NEXT:    fadd.s ft1, ft2, ft1
 ; RV32IF-NEXT:    fmv.w.x ft2, a1
 ; RV32IF-NEXT:    fnmadd.s ft0, ft1, ft2, ft0
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fnmadd_s:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a2
+; RV64IF-NEXT:    lui a2, %hi(.LCPI16_0)
+; RV64IF-NEXT:    addi a2, a2, %lo(.LCPI16_0)
+; RV64IF-NEXT:    flw ft1, 0(a2)
+; RV64IF-NEXT:    fadd.s ft0, ft0, ft1
+; RV64IF-NEXT:    fmv.w.x ft2, a0
+; RV64IF-NEXT:    fadd.s ft1, ft2, ft1
+; RV64IF-NEXT:    fmv.w.x ft2, a1
+; RV64IF-NEXT:    fnmadd.s ft0, ft1, ft2, ft0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %a_ = fadd float 0.0, %a
   %c_ = fadd float 0.0, %c
   %nega = fsub float -0.0, %a_
   %negc = fsub float -0.0, %c_
   %1 = call float @llvm.fma.f32(float %nega, float %b, float %negc)
   ret float %1
 }
 
 define float @fnmsub_s(float %a, float %b, float %c) nounwind {
 ; RV32IF-LABEL: fnmsub_s:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    lui a0, %hi(.LCPI17_0)
 ; RV32IF-NEXT:    addi a0, a0, %lo(.LCPI17_0)
 ; RV32IF-NEXT:    flw ft1, 0(a0)
 ; RV32IF-NEXT:    fadd.s ft0, ft0, ft1
 ; RV32IF-NEXT:    fmv.w.x ft1, a2
 ; RV32IF-NEXT:    fmv.w.x ft2, a1
 ; RV32IF-NEXT:    fnmsub.s ft0, ft0, ft2, ft1
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fnmsub_s:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    lui a0, %hi(.LCPI17_0)
+; RV64IF-NEXT:    addi a0, a0, %lo(.LCPI17_0)
+; RV64IF-NEXT:    flw ft1, 0(a0)
+; RV64IF-NEXT:    fadd.s ft0, ft0, ft1
+; RV64IF-NEXT:    fmv.w.x ft1, a2
+; RV64IF-NEXT:    fmv.w.x ft2, a1
+; RV64IF-NEXT:    fnmsub.s ft0, ft0, ft2, ft1
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %a_ = fadd float 0.0, %a
   %nega = fsub float -0.0, %a_
   %1 = call float @llvm.fma.f32(float %nega, float %b, float %c)
   ret float %1
 }

llvm/trunk/test/CodeGen/RISCV/float-bitmanip-dagcombines.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s \
 ; RUN:   | FileCheck -check-prefix=RV32I %s
 ; RUN: llc -mtriple=riscv32 -mattr=+f -verify-machineinstrs < %s \
 ; RUN:   | FileCheck -check-prefix=RV32IF %s
 ; RUN: llc -mtriple=riscv64 -verify-machineinstrs < %s \
 ; RUN:   | FileCheck -check-prefix=RV64I %s
+; RUN: llc -mtriple=riscv64 -mattr=+f -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefix=RV64IF %s
 
 ; This file tests cases where simple floating point operations can be
 ; profitably handled though bit manipulation if a soft-float ABI is being used
 ; (e.g. fneg implemented by XORing the sign bit). This is typically handled in
 ; DAGCombiner::visitBITCAST, but this target-independent code may not trigger
 ; in cases where we perform custom legalisation (e.g. RV64F).
 
 define float @fneg(float %a) nounwind {

 ; RV32IF-NEXT:    xor a0, a0, a1
 ; RV32IF-NEXT:    ret
 ;
 ; RV64I-LABEL: fneg:
 ; RV64I:       # %bb.0:
 ; RV64I-NEXT:    lui a1, 524288
 ; RV64I-NEXT:    xor a0, a0, a1
 ; RV64I-NEXT:    ret
+;
+; RV64IF-LABEL: fneg:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    lui a1, 524288
+; RV64IF-NEXT:    xor a0, a0, a1
+; RV64IF-NEXT:    ret
   %1 = fneg float %a
   ret float %1
 }
 
 declare float @llvm.fabs.f32(float)
 
 define float @fabs(float %a) nounwind {
 ; RV32I-LABEL: fabs:

 ; RV32IF-NEXT:    ret
 ;
 ; RV64I-LABEL: fabs:
 ; RV64I:       # %bb.0:
 ; RV64I-NEXT:    lui a1, 524288
 ; RV64I-NEXT:    addiw a1, a1, -1
 ; RV64I-NEXT:    and a0, a0, a1
 ; RV64I-NEXT:    ret
+;
+; RV64IF-LABEL: fabs:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    lui a1, 524288
+; RV64IF-NEXT:    addiw a1, a1, -1
+; RV64IF-NEXT:    and a0, a0, a1
+; RV64IF-NEXT:    ret
   %1 = call float @llvm.fabs.f32(float %a)
   ret float %1
 }
 
 declare float @llvm.copysign.f32(float, float)
 
 ; DAGTypeLegalizer::SoftenFloatRes_FCOPYSIGN will convert to bitwise
 ; operations if floating point isn't supported. A combine could be written to

 ; RV64I:       # %bb.0:
 ; RV64I-NEXT:    not a1, a1
 ; RV64I-NEXT:    lui a2, 524288
 ; RV64I-NEXT:    and a1, a1, a2
 ; RV64I-NEXT:    addiw a2, a2, -1
 ; RV64I-NEXT:    and a0, a0, a2
 ; RV64I-NEXT:    or a0, a0, a1
 ; RV64I-NEXT:    ret
+;
+; RV64IF-LABEL: fcopysign_fneg:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    fsgnjn.s ft0, ft1, ft0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = fneg float %b
   %2 = call float @llvm.copysign.f32(float %a, float %1)
   ret float %2
 }

llvm/trunk/test/CodeGen/RISCV/float-br-fcmp.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -mattr=+f -verify-machineinstrs < %s \
 ; RUN:   | FileCheck -check-prefix=RV32IF %s
+; RUN: llc -mtriple=riscv64 -mattr=+f -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefix=RV64IF %s
 
 declare void @abort()
 declare void @exit(i32)
 declare float @dummy(float)
 
 define void @br_fcmp_false(float %a, float %b) nounwind {
 ; RV32IF-LABEL: br_fcmp_false:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    addi sp, sp, -16
 ; RV32IF-NEXT:    sw ra, 12(sp)
 ; RV32IF-NEXT:    addi a0, zero, 1
 ; RV32IF-NEXT:    bnez a0, .LBB0_2
 ; RV32IF-NEXT:  # %bb.1: # %if.then
 ; RV32IF-NEXT:    lw ra, 12(sp)
 ; RV32IF-NEXT:    addi sp, sp, 16
 ; RV32IF-NEXT:    ret
 ; RV32IF-NEXT:  .LBB0_2: # %if.else
 ; RV32IF-NEXT:    call abort
+;
+; RV64IF-LABEL: br_fcmp_false:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    addi sp, sp, -16
+; RV64IF-NEXT:    sd ra, 8(sp)
+; RV64IF-NEXT:    addi a0, zero, 1
+; RV64IF-NEXT:    bnez a0, .LBB0_2
+; RV64IF-NEXT:  # %bb.1: # %if.then
+; RV64IF-NEXT:    ld ra, 8(sp)
+; RV64IF-NEXT:    addi sp, sp, 16
+; RV64IF-NEXT:    ret
+; RV64IF-NEXT:  .LBB0_2: # %if.else
+; RV64IF-NEXT:    call abort
   %1 = fcmp false float %a, %b
   br i1 %1, label %if.then, label %if.else
 if.then:
   ret void
 if.else:
   tail call void @abort()
   unreachable
 }
 
 define void @br_fcmp_oeq(float %a, float %b) nounwind {
 ; RV32IF-LABEL: br_fcmp_oeq:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    addi sp, sp, -16
 ; RV32IF-NEXT:    sw ra, 12(sp)
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    feq.s a0, ft1, ft0
 ; RV32IF-NEXT:    bnez a0, .LBB1_2
 ; RV32IF-NEXT:  # %bb.1: # %if.else
 ; RV32IF-NEXT:    lw ra, 12(sp)
 ; RV32IF-NEXT:    addi sp, sp, 16
 ; RV32IF-NEXT:    ret
 ; RV32IF-NEXT:  .LBB1_2: # %if.then
 ; RV32IF-NEXT:    call abort
+;
+; RV64IF-LABEL: br_fcmp_oeq:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    addi sp, sp, -16
+; RV64IF-NEXT:    sd ra, 8(sp)
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    feq.s a0, ft1, ft0
+; RV64IF-NEXT:    bnez a0, .LBB1_2
+; RV64IF-NEXT:  # %bb.1: # %if.else
+; RV64IF-NEXT:    ld ra, 8(sp)
+; RV64IF-NEXT:    addi sp, sp, 16
+; RV64IF-NEXT:    ret
+; RV64IF-NEXT:  .LBB1_2: # %if.then
+; RV64IF-NEXT:    call abort
   %1 = fcmp oeq float %a, %b
   br i1 %1, label %if.then, label %if.else
 if.else:
   ret void
 if.then:
   tail call void @abort()
   unreachable
 }

 ; RV32IF-NEXT:    xori a0, a0, 1
 ; RV32IF-NEXT:    beqz a0, .LBB2_2
 ; RV32IF-NEXT:  # %bb.1: # %if.else
 ; RV32IF-NEXT:    lw ra, 12(sp)
 ; RV32IF-NEXT:    addi sp, sp, 16
 ; RV32IF-NEXT:    ret
 ; RV32IF-NEXT:  .LBB2_2: # %if.then
 ; RV32IF-NEXT:    call abort
+;
+; RV64IF-LABEL: br_fcmp_oeq_alt:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    addi sp, sp, -16
+; RV64IF-NEXT:    sd ra, 8(sp)
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    feq.s a0, ft1, ft0
+; RV64IF-NEXT:    xori a0, a0, 1
+; RV64IF-NEXT:    beqz a0, .LBB2_2
+; RV64IF-NEXT:  # %bb.1: # %if.else
+; RV64IF-NEXT:    ld ra, 8(sp)
+; RV64IF-NEXT:    addi sp, sp, 16
+; RV64IF-NEXT:    ret
+; RV64IF-NEXT:  .LBB2_2: # %if.then
+; RV64IF-NEXT:    call abort
   %1 = fcmp oeq float %a, %b
   br i1 %1, label %if.then, label %if.else
 if.then:
   tail call void @abort()
   unreachable
 if.else:
   ret void
 }
 
 define void @br_fcmp_ogt(float %a, float %b) nounwind {
 ; RV32IF-LABEL: br_fcmp_ogt:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    addi sp, sp, -16
 ; RV32IF-NEXT:    sw ra, 12(sp)
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    fmv.w.x ft1, a1
 ; RV32IF-NEXT:    flt.s a0, ft1, ft0
 ; RV32IF-NEXT:    bnez a0, .LBB3_2
 ; RV32IF-NEXT:  # %bb.1: # %if.else
 ; RV32IF-NEXT:    lw ra, 12(sp)
 ; RV32IF-NEXT:    addi sp, sp, 16
 ; RV32IF-NEXT:    ret
 ; RV32IF-NEXT:  .LBB3_2: # %if.then
 ; RV32IF-NEXT:    call abort
+;
+; RV64IF-LABEL: br_fcmp_ogt:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    addi sp, sp, -16
+; RV64IF-NEXT:    sd ra, 8(sp)
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    flt.s a0, ft1, ft0
+; RV64IF-NEXT:    bnez a0, .LBB3_2
+; RV64IF-NEXT:  # %bb.1: # %if.else
+; RV64IF-NEXT:    ld ra, 8(sp)
+; RV64IF-NEXT:    addi sp, sp, 16
+; RV64IF-NEXT:    ret
+; RV64IF-NEXT:  .LBB3_2: # %if.then
+; RV64IF-NEXT:    call abort
   %1 = fcmp ogt float %a, %b
   br i1 %1, label %if.then, label %if.else
 if.else:
   ret void
 if.then:
   tail call void @abort()
   unreachable
 }
 
 define void @br_fcmp_oge(float %a, float %b) nounwind {
 ; RV32IF-LABEL: br_fcmp_oge:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    addi sp, sp, -16
 ; RV32IF-NEXT:    sw ra, 12(sp)
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    fmv.w.x ft1, a1
 ; RV32IF-NEXT:    fle.s a0, ft1, ft0
 ; RV32IF-NEXT:    bnez a0, .LBB4_2
 ; RV32IF-NEXT:  # %bb.1: # %if.else
 ; RV32IF-NEXT:    lw ra, 12(sp)
 ; RV32IF-NEXT:    addi sp, sp, 16
 ; RV32IF-NEXT:    ret
 ; RV32IF-NEXT:  .LBB4_2: # %if.then
 ; RV32IF-NEXT:    call abort
+;
+; RV64IF-LABEL: br_fcmp_oge:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    addi sp, sp, -16
+; RV64IF-NEXT:    sd ra, 8(sp)
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    fle.s a0, ft1, ft0
+; RV64IF-NEXT:    bnez a0, .LBB4_2
+; RV64IF-NEXT:  # %bb.1: # %if.else
+; RV64IF-NEXT:    ld ra, 8(sp)
+; RV64IF-NEXT:    addi sp, sp, 16
+; RV64IF-NEXT:    ret
+; RV64IF-NEXT:  .LBB4_2: # %if.then
+; RV64IF-NEXT:    call abort
   %1 = fcmp oge float %a, %b
   br i1 %1, label %if.then, label %if.else
 if.else:
   ret void
 if.then:
   tail call void @abort()
   unreachable
 }
 
 define void @br_fcmp_olt(float %a, float %b) nounwind {
 ; RV32IF-LABEL: br_fcmp_olt:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    addi sp, sp, -16
 ; RV32IF-NEXT:    sw ra, 12(sp)
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    flt.s a0, ft1, ft0
 ; RV32IF-NEXT:    bnez a0, .LBB5_2
 ; RV32IF-NEXT:  # %bb.1: # %if.else
 ; RV32IF-NEXT:    lw ra, 12(sp)
 ; RV32IF-NEXT:    addi sp, sp, 16
 ; RV32IF-NEXT:    ret
 ; RV32IF-NEXT:  .LBB5_2: # %if.then
 ; RV32IF-NEXT:    call abort
+;
+; RV64IF-LABEL: br_fcmp_olt:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    addi sp, sp, -16
+; RV64IF-NEXT:    sd ra, 8(sp)
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    flt.s a0, ft1, ft0
+; RV64IF-NEXT:    bnez a0, .LBB5_2
+; RV64IF-NEXT:  # %bb.1: # %if.else
+; RV64IF-NEXT:    ld ra, 8(sp)
+; RV64IF-NEXT:    addi sp, sp, 16
+; RV64IF-NEXT:    ret
+; RV64IF-NEXT:  .LBB5_2: # %if.then
+; RV64IF-NEXT:    call abort
   %1 = fcmp olt float %a, %b
   br i1 %1, label %if.then, label %if.else
 if.else:
   ret void
 if.then:
   tail call void @abort()
   unreachable
 }
 
 define void @br_fcmp_ole(float %a, float %b) nounwind {
 ; RV32IF-LABEL: br_fcmp_ole:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    addi sp, sp, -16
 ; RV32IF-NEXT:    sw ra, 12(sp)
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    fle.s a0, ft1, ft0
 ; RV32IF-NEXT:    bnez a0, .LBB6_2
 ; RV32IF-NEXT:  # %bb.1: # %if.else
 ; RV32IF-NEXT:    lw ra, 12(sp)
 ; RV32IF-NEXT:    addi sp, sp, 16
 ; RV32IF-NEXT:    ret
 ; RV32IF-NEXT:  .LBB6_2: # %if.then
 ; RV32IF-NEXT:    call abort
+;
+; RV64IF-LABEL: br_fcmp_ole:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    addi sp, sp, -16
+; RV64IF-NEXT:    sd ra, 8(sp)
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    fle.s a0, ft1, ft0
+; RV64IF-NEXT:    bnez a0, .LBB6_2
+; RV64IF-NEXT:  # %bb.1: # %if.else
+; RV64IF-NEXT:    ld ra, 8(sp)
+; RV64IF-NEXT:    addi sp, sp, 16
+; RV64IF-NEXT:    ret
+; RV64IF-NEXT:  .LBB6_2: # %if.then
+; RV64IF-NEXT:    call abort
   %1 = fcmp ole float %a, %b
   br i1 %1, label %if.then, label %if.else
 if.else:
   ret void
 if.then:
   tail call void @abort()
   unreachable
 }

 ; RV32IF-NEXT:    and a0, a1, a0
 ; RV32IF-NEXT:    bnez a0, .LBB7_2
 ; RV32IF-NEXT:  # %bb.1: # %if.else
 ; RV32IF-NEXT:    lw ra, 12(sp)
 ; RV32IF-NEXT:    addi sp, sp, 16
 ; RV32IF-NEXT:    ret
 ; RV32IF-NEXT:  .LBB7_2: # %if.then
 ; RV32IF-NEXT:    call abort
+;
+; RV64IF-LABEL: br_fcmp_one:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    addi sp, sp, -16
+; RV64IF-NEXT:    sd ra, 8(sp)
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    feq.s a0, ft1, ft1
+; RV64IF-NEXT:    feq.s a1, ft0, ft0
+; RV64IF-NEXT:    and a0, a1, a0
+; RV64IF-NEXT:    feq.s a1, ft0, ft1
+; RV64IF-NEXT:    not a1, a1
+; RV64IF-NEXT:    seqz a0, a0
+; RV64IF-NEXT:    xori a0, a0, 1
+; RV64IF-NEXT:    and a0, a1, a0
+; RV64IF-NEXT:    bnez a0, .LBB7_2
+; RV64IF-NEXT:  # %bb.1: # %if.else
+; RV64IF-NEXT:    ld ra, 8(sp)
+; RV64IF-NEXT:    addi sp, sp, 16
+; RV64IF-NEXT:    ret
+; RV64IF-NEXT:  .LBB7_2: # %if.then
+; RV64IF-NEXT:    call abort
   %1 = fcmp one float %a, %b
   br i1 %1, label %if.then, label %if.else
 if.else:
   ret void
 if.then:
   tail call void @abort()
   unreachable
 }

 ; RV32IF-NEXT:    xori a0, a0, 1
 ; RV32IF-NEXT:    bnez a0, .LBB8_2
 ; RV32IF-NEXT:  # %bb.1: # %if.else
 ; RV32IF-NEXT:    lw ra, 12(sp)
 ; RV32IF-NEXT:    addi sp, sp, 16
 ; RV32IF-NEXT:    ret
 ; RV32IF-NEXT:  .LBB8_2: # %if.then
 ; RV32IF-NEXT:    call abort
+;
+; RV64IF-LABEL: br_fcmp_ord:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    addi sp, sp, -16
+; RV64IF-NEXT:    sd ra, 8(sp)
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    feq.s a1, ft0, ft0
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    feq.s a0, ft0, ft0
+; RV64IF-NEXT:    and a0, a0, a1
+; RV64IF-NEXT:    seqz a0, a0
+; RV64IF-NEXT:    xori a0, a0, 1
+; RV64IF-NEXT:    bnez a0, .LBB8_2
+; RV64IF-NEXT:  # %bb.1: # %if.else
+; RV64IF-NEXT:    ld ra, 8(sp)
+; RV64IF-NEXT:    addi sp, sp, 16
+; RV64IF-NEXT:    ret
+; RV64IF-NEXT:  .LBB8_2: # %if.then
+; RV64IF-NEXT:    call abort
   %1 = fcmp ord float %a, %b
   br i1 %1, label %if.then, label %if.else
 if.else:
   ret void
 if.then:
   tail call void @abort()
   unreachable
 }

 ; RV32IF-NEXT:    or a0, a0, a1
 ; RV32IF-NEXT:    bnez a0, .LBB9_2
 ; RV32IF-NEXT:  # %bb.1: # %if.else
 ; RV32IF-NEXT:    lw ra, 12(sp)
 ; RV32IF-NEXT:    addi sp, sp, 16
 ; RV32IF-NEXT:    ret
 ; RV32IF-NEXT:  .LBB9_2: # %if.then
 ; RV32IF-NEXT:    call abort
+;
+; RV64IF-LABEL: br_fcmp_ueq:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    addi sp, sp, -16
+; RV64IF-NEXT:    sd ra, 8(sp)
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    feq.s a0, ft1, ft0
+; RV64IF-NEXT:    feq.s a1, ft0, ft0
+; RV64IF-NEXT:    feq.s a2, ft1, ft1
+; RV64IF-NEXT:    and a1, a2, a1
+; RV64IF-NEXT:    seqz a1, a1
+; RV64IF-NEXT:    or a0, a0, a1
+; RV64IF-NEXT:    bnez a0, .LBB9_2
+; RV64IF-NEXT:  # %bb.1: # %if.else
+; RV64IF-NEXT:    ld ra, 8(sp)
+; RV64IF-NEXT:    addi sp, sp, 16
+; RV64IF-NEXT:    ret
+; RV64IF-NEXT:  .LBB9_2: # %if.then
+; RV64IF-NEXT:    call abort
   %1 = fcmp ueq float %a, %b
   br i1 %1, label %if.then, label %if.else
 if.else:
   ret void
 if.then:
   tail call void @abort()
   unreachable
 }

 ; RV32IF-NEXT:    xori a0, a0, 1
 ; RV32IF-NEXT:    bnez a0, .LBB10_2
 ; RV32IF-NEXT:  # %bb.1: # %if.else
 ; RV32IF-NEXT:    lw ra, 12(sp)
 ; RV32IF-NEXT:    addi sp, sp, 16
 ; RV32IF-NEXT:    ret
 ; RV32IF-NEXT:  .LBB10_2: # %if.then
 ; RV32IF-NEXT:    call abort
+;
+; RV64IF-LABEL: br_fcmp_ugt:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    addi sp, sp, -16
+; RV64IF-NEXT:    sd ra, 8(sp)
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    fle.s a0, ft1, ft0
+; RV64IF-NEXT:    xori a0, a0, 1
+; RV64IF-NEXT:    bnez a0, .LBB10_2
+; RV64IF-NEXT:  # %bb.1: # %if.else
+; RV64IF-NEXT:    ld ra, 8(sp)
+; RV64IF-NEXT:    addi sp, sp, 16
+; RV64IF-NEXT:    ret
+; RV64IF-NEXT:  .LBB10_2: # %if.then
+; RV64IF-NEXT:    call abort
   %1 = fcmp ugt float %a, %b
   br i1 %1, label %if.then, label %if.else
 if.else:
   ret void
 if.then:
   tail call void @abort()
   unreachable
 }

 ; RV32IF-NEXT:    xori a0, a0, 1
 ; RV32IF-NEXT:    bnez a0, .LBB11_2
 ; RV32IF-NEXT:  # %bb.1: # %if.else
 ; RV32IF-NEXT:    lw ra, 12(sp)
 ; RV32IF-NEXT:    addi sp, sp, 16
 ; RV32IF-NEXT:    ret
 ; RV32IF-NEXT:  .LBB11_2: # %if.then
 ; RV32IF-NEXT:    call abort
+;
+; RV64IF-LABEL: br_fcmp_uge:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    addi sp, sp, -16
+; RV64IF-NEXT:    sd ra, 8(sp)
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    flt.s a0, ft1, ft0
+; RV64IF-NEXT:    xori a0, a0, 1
+; RV64IF-NEXT:    bnez a0, .LBB11_2
+; RV64IF-NEXT:  # %bb.1: # %if.else
+; RV64IF-NEXT:    ld ra, 8(sp)
+; RV64IF-NEXT:    addi sp, sp, 16
+; RV64IF-NEXT:    ret
+; RV64IF-NEXT:  .LBB11_2: # %if.then
+; RV64IF-NEXT:    call abort
   %1 = fcmp uge float %a, %b
   br i1 %1, label %if.then, label %if.else
 if.else:
   ret void
 if.then:
   tail call void @abort()
   unreachable
 }

 ; RV32IF-NEXT:    xori a0, a0, 1
 ; RV32IF-NEXT:    bnez a0, .LBB12_2
 ; RV32IF-NEXT:  # %bb.1: # %if.else
 ; RV32IF-NEXT:    lw ra, 12(sp)
 ; RV32IF-NEXT:    addi sp, sp, 16
 ; RV32IF-NEXT:    ret
 ; RV32IF-NEXT:  .LBB12_2: # %if.then
 ; RV32IF-NEXT:    call abort
+;
+; RV64IF-LABEL: br_fcmp_ult:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    addi sp, sp, -16
+; RV64IF-NEXT:    sd ra, 8(sp)
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    fle.s a0, ft1, ft0
+; RV64IF-NEXT:    xori a0, a0, 1
+; RV64IF-NEXT:    bnez a0, .LBB12_2
+; RV64IF-NEXT:  # %bb.1: # %if.else
+; RV64IF-NEXT:    ld ra, 8(sp)
+; RV64IF-NEXT:    addi sp, sp, 16
+; RV64IF-NEXT:    ret
+; RV64IF-NEXT:  .LBB12_2: # %if.then
+; RV64IF-NEXT:    call abort
   %1 = fcmp ult float %a, %b
   br i1 %1, label %if.then, label %if.else
 if.else:
   ret void
 if.then:
   tail call void @abort()
   unreachable
 }

 ; RV32IF-NEXT:    xori a0, a0, 1
 ; RV32IF-NEXT:    bnez a0, .LBB13_2
 ; RV32IF-NEXT:  # %bb.1: # %if.else
 ; RV32IF-NEXT:    lw ra, 12(sp)
 ; RV32IF-NEXT:    addi sp, sp, 16
 ; RV32IF-NEXT:    ret
 ; RV32IF-NEXT:  .LBB13_2: # %if.then
 ; RV32IF-NEXT:    call abort
+;
+; RV64IF-LABEL: br_fcmp_ule:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    addi sp, sp, -16
+; RV64IF-NEXT:    sd ra, 8(sp)
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    flt.s a0, ft1, ft0
+; RV64IF-NEXT:    xori a0, a0, 1
+; RV64IF-NEXT:    bnez a0, .LBB13_2
+; RV64IF-NEXT:  # %bb.1: # %if.else
+; RV64IF-NEXT:    ld ra, 8(sp)
+; RV64IF-NEXT:    addi sp, sp, 16
+; RV64IF-NEXT:    ret
+; RV64IF-NEXT:  .LBB13_2: # %if.then
+; RV64IF-NEXT:    call abort
   %1 = fcmp ule float %a, %b
   br i1 %1, label %if.then, label %if.else
 if.else:
   ret void
 if.then:
   tail call void @abort()
   unreachable
 }

 ; RV32IF-NEXT:    xori a0, a0, 1
 ; RV32IF-NEXT:    bnez a0, .LBB14_2
 ; RV32IF-NEXT:  # %bb.1: # %if.else
 ; RV32IF-NEXT:    lw ra, 12(sp)
 ; RV32IF-NEXT:    addi sp, sp, 16
 ; RV32IF-NEXT:    ret
 ; RV32IF-NEXT:  .LBB14_2: # %if.then
 ; RV32IF-NEXT:    call abort
+;
+; RV64IF-LABEL: br_fcmp_une:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    addi sp, sp, -16
+; RV64IF-NEXT:    sd ra, 8(sp)
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    feq.s a0, ft1, ft0
+; RV64IF-NEXT:    xori a0, a0, 1
+; RV64IF-NEXT:    bnez a0, .LBB14_2
+; RV64IF-NEXT:  # %bb.1: # %if.else
+; RV64IF-NEXT:    ld ra, 8(sp)
+; RV64IF-NEXT:    addi sp, sp, 16
+; RV64IF-NEXT:    ret
+; RV64IF-NEXT:  .LBB14_2: # %if.then
+; RV64IF-NEXT:    call abort
   %1 = fcmp une float %a, %b
   br i1 %1, label %if.then, label %if.else
 if.else:
   ret void
 if.then:
   tail call void @abort()
   unreachable
 }

 ; RV32IF-NEXT:    seqz a0, a0
 ; RV32IF-NEXT:    bnez a0, .LBB15_2
 ; RV32IF-NEXT:  # %bb.1: # %if.else
 ; RV32IF-NEXT:    lw ra, 12(sp)
 ; RV32IF-NEXT:    addi sp, sp, 16
 ; RV32IF-NEXT:    ret
 ; RV32IF-NEXT:  .LBB15_2: # %if.then
 ; RV32IF-NEXT:    call abort
+;
+; RV64IF-LABEL: br_fcmp_uno:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    addi sp, sp, -16
+; RV64IF-NEXT:    sd ra, 8(sp)
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    feq.s a1, ft0, ft0
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    feq.s a0, ft0, ft0
+; RV64IF-NEXT:    and a0, a0, a1
+; RV64IF-NEXT:    seqz a0, a0
+; RV64IF-NEXT:    bnez a0, .LBB15_2
+; RV64IF-NEXT:  # %bb.1: # %if.else
+; RV64IF-NEXT:    ld ra, 8(sp)
+; RV64IF-NEXT:    addi sp, sp, 16
+; RV64IF-NEXT:    ret
+; RV64IF-NEXT:  .LBB15_2: # %if.then
+; RV64IF-NEXT:    call abort
   %1 = fcmp uno float %a, %b
   br i1 %1, label %if.then, label %if.else
 if.else:
   ret void
 if.then:
   tail call void @abort()
   unreachable
 }
 
 define void @br_fcmp_true(float %a, float %b) nounwind {
 ; RV32IF-LABEL: br_fcmp_true:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    addi sp, sp, -16
 ; RV32IF-NEXT:    sw ra, 12(sp)
 ; RV32IF-NEXT:    addi a0, zero, 1
 ; RV32IF-NEXT:    bnez a0, .LBB16_2
 ; RV32IF-NEXT:  # %bb.1: # %if.else
 ; RV32IF-NEXT:    lw ra, 12(sp)
 ; RV32IF-NEXT:    addi sp, sp, 16
 ; RV32IF-NEXT:    ret
 ; RV32IF-NEXT:  .LBB16_2: # %if.then
 ; RV32IF-NEXT:    call abort
+;
+; RV64IF-LABEL: br_fcmp_true:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    addi sp, sp, -16
+; RV64IF-NEXT:    sd ra, 8(sp)
+; RV64IF-NEXT:    addi a0, zero, 1
+; RV64IF-NEXT:    bnez a0, .LBB16_2
+; RV64IF-NEXT:  # %bb.1: # %if.else
+; RV64IF-NEXT:    ld ra, 8(sp)
+; RV64IF-NEXT:    addi sp, sp, 16
+; RV64IF-NEXT:    ret
+; RV64IF-NEXT:  .LBB16_2: # %if.then
+; RV64IF-NEXT:    call abort
   %1 = fcmp true float %a, %b
   br i1 %1, label %if.then, label %if.else
 if.else:
   ret void
 if.then:
   tail call void @abort()
   unreachable
 }

 ; RV32IF-NEXT:    beqz a0, .LBB17_3
 ; RV32IF-NEXT:  # %bb.2: # %if.end4
 ; RV32IF-NEXT:    mv a0, zero
 ; RV32IF-NEXT:    lw ra, 12(sp)
 ; RV32IF-NEXT:    addi sp, sp, 16
 ; RV32IF-NEXT:    ret
 ; RV32IF-NEXT:  .LBB17_3: # %if.then
 ; RV32IF-NEXT:    call abort
+;
+; RV64IF-LABEL: br_fcmp_store_load_stack_slot:
+; RV64IF:       # %bb.0: # %entry
+; RV64IF-NEXT:    addi sp, sp, -32
+; RV64IF-NEXT:    sd ra, 24(sp)
+; RV64IF-NEXT:    sd s1, 16(sp)
+; RV64IF-NEXT:    lui a0, %hi(.LCPI17_0)
+; RV64IF-NEXT:    addi a0, a0, %lo(.LCPI17_0)
+; RV64IF-NEXT:    flw ft0, 0(a0)
+; RV64IF-NEXT:    fsw ft0, 12(sp)
+; RV64IF-NEXT:    fmv.x.w s1, ft0
+; RV64IF-NEXT:    mv a0, s1
+; RV64IF-NEXT:    call dummy
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    flw ft1, 12(sp)
+; RV64IF-NEXT:    feq.s a0, ft0, ft1
+; RV64IF-NEXT:    beqz a0, .LBB17_3
+; RV64IF-NEXT:  # %bb.1: # %if.end
+; RV64IF-NEXT:    mv a0, s1
+; RV64IF-NEXT:    call dummy
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    flw ft1, 12(sp)
+; RV64IF-NEXT:    feq.s a0, ft0, ft1
+; RV64IF-NEXT:    beqz a0, .LBB17_3
+; RV64IF-NEXT:  # %bb.2: # %if.end4
+; RV64IF-NEXT:    mv a0, zero
+; RV64IF-NEXT:    ld s1, 16(sp)
+; RV64IF-NEXT:    ld ra, 24(sp)
+; RV64IF-NEXT:    addi sp, sp, 32
+; RV64IF-NEXT:    ret
+; RV64IF-NEXT:  .LBB17_3: # %if.then
+; RV64IF-NEXT:    call abort
 entry:
   %call = call float @dummy(float 0.000000e+00)
   %cmp = fcmp une float %call, 0.000000e+00
   br i1 %cmp, label %if.then, label %if.end
 
 if.then:
   call void @abort()
   unreachable

llvm/trunk/test/CodeGen/RISCV/float-convert.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -mattr=+f -verify-machineinstrs < %s \
 ; RUN:   | FileCheck -check-prefix=RV32IF %s
+; RUN: llc -mtriple=riscv64 -mattr=+f -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefix=RV64IF %s
 
+; For RV64F, fcvt.l.s is semantically equivalent to fcvt.w.s in this case
+; because fptosi will produce poison if the result doesn't fit into an i32.
 define i32 @fcvt_w_s(float %a) nounwind {
 ; RV32IF-LABEL: fcvt_w_s:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    fcvt.w.s a0, ft0, rtz
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcvt_w_s:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fcvt.l.s a0, ft0, rtz
+; RV64IF-NEXT:    ret
   %1 = fptosi float %a to i32
   ret i32 %1
 }
 
+; For RV64F, fcvt.lu.s is semantically equivalent to fcvt.wu.s in this case
+; because fptoui will produce poison if the result doesn't fit into an i32.
 define i32 @fcvt_wu_s(float %a) nounwind {
 ; RV32IF-LABEL: fcvt_wu_s:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    fcvt.wu.s a0, ft0, rtz
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcvt_wu_s:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fcvt.lu.s a0, ft0, rtz
+; RV64IF-NEXT:    ret
   %1 = fptoui float %a to i32
   ret i32 %1
 }
 
 define i32 @fmv_x_w(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fmv_x_w:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    fadd.s ft0, ft1, ft0
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fmv_x_w:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    fadd.s ft0, ft1, ft0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
 ; Ensure fmv.x.w is generated even for a soft float calling convention
   %1 = fadd float %a, %b
   %2 = bitcast float %1 to i32
   ret i32 %2
 }
 
 define float @fcvt_s_w(i32 %a) nounwind {
 ; RV32IF-LABEL: fcvt_s_w:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fcvt.s.w ft0, a0
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcvt_s_w:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fcvt.s.w ft0, a0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = sitofp i32 %a to float
   ret float %1
 }
 
 define float @fcvt_s_wu(i32 %a) nounwind {
 ; RV32IF-LABEL: fcvt_s_wu:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fcvt.s.wu ft0, a0
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcvt_s_wu:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fcvt.s.wu ft0, a0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = uitofp i32 %a to float
   ret float %1
 }
 
 define float @fmv_w_x(i32 %a, i32 %b) nounwind {
 ; RV32IF-LABEL: fmv_w_x:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    fadd.s ft0, ft1, ft0
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fmv_w_x:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    fadd.s ft0, ft1, ft0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
 ; Ensure fmv.w.x is generated even for a soft float calling convention
   %1 = bitcast i32 %a to float
   %2 = bitcast i32 %b to float
   %3 = fadd float %1, %2
   ret float %3
 }
+
+define i64 @fcvt_l_s(float %a) nounwind {
+; RV32IF-LABEL: fcvt_l_s:
+; RV32IF:       # %bb.0:
+; RV32IF-NEXT:    addi sp, sp, -16
+; RV32IF-NEXT:    sw ra, 12(sp)
+; RV32IF-NEXT:    call __fixsfdi
+; RV32IF-NEXT:    lw ra, 12(sp)
+; RV32IF-NEXT:    addi sp, sp, 16
+; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcvt_l_s:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fcvt.l.s a0, ft0, rtz
+; RV64IF-NEXT:    ret
+  %1 = fptosi float %a to i64
+  ret i64 %1
+}
+
+define i64 @fcvt_lu_s(float %a) nounwind {
+; RV32IF-LABEL: fcvt_lu_s:
+; RV32IF:       # %bb.0:
+; RV32IF-NEXT:    addi sp, sp, -16
+; RV32IF-NEXT:    sw ra, 12(sp)
+; RV32IF-NEXT:    call __fixunssfdi
+; RV32IF-NEXT:    lw ra, 12(sp)
+; RV32IF-NEXT:    addi sp, sp, 16
+; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcvt_lu_s:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fcvt.lu.s a0, ft0, rtz
+; RV64IF-NEXT:    ret
+  %1 = fptoui float %a to i64
+  ret i64 %1
+}
+
+define float @fcvt_s_l(i64 %a) nounwind {
+; RV32IF-LABEL: fcvt_s_l:
+; RV32IF:       # %bb.0:
+; RV32IF-NEXT:    addi sp, sp, -16
+; RV32IF-NEXT:    sw ra, 12(sp)
+; RV32IF-NEXT:    call __floatdisf
+; RV32IF-NEXT:    lw ra, 12(sp)
+; RV32IF-NEXT:    addi sp, sp, 16
+; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcvt_s_l:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fcvt.s.l ft0, a0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
+  %1 = sitofp i64 %a to float
+  ret float %1
+}
+
+define float @fcvt_s_lu(i64 %a) nounwind {
+; RV32IF-LABEL: fcvt_s_lu:
+; RV32IF:       # %bb.0:
+; RV32IF-NEXT:    addi sp, sp, -16
+; RV32IF-NEXT:    sw ra, 12(sp)
+; RV32IF-NEXT:    call __floatundisf
+; RV32IF-NEXT:    lw ra, 12(sp)
+; RV32IF-NEXT:    addi sp, sp, 16
+; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcvt_s_lu:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fcvt.s.lu ft0, a0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
+  %1 = uitofp i64 %a to float
+  ret float %1
+}

llvm/trunk/test/CodeGen/RISCV/float-fcmp.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -mattr=+f -verify-machineinstrs < %s \
 ; RUN:   | FileCheck -check-prefix=RV32IF %s
+; RUN: llc -mtriple=riscv64 -mattr=+f -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefix=RV64IF %s
 
 define i32 @fcmp_false(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fcmp_false:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    mv a0, zero
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcmp_false:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    mv a0, zero
+; RV64IF-NEXT:    ret
   %1 = fcmp false float %a, %b
   %2 = zext i1 %1 to i32
   ret i32 %2
 }
 
 define i32 @fcmp_oeq(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fcmp_oeq:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    feq.s a0, ft1, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcmp_oeq:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    feq.s a0, ft1, ft0
+; RV64IF-NEXT:    ret
   %1 = fcmp oeq float %a, %b
   %2 = zext i1 %1 to i32
   ret i32 %2
 }
 
 define i32 @fcmp_ogt(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fcmp_ogt:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    fmv.w.x ft1, a1
 ; RV32IF-NEXT:    flt.s a0, ft1, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcmp_ogt:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    flt.s a0, ft1, ft0
+; RV64IF-NEXT:    ret
   %1 = fcmp ogt float %a, %b
   %2 = zext i1 %1 to i32
   ret i32 %2
 }
 
 define i32 @fcmp_oge(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fcmp_oge:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    fmv.w.x ft1, a1
 ; RV32IF-NEXT:    fle.s a0, ft1, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcmp_oge:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    fle.s a0, ft1, ft0
+; RV64IF-NEXT:    ret
   %1 = fcmp oge float %a, %b
   %2 = zext i1 %1 to i32
   ret i32 %2
 }
 
 define i32 @fcmp_olt(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fcmp_olt:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    flt.s a0, ft1, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcmp_olt:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    flt.s a0, ft1, ft0
+; RV64IF-NEXT:    ret
   %1 = fcmp olt float %a, %b
   %2 = zext i1 %1 to i32
   ret i32 %2
 }
 
 define i32 @fcmp_ole(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fcmp_ole:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    fle.s a0, ft1, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcmp_ole:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    fle.s a0, ft1, ft0
+; RV64IF-NEXT:    ret
   %1 = fcmp ole float %a, %b
   %2 = zext i1 %1 to i32
   ret i32 %2
 }
 
 define i32 @fcmp_one(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fcmp_one:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    fmv.w.x ft1, a1
 ; RV32IF-NEXT:    feq.s a0, ft1, ft1
 ; RV32IF-NEXT:    feq.s a1, ft0, ft0
 ; RV32IF-NEXT:    and a0, a1, a0
 ; RV32IF-NEXT:    feq.s a1, ft0, ft1
 ; RV32IF-NEXT:    not a1, a1
 ; RV32IF-NEXT:    seqz a0, a0
 ; RV32IF-NEXT:    xori a0, a0, 1
 ; RV32IF-NEXT:    and a0, a1, a0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcmp_one:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    feq.s a0, ft1, ft1
+; RV64IF-NEXT:    feq.s a1, ft0, ft0
+; RV64IF-NEXT:    and a0, a1, a0
+; RV64IF-NEXT:    feq.s a1, ft0, ft1
+; RV64IF-NEXT:    not a1, a1
+; RV64IF-NEXT:    seqz a0, a0
+; RV64IF-NEXT:    xori a0, a0, 1
+; RV64IF-NEXT:    and a0, a1, a0
+; RV64IF-NEXT:    ret
   %1 = fcmp one float %a, %b
   %2 = zext i1 %1 to i32
   ret i32 %2
 }
 
 define i32 @fcmp_ord(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fcmp_ord:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    feq.s a1, ft0, ft0
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    feq.s a0, ft0, ft0
 ; RV32IF-NEXT:    and a0, a0, a1
 ; RV32IF-NEXT:    seqz a0, a0
 ; RV32IF-NEXT:    xori a0, a0, 1
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcmp_ord:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    feq.s a1, ft0, ft0
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    feq.s a0, ft0, ft0
+; RV64IF-NEXT:    and a0, a0, a1
+; RV64IF-NEXT:    seqz a0, a0
+; RV64IF-NEXT:    xori a0, a0, 1
+; RV64IF-NEXT:    ret
   %1 = fcmp ord float %a, %b
   %2 = zext i1 %1 to i32
   ret i32 %2
 }
 
 define i32 @fcmp_ueq(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fcmp_ueq:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    feq.s a0, ft1, ft0
 ; RV32IF-NEXT:    feq.s a1, ft0, ft0
 ; RV32IF-NEXT:    feq.s a2, ft1, ft1
 ; RV32IF-NEXT:    and a1, a2, a1
 ; RV32IF-NEXT:    seqz a1, a1
 ; RV32IF-NEXT:    or a0, a0, a1
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcmp_ueq:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    feq.s a0, ft1, ft0
+; RV64IF-NEXT:    feq.s a1, ft0, ft0
+; RV64IF-NEXT:    feq.s a2, ft1, ft1
+; RV64IF-NEXT:    and a1, a2, a1
+; RV64IF-NEXT:    seqz a1, a1
+; RV64IF-NEXT:    or a0, a0, a1
+; RV64IF-NEXT:    ret
   %1 = fcmp ueq float %a, %b
   %2 = zext i1 %1 to i32
   ret i32 %2
 }
 
 define i32 @fcmp_ugt(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fcmp_ugt:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    fle.s a0, ft1, ft0
 ; RV32IF-NEXT:    xori a0, a0, 1
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcmp_ugt:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    fle.s a0, ft1, ft0
+; RV64IF-NEXT:    xori a0, a0, 1
+; RV64IF-NEXT:    ret
   %1 = fcmp ugt float %a, %b
   %2 = zext i1 %1 to i32
   ret i32 %2
 }
 
 define i32 @fcmp_uge(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fcmp_uge:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    flt.s a0, ft1, ft0
 ; RV32IF-NEXT:    xori a0, a0, 1
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcmp_uge:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    flt.s a0, ft1, ft0
+; RV64IF-NEXT:    xori a0, a0, 1
+; RV64IF-NEXT:    ret
   %1 = fcmp uge float %a, %b
   %2 = zext i1 %1 to i32
   ret i32 %2
 }
 
 define i32 @fcmp_ult(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fcmp_ult:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    fmv.w.x ft1, a1
 ; RV32IF-NEXT:    fle.s a0, ft1, ft0
 ; RV32IF-NEXT:    xori a0, a0, 1
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcmp_ult:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    fle.s a0, ft1, ft0
+; RV64IF-NEXT:    xori a0, a0, 1
+; RV64IF-NEXT:    ret
   %1 = fcmp ult float %a, %b
   %2 = zext i1 %1 to i32
   ret i32 %2
 }
 
 define i32 @fcmp_ule(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fcmp_ule:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    fmv.w.x ft1, a1
 ; RV32IF-NEXT:    flt.s a0, ft1, ft0
 ; RV32IF-NEXT:    xori a0, a0, 1
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcmp_ule:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    flt.s a0, ft1, ft0
+; RV64IF-NEXT:    xori a0, a0, 1
+; RV64IF-NEXT:    ret
   %1 = fcmp ule float %a, %b
   %2 = zext i1 %1 to i32
   ret i32 %2
 }
 
 define i32 @fcmp_une(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fcmp_une:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    feq.s a0, ft1, ft0
 ; RV32IF-NEXT:    xori a0, a0, 1
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcmp_une:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    feq.s a0, ft1, ft0
+; RV64IF-NEXT:    xori a0, a0, 1
+; RV64IF-NEXT:    ret
   %1 = fcmp une float %a, %b
   %2 = zext i1 %1 to i32
   ret i32 %2
 }
 
 define i32 @fcmp_uno(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fcmp_uno:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    feq.s a1, ft0, ft0
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    feq.s a0, ft0, ft0
 ; RV32IF-NEXT:    and a0, a0, a1
 ; RV32IF-NEXT:    seqz a0, a0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcmp_uno:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    feq.s a1, ft0, ft0
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    feq.s a0, ft0, ft0
+; RV64IF-NEXT:    and a0, a0, a1
+; RV64IF-NEXT:    seqz a0, a0
+; RV64IF-NEXT:    ret
   %1 = fcmp uno float %a, %b
   %2 = zext i1 %1 to i32
   ret i32 %2
 }
 
 define i32 @fcmp_true(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fcmp_true:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    addi a0, zero, 1
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fcmp_true:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    addi a0, zero, 1
+; RV64IF-NEXT:    ret
   %1 = fcmp true float %a, %b
   %2 = zext i1 %1 to i32
   ret i32 %2
 }

llvm/trunk/test/CodeGen/RISCV/float-imm.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -mattr=+f -verify-machineinstrs < %s \
 ; RUN:   | FileCheck -check-prefix=RV32IF %s
+; RUN: llc -mtriple=riscv64 -mattr=+f -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefix=RV64IF %s
 
+; TODO: constant pool shouldn't be necessary for RV64IF.
 define float @float_imm() nounwind {
 ; RV32IF-LABEL: float_imm:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    lui a0, 263313
 ; RV32IF-NEXT:    addi a0, a0, -37
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: float_imm:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    lui a0, %hi(.LCPI0_0)
+; RV64IF-NEXT:    addi a0, a0, %lo(.LCPI0_0)
+; RV64IF-NEXT:    flw ft0, 0(a0)
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   ret float 3.14159274101257324218750
 }
 
 define float @float_imm_op(float %a) nounwind {
 ; TODO: addi should be folded in to the flw
 ; RV32IF-LABEL: float_imm_op:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    lui a0, %hi(.LCPI1_0)
 ; RV32IF-NEXT:    addi a0, a0, %lo(.LCPI1_0)
 ; RV32IF-NEXT:    flw ft1, 0(a0)
 ; RV32IF-NEXT:    fadd.s ft0, ft0, ft1
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: float_imm_op:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    lui a0, %hi(.LCPI1_0)
+; RV64IF-NEXT:    addi a0, a0, %lo(.LCPI1_0)
+; RV64IF-NEXT:    flw ft1, 0(a0)
+; RV64IF-NEXT:    fadd.s ft0, ft0, ft1
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = fadd float %a, 1.0
   ret float %1
 }

llvm/trunk/test/CodeGen/RISCV/float-mem.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -mattr=+f -verify-machineinstrs < %s \
 ; RUN:   | FileCheck -check-prefix=RV32IF %s
+; RUN: llc -mtriple=riscv64 -mattr=+f -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefix=RV64IF %s
 
 define float @flw(float *%a) nounwind {
 ; RV32IF-LABEL: flw:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    flw ft0, 12(a0)
 ; RV32IF-NEXT:    flw ft1, 0(a0)
 ; RV32IF-NEXT:    fadd.s ft0, ft1, ft0
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: flw:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    flw ft0, 12(a0)
+; RV64IF-NEXT:    flw ft1, 0(a0)
+; RV64IF-NEXT:    fadd.s ft0, ft1, ft0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = load float, float* %a
   %2 = getelementptr float, float* %a, i32 3
   %3 = load float, float* %2
 ; Use both loaded values in an FP op to ensure an flw is used, even for the
 ; soft float ABI
   %4 = fadd float %1, %3
   ret float %4
 }
 
 define void @fsw(float *%a, float %b, float %c) nounwind {
 ; Use %b and %c in an FP op to ensure floating point registers are used, even
 ; for the soft float ABI
 ; RV32IF-LABEL: fsw:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a2
 ; RV32IF-NEXT:    fmv.w.x ft1, a1
 ; RV32IF-NEXT:    fadd.s ft0, ft1, ft0
 ; RV32IF-NEXT:    fsw ft0, 32(a0)
 ; RV32IF-NEXT:    fsw ft0, 0(a0)
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fsw:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a2
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    fadd.s ft0, ft1, ft0
+; RV64IF-NEXT:    fsw ft0, 32(a0)
+; RV64IF-NEXT:    fsw ft0, 0(a0)
+; RV64IF-NEXT:    ret
   %1 = fadd float %b, %c
   store float %1, float* %a
   %2 = getelementptr float, float* %a, i32 8
   store float %1, float* %2
   ret void
 }
 
 ; Check load and store to a global

 ; RV32IF-NEXT:    lui a0, %hi(G)
 ; RV32IF-NEXT:    flw ft1, %lo(G)(a0)
 ; RV32IF-NEXT:    fsw ft0, %lo(G)(a0)
 ; RV32IF-NEXT:    addi a0, a0, %lo(G)
 ; RV32IF-NEXT:    flw ft1, 36(a0)
 ; RV32IF-NEXT:    fsw ft0, 36(a0)
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: flw_fsw_global:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    fadd.s ft0, ft1, ft0
+; RV64IF-NEXT:    lui a0, %hi(G)
+; RV64IF-NEXT:    flw ft1, %lo(G)(a0)
+; RV64IF-NEXT:    fsw ft0, %lo(G)(a0)
+; RV64IF-NEXT:    addi a0, a0, %lo(G)
+; RV64IF-NEXT:    flw ft1, 36(a0)
+; RV64IF-NEXT:    fsw ft0, 36(a0)
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = fadd float %a, %b
   %2 = load volatile float, float* @G
   store float %1, float* @G
   %3 = getelementptr float, float* @G, i32 9
   %4 = load volatile float, float* %3
   store float %1, float* %3
   ret float %1
 }
 
 ; Ensure that 1 is added to the high 20 bits if bit 11 of the low part is 1
 define float @flw_fsw_constant(float %a) nounwind {
 ; RV32IF-LABEL: flw_fsw_constant:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    lui a0, 912092
 ; RV32IF-NEXT:    flw ft1, -273(a0)
 ; RV32IF-NEXT:    fadd.s ft0, ft0, ft1
 ; RV32IF-NEXT:    fsw ft0, -273(a0)
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: flw_fsw_constant:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    lui a0, 56
+; RV64IF-NEXT:    addiw a0, a0, -1353
+; RV64IF-NEXT:    slli a0, a0, 14
+; RV64IF-NEXT:    flw ft1, -273(a0)
+; RV64IF-NEXT:    fadd.s ft0, ft0, ft1
+; RV64IF-NEXT:    fsw ft0, -273(a0)
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = inttoptr i32 3735928559 to float*
   %2 = load volatile float, float* %1
   %3 = fadd float %a, %2
   store float %3, float* %1
   ret float %3
 }
 
 declare void @notdead(i8*)

 ; RV32IF-NEXT:    fmv.w.x ft0, s1
 ; RV32IF-NEXT:    flw ft1, 4(sp)
 ; RV32IF-NEXT:    fadd.s ft0, ft1, ft0
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    lw s1, 8(sp)
 ; RV32IF-NEXT:    lw ra, 12(sp)
 ; RV32IF-NEXT:    addi sp, sp, 16
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: flw_stack:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    addi sp, sp, -32
+; RV64IF-NEXT:    sd ra, 24(sp)
+; RV64IF-NEXT:    sd s1, 16(sp)
+; RV64IF-NEXT:    mv s1, a0
+; RV64IF-NEXT:    addi a0, sp, 12
+; RV64IF-NEXT:    call notdead
+; RV64IF-NEXT:    fmv.w.x ft0, s1
+; RV64IF-NEXT:    flw ft1, 12(sp)
+; RV64IF-NEXT:    fadd.s ft0, ft1, ft0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ld s1, 16(sp)
+; RV64IF-NEXT:    ld ra, 24(sp)
+; RV64IF-NEXT:    addi sp, sp, 32
+; RV64IF-NEXT:    ret
   %1 = alloca float, align 4
   %2 = bitcast float* %1 to i8*
   call void @notdead(i8* %2)
   %3 = load float, float* %1
   %4 = fadd float %3, %a ; force load in to FPR32
   ret float %4
 }
 
 define void @fsw_stack(float %a, float %b) nounwind {
 ; RV32IF-LABEL: fsw_stack:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    addi sp, sp, -16
 ; RV32IF-NEXT:    sw ra, 12(sp)
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    fadd.s ft0, ft1, ft0
 ; RV32IF-NEXT:    fsw ft0, 8(sp)
 ; RV32IF-NEXT:    addi a0, sp, 8
 ; RV32IF-NEXT:    call notdead
 ; RV32IF-NEXT:    lw ra, 12(sp)
 ; RV32IF-NEXT:    addi sp, sp, 16
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: fsw_stack:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    addi sp, sp, -16
+; RV64IF-NEXT:    sd ra, 8(sp)
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    fadd.s ft0, ft1, ft0
+; RV64IF-NEXT:    fsw ft0, 4(sp)
+; RV64IF-NEXT:    addi a0, sp, 4
+; RV64IF-NEXT:    call notdead
+; RV64IF-NEXT:    ld ra, 8(sp)
+; RV64IF-NEXT:    addi sp, sp, 16
+; RV64IF-NEXT:    ret
   %1 = fadd float %a, %b ; force store from FPR32
   %2 = alloca float, align 4
   store float %1, float* %2
   %3 = bitcast float* %2 to i8*
   call void @notdead(i8* %3)
   ret void
 }

llvm/trunk/test/CodeGen/RISCV/float-select-fcmp.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -mattr=+f -verify-machineinstrs < %s \
 ; RUN:   | FileCheck -check-prefix=RV32IF %s
+; RUN: llc -mtriple=riscv64 -mattr=+f -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefix=RV64IF %s
 
 define float @select_fcmp_false(float %a, float %b) nounwind {
 ; RV32IF-LABEL: select_fcmp_false:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    mv a0, a1
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: select_fcmp_false:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    mv a0, a1
+; RV64IF-NEXT:    ret
   %1 = fcmp false float %a, %b
   %2 = select i1 %1, float %a, float %b
   ret float %2
 }
 
 define float @select_fcmp_oeq(float %a, float %b) nounwind {
 ; RV32IF-LABEL: select_fcmp_oeq:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft1, a1
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    feq.s a0, ft0, ft1
 ; RV32IF-NEXT:    bnez a0, .LBB1_2
 ; RV32IF-NEXT:  # %bb.1:
 ; RV32IF-NEXT:    fmv.s ft0, ft1
 ; RV32IF-NEXT:  .LBB1_2:
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: select_fcmp_oeq:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    feq.s a0, ft0, ft1
+; RV64IF-NEXT:    bnez a0, .LBB1_2
+; RV64IF-NEXT:  # %bb.1:
+; RV64IF-NEXT:    fmv.s ft0, ft1
+; RV64IF-NEXT:  .LBB1_2:
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = fcmp oeq float %a, %b
   %2 = select i1 %1, float %a, float %b
   ret float %2
 }
 
 define float @select_fcmp_ogt(float %a, float %b) nounwind {
 ; RV32IF-LABEL: select_fcmp_ogt:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    fmv.w.x ft1, a1
 ; RV32IF-NEXT:    flt.s a0, ft1, ft0
 ; RV32IF-NEXT:    bnez a0, .LBB2_2
 ; RV32IF-NEXT:  # %bb.1:
 ; RV32IF-NEXT:    fmv.s ft0, ft1
 ; RV32IF-NEXT:  .LBB2_2:
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: select_fcmp_ogt:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    flt.s a0, ft1, ft0
+; RV64IF-NEXT:    bnez a0, .LBB2_2
+; RV64IF-NEXT:  # %bb.1:
+; RV64IF-NEXT:    fmv.s ft0, ft1
+; RV64IF-NEXT:  .LBB2_2:
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = fcmp ogt float %a, %b
   %2 = select i1 %1, float %a, float %b
   ret float %2
 }
 
 define float @select_fcmp_oge(float %a, float %b) nounwind {
 ; RV32IF-LABEL: select_fcmp_oge:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    fmv.w.x ft1, a1
 ; RV32IF-NEXT:    fle.s a0, ft1, ft0
 ; RV32IF-NEXT:    bnez a0, .LBB3_2
 ; RV32IF-NEXT:  # %bb.1:
 ; RV32IF-NEXT:    fmv.s ft0, ft1
 ; RV32IF-NEXT:  .LBB3_2:
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: select_fcmp_oge:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    fle.s a0, ft1, ft0
+; RV64IF-NEXT:    bnez a0, .LBB3_2
+; RV64IF-NEXT:  # %bb.1:
+; RV64IF-NEXT:    fmv.s ft0, ft1
+; RV64IF-NEXT:  .LBB3_2:
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = fcmp oge float %a, %b
   %2 = select i1 %1, float %a, float %b
   ret float %2
 }
 
 define float @select_fcmp_olt(float %a, float %b) nounwind {
 ; RV32IF-LABEL: select_fcmp_olt:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft1, a1
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    flt.s a0, ft0, ft1
 ; RV32IF-NEXT:    bnez a0, .LBB4_2
 ; RV32IF-NEXT:  # %bb.1:
 ; RV32IF-NEXT:    fmv.s ft0, ft1
 ; RV32IF-NEXT:  .LBB4_2:
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: select_fcmp_olt:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    flt.s a0, ft0, ft1
+; RV64IF-NEXT:    bnez a0, .LBB4_2
+; RV64IF-NEXT:  # %bb.1:
+; RV64IF-NEXT:    fmv.s ft0, ft1
+; RV64IF-NEXT:  .LBB4_2:
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = fcmp olt float %a, %b
   %2 = select i1 %1, float %a, float %b
   ret float %2
 }
 
 define float @select_fcmp_ole(float %a, float %b) nounwind {
 ; RV32IF-LABEL: select_fcmp_ole:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft1, a1
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    fle.s a0, ft0, ft1
 ; RV32IF-NEXT:    bnez a0, .LBB5_2
 ; RV32IF-NEXT:  # %bb.1:
 ; RV32IF-NEXT:    fmv.s ft0, ft1
 ; RV32IF-NEXT:  .LBB5_2:
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: select_fcmp_ole:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fle.s a0, ft0, ft1
+; RV64IF-NEXT:    bnez a0, .LBB5_2
+; RV64IF-NEXT:  # %bb.1:
+; RV64IF-NEXT:    fmv.s ft0, ft1
+; RV64IF-NEXT:  .LBB5_2:
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = fcmp ole float %a, %b
   %2 = select i1 %1, float %a, float %b
   ret float %2
 }
 
 define float @select_fcmp_one(float %a, float %b) nounwind {
 ; TODO: feq.s+sltiu+bne sequence could be optimised
 ; RV32IF-LABEL: select_fcmp_one:

 ; RV32IF-NEXT:    xori a0, a0, 1
 ; RV32IF-NEXT:    and a0, a1, a0
 ; RV32IF-NEXT:    bnez a0, .LBB6_2
 ; RV32IF-NEXT:  # %bb.1:
 ; RV32IF-NEXT:    fmv.s ft0, ft1
 ; RV32IF-NEXT:  .LBB6_2:
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: select_fcmp_one:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    feq.s a0, ft1, ft1
+; RV64IF-NEXT:    feq.s a1, ft0, ft0
+; RV64IF-NEXT:    and a0, a1, a0
+; RV64IF-NEXT:    feq.s a1, ft0, ft1
+; RV64IF-NEXT:    not a1, a1
+; RV64IF-NEXT:    seqz a0, a0
+; RV64IF-NEXT:    xori a0, a0, 1
+; RV64IF-NEXT:    and a0, a1, a0
+; RV64IF-NEXT:    bnez a0, .LBB6_2
+; RV64IF-NEXT:  # %bb.1:
+; RV64IF-NEXT:    fmv.s ft0, ft1
+; RV64IF-NEXT:  .LBB6_2:
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = fcmp one float %a, %b
   %2 = select i1 %1, float %a, float %b
   ret float %2
 }
 
 define float @select_fcmp_ord(float %a, float %b) nounwind {
 ; RV32IF-LABEL: select_fcmp_ord:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    fmv.w.x ft1, a1
 ; RV32IF-NEXT:    feq.s a0, ft1, ft1
 ; RV32IF-NEXT:    feq.s a1, ft0, ft0
 ; RV32IF-NEXT:    and a0, a1, a0
 ; RV32IF-NEXT:    seqz a0, a0
 ; RV32IF-NEXT:    xori a0, a0, 1
 ; RV32IF-NEXT:    bnez a0, .LBB7_2
 ; RV32IF-NEXT:  # %bb.1:
 ; RV32IF-NEXT:    fmv.s ft0, ft1
 ; RV32IF-NEXT:  .LBB7_2:
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: select_fcmp_ord:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    feq.s a0, ft1, ft1
+; RV64IF-NEXT:    feq.s a1, ft0, ft0
+; RV64IF-NEXT:    and a0, a1, a0
+; RV64IF-NEXT:    seqz a0, a0
+; RV64IF-NEXT:    xori a0, a0, 1
+; RV64IF-NEXT:    bnez a0, .LBB7_2
+; RV64IF-NEXT:  # %bb.1:
+; RV64IF-NEXT:    fmv.s ft0, ft1
+; RV64IF-NEXT:  .LBB7_2:
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = fcmp ord float %a, %b
   %2 = select i1 %1, float %a, float %b
   ret float %2
 }
 
 define float @select_fcmp_ueq(float %a, float %b) nounwind {
 ; RV32IF-LABEL: select_fcmp_ueq:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    fmv.w.x ft1, a1
 ; RV32IF-NEXT:    feq.s a0, ft1, ft1
 ; RV32IF-NEXT:    feq.s a1, ft0, ft0
 ; RV32IF-NEXT:    and a0, a1, a0
 ; RV32IF-NEXT:    seqz a0, a0
 ; RV32IF-NEXT:    feq.s a1, ft0, ft1
 ; RV32IF-NEXT:    or a0, a1, a0
 ; RV32IF-NEXT:    bnez a0, .LBB8_2
 ; RV32IF-NEXT:  # %bb.1:
 ; RV32IF-NEXT:    fmv.s ft0, ft1
 ; RV32IF-NEXT:  .LBB8_2:
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: select_fcmp_ueq:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    feq.s a0, ft1, ft1
+; RV64IF-NEXT:    feq.s a1, ft0, ft0
+; RV64IF-NEXT:    and a0, a1, a0
+; RV64IF-NEXT:    seqz a0, a0
+; RV64IF-NEXT:    feq.s a1, ft0, ft1
+; RV64IF-NEXT:    or a0, a1, a0
+; RV64IF-NEXT:    bnez a0, .LBB8_2
+; RV64IF-NEXT:  # %bb.1:
+; RV64IF-NEXT:    fmv.s ft0, ft1
+; RV64IF-NEXT:  .LBB8_2:
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = fcmp ueq float %a, %b
   %2 = select i1 %1, float %a, float %b
   ret float %2
 }
 
 define float @select_fcmp_ugt(float %a, float %b) nounwind {
 ; RV32IF-LABEL: select_fcmp_ugt:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft1, a1
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    fle.s a0, ft0, ft1
 ; RV32IF-NEXT:    xori a0, a0, 1
 ; RV32IF-NEXT:    bnez a0, .LBB9_2
 ; RV32IF-NEXT:  # %bb.1:
 ; RV32IF-NEXT:    fmv.s ft0, ft1
 ; RV32IF-NEXT:  .LBB9_2:
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: select_fcmp_ugt:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fle.s a0, ft0, ft1
+; RV64IF-NEXT:    xori a0, a0, 1
+; RV64IF-NEXT:    bnez a0, .LBB9_2
+; RV64IF-NEXT:  # %bb.1:
+; RV64IF-NEXT:    fmv.s ft0, ft1
+; RV64IF-NEXT:  .LBB9_2:
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = fcmp ugt float %a, %b
   %2 = select i1 %1, float %a, float %b
   ret float %2
 }
 
 define float @select_fcmp_uge(float %a, float %b) nounwind {
 ; RV32IF-LABEL: select_fcmp_uge:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft1, a1
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    flt.s a0, ft0, ft1
 ; RV32IF-NEXT:    xori a0, a0, 1
 ; RV32IF-NEXT:    bnez a0, .LBB10_2
 ; RV32IF-NEXT:  # %bb.1:
 ; RV32IF-NEXT:    fmv.s ft0, ft1
 ; RV32IF-NEXT:  .LBB10_2:
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: select_fcmp_uge:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    flt.s a0, ft0, ft1
+; RV64IF-NEXT:    xori a0, a0, 1
+; RV64IF-NEXT:    bnez a0, .LBB10_2
+; RV64IF-NEXT:  # %bb.1:
+; RV64IF-NEXT:    fmv.s ft0, ft1
+; RV64IF-NEXT:  .LBB10_2:
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = fcmp uge float %a, %b
   %2 = select i1 %1, float %a, float %b
   ret float %2
 }
 
 define float @select_fcmp_ult(float %a, float %b) nounwind {
 ; RV32IF-LABEL: select_fcmp_ult:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    fmv.w.x ft1, a1
 ; RV32IF-NEXT:    fle.s a0, ft1, ft0
 ; RV32IF-NEXT:    xori a0, a0, 1
 ; RV32IF-NEXT:    bnez a0, .LBB11_2
 ; RV32IF-NEXT:  # %bb.1:
 ; RV32IF-NEXT:    fmv.s ft0, ft1
 ; RV32IF-NEXT:  .LBB11_2:
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: select_fcmp_ult:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    fle.s a0, ft1, ft0
+; RV64IF-NEXT:    xori a0, a0, 1
+; RV64IF-NEXT:    bnez a0, .LBB11_2
+; RV64IF-NEXT:  # %bb.1:
+; RV64IF-NEXT:    fmv.s ft0, ft1
+; RV64IF-NEXT:  .LBB11_2:
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = fcmp ult float %a, %b
   %2 = select i1 %1, float %a, float %b
   ret float %2
 }
 
 define float @select_fcmp_ule(float %a, float %b) nounwind {
 ; RV32IF-LABEL: select_fcmp_ule:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    fmv.w.x ft1, a1
 ; RV32IF-NEXT:    flt.s a0, ft1, ft0
 ; RV32IF-NEXT:    xori a0, a0, 1
 ; RV32IF-NEXT:    bnez a0, .LBB12_2
 ; RV32IF-NEXT:  # %bb.1:
 ; RV32IF-NEXT:    fmv.s ft0, ft1
 ; RV32IF-NEXT:  .LBB12_2:
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: select_fcmp_ule:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    flt.s a0, ft1, ft0
+; RV64IF-NEXT:    xori a0, a0, 1
+; RV64IF-NEXT:    bnez a0, .LBB12_2
+; RV64IF-NEXT:  # %bb.1:
+; RV64IF-NEXT:    fmv.s ft0, ft1
+; RV64IF-NEXT:  .LBB12_2:
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = fcmp ule float %a, %b
   %2 = select i1 %1, float %a, float %b
   ret float %2
 }
 
 define float @select_fcmp_une(float %a, float %b) nounwind {
 ; RV32IF-LABEL: select_fcmp_une:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft1, a1
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    feq.s a0, ft0, ft1
 ; RV32IF-NEXT:    xori a0, a0, 1
 ; RV32IF-NEXT:    bnez a0, .LBB13_2
 ; RV32IF-NEXT:  # %bb.1:
 ; RV32IF-NEXT:    fmv.s ft0, ft1
 ; RV32IF-NEXT:  .LBB13_2:
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: select_fcmp_une:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    feq.s a0, ft0, ft1
+; RV64IF-NEXT:    xori a0, a0, 1
+; RV64IF-NEXT:    bnez a0, .LBB13_2
+; RV64IF-NEXT:  # %bb.1:
+; RV64IF-NEXT:    fmv.s ft0, ft1
+; RV64IF-NEXT:  .LBB13_2:
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = fcmp une float %a, %b
   %2 = select i1 %1, float %a, float %b
   ret float %2
 }
 
 define float @select_fcmp_uno(float %a, float %b) nounwind {
 ; TODO: sltiu+bne could be optimized
 ; RV32IF-LABEL: select_fcmp_uno:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a0
 ; RV32IF-NEXT:    fmv.w.x ft1, a1
 ; RV32IF-NEXT:    feq.s a0, ft1, ft1
 ; RV32IF-NEXT:    feq.s a1, ft0, ft0
 ; RV32IF-NEXT:    and a0, a1, a0
 ; RV32IF-NEXT:    seqz a0, a0
 ; RV32IF-NEXT:    bnez a0, .LBB14_2
 ; RV32IF-NEXT:  # %bb.1:
 ; RV32IF-NEXT:    fmv.s ft0, ft1
 ; RV32IF-NEXT:  .LBB14_2:
 ; RV32IF-NEXT:    fmv.x.w a0, ft0
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: select_fcmp_uno:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fmv.w.x ft1, a1
+; RV64IF-NEXT:    feq.s a0, ft1, ft1
+; RV64IF-NEXT:    feq.s a1, ft0, ft0
+; RV64IF-NEXT:    and a0, a1, a0
+; RV64IF-NEXT:    seqz a0, a0
+; RV64IF-NEXT:    bnez a0, .LBB14_2
+; RV64IF-NEXT:  # %bb.1:
+; RV64IF-NEXT:    fmv.s ft0, ft1
+; RV64IF-NEXT:  .LBB14_2:
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
   %1 = fcmp uno float %a, %b
   %2 = select i1 %1, float %a, float %b
   ret float %2
 }
 
 define float @select_fcmp_true(float %a, float %b) nounwind {
 ; RV32IF-LABEL: select_fcmp_true:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: select_fcmp_true:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    ret
   %1 = fcmp true float %a, %b
   %2 = select i1 %1, float %a, float %b
   ret float %2
 }
 
 ; Ensure that ISel succeeds for a select+fcmp that has an i32 result type.
 define i32 @i32_select_fcmp_oeq(float %a, float %b, i32 %c, i32 %d) nounwind {
 ; RV32IF-LABEL: i32_select_fcmp_oeq:
 ; RV32IF:       # %bb.0:
 ; RV32IF-NEXT:    fmv.w.x ft0, a1
 ; RV32IF-NEXT:    fmv.w.x ft1, a0
 ; RV32IF-NEXT:    feq.s a0, ft1, ft0
 ; RV32IF-NEXT:    bnez a0, .LBB16_2
 ; RV32IF-NEXT:  # %bb.1:
 ; RV32IF-NEXT:    mv a2, a3
 ; RV32IF-NEXT:  .LBB16_2:
 ; RV32IF-NEXT:    mv a0, a2
 ; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: i32_select_fcmp_oeq:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    feq.s a0, ft1, ft0
+; RV64IF-NEXT:    bnez a0, .LBB16_2
+; RV64IF-NEXT:  # %bb.1:
+; RV64IF-NEXT:    mv a2, a3
+; RV64IF-NEXT:  .LBB16_2:
+; RV64IF-NEXT:    mv a0, a2
+; RV64IF-NEXT:    ret
   %1 = fcmp oeq float %a, %b
   %2 = select i1 %1, i32 %c, i32 %d
   ret i32 %2
 }

llvm/trunk/test/CodeGen/RISCV/rv32i-rv64i-float-double.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefix=RV32IF %s
+; RUN: llc -mtriple=riscv64 -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefix=RV64IF %s
+
+; This file provides a simple sanity check of float and double operations for
+; RV32I and RV64I. This is primarily intended to ensure that custom
+; legalisation or DAG combines aren't incorrectly triggered when the F
+; extension isn't enabled.
+
+; TODO: f32 parameters on RV64 with a soft-float ABI are anyext.
+
+define float @float_test(float %a, float %b) nounwind {
+; RV32IF-LABEL: float_test:
+; RV32IF:       # %bb.0:
+; RV32IF-NEXT:    addi sp, sp, -16
+; RV32IF-NEXT:    sw ra, 12(sp)
+; RV32IF-NEXT:    sw s1, 8(sp)
+; RV32IF-NEXT:    mv s1, a1
+; RV32IF-NEXT:    call __addsf3
+; RV32IF-NEXT:    mv a1, s1
+; RV32IF-NEXT:    call __divsf3
+; RV32IF-NEXT:    lw s1, 8(sp)
+; RV32IF-NEXT:    lw ra, 12(sp)
+; RV32IF-NEXT:    addi sp, sp, 16
+; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: float_test:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    addi sp, sp, -16
+; RV64IF-NEXT:    sd ra, 8(sp)
+; RV64IF-NEXT:    sd s1, 0(sp)
+; RV64IF-NEXT:    slli a0, a0, 32
+; RV64IF-NEXT:    srli a0, a0, 32
+; RV64IF-NEXT:    slli a1, a1, 32
+; RV64IF-NEXT:    srli s1, a1, 32
+; RV64IF-NEXT:    mv a1, s1
+; RV64IF-NEXT:    call __addsf3
+; RV64IF-NEXT:    mv a1, s1
+; RV64IF-NEXT:    call __divsf3
+; RV64IF-NEXT:    ld s1, 0(sp)
+; RV64IF-NEXT:    ld ra, 8(sp)
+; RV64IF-NEXT:    addi sp, sp, 16
+; RV64IF-NEXT:    ret
+  %1 = fadd float %a, %b
+  %2 = fdiv float %1, %b
+  ret float %2
+}
+
+define double @double_test(double %a, double %b) nounwind {
+; RV32IF-LABEL: double_test:
+; RV32IF:       # %bb.0:
+; RV32IF-NEXT:    addi sp, sp, -16
+; RV32IF-NEXT:    sw ra, 12(sp)
+; RV32IF-NEXT:    sw s1, 8(sp)
+; RV32IF-NEXT:    sw s2, 4(sp)
+; RV32IF-NEXT:    mv s2, a3
+; RV32IF-NEXT:    mv s1, a2
+; RV32IF-NEXT:    call __adddf3
+; RV32IF-NEXT:    mv a2, s1
+; RV32IF-NEXT:    mv a3, s2
+; RV32IF-NEXT:    call __divdf3
+; RV32IF-NEXT:    lw s2, 4(sp)
+; RV32IF-NEXT:    lw s1, 8(sp)
+; RV32IF-NEXT:    lw ra, 12(sp)
+; RV32IF-NEXT:    addi sp, sp, 16
+; RV32IF-NEXT:    ret
+;
+; RV64IF-LABEL: double_test:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    addi sp, sp, -16
+; RV64IF-NEXT:    sd ra, 8(sp)
+; RV64IF-NEXT:    sd s1, 0(sp)
+; RV64IF-NEXT:    mv s1, a1
+; RV64IF-NEXT:    call __adddf3
+; RV64IF-NEXT:    mv a1, s1
+; RV64IF-NEXT:    call __divdf3
+; RV64IF-NEXT:    ld s1, 0(sp)
+; RV64IF-NEXT:    ld ra, 8(sp)
+; RV64IF-NEXT:    addi sp, sp, 16
+; RV64IF-NEXT:    ret
+  %1 = fadd double %a, %b
+  %2 = fdiv double %1, %b
+  ret double %2
+}

llvm/trunk/test/CodeGen/RISCV/rv64f-float-convert.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv64 -mattr=+f -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s -check-prefix=RV64IF
+
+; This file exhaustively checks float<->i32 conversions. In general,
+; fcvt.l[u].s can be selected instead of fcvt.w[u].s because poison is
+; generated for an fpto[s|u]i conversion if the result doesn't fit in the
+; target type.
+
+define i32 @aext_fptosi(float %a) nounwind {
+; RV64IF-LABEL: aext_fptosi:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fcvt.l.s a0, ft0, rtz
+; RV64IF-NEXT:    ret
+  %1 = fptosi float %a to i32
+  ret i32 %1
+}
+
+define signext i32 @sext_fptosi(float %a) nounwind {
+; RV64IF-LABEL: sext_fptosi:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fcvt.l.s a0, ft0, rtz
+; RV64IF-NEXT:    ret
+  %1 = fptosi float %a to i32
+  ret i32 %1
+}
+
+define zeroext i32 @zext_fptosi(float %a) nounwind {
+; RV64IF-LABEL: zext_fptosi:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fcvt.l.s a0, ft0, rtz
+; RV64IF-NEXT:    slli a0, a0, 32
+; RV64IF-NEXT:    srli a0, a0, 32
+; RV64IF-NEXT:    ret
+  %1 = fptosi float %a to i32
+  ret i32 %1
+}
+
+define i32 @aext_fptoui(float %a) nounwind {
+; RV64IF-LABEL: aext_fptoui:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fcvt.lu.s a0, ft0, rtz
+; RV64IF-NEXT:    ret
+  %1 = fptoui float %a to i32
+  ret i32 %1
+}
+
+define signext i32 @sext_fptoui(float %a) nounwind {
+; RV64IF-LABEL: sext_fptoui:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fcvt.wu.s a0, ft0, rtz
+; RV64IF-NEXT:    ret
+  %1 = fptoui float %a to i32
+  ret i32 %1
+}
+
+define zeroext i32 @zext_fptoui(float %a) nounwind {
+; RV64IF-LABEL: zext_fptoui:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a0
+; RV64IF-NEXT:    fcvt.lu.s a0, ft0, rtz
+; RV64IF-NEXT:    ret
+  %1 = fptoui float %a to i32
+  ret i32 %1
+}
+
+define i32 @bcvt_f32_to_aext_i32(float %a, float %b) nounwind {
+; RV64IF-LABEL: bcvt_f32_to_aext_i32:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    fadd.s ft0, ft1, ft0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
+  %1 = fadd float %a, %b
+  %2 = bitcast float %1 to i32
+  ret i32 %2
+}
+
+define signext i32 @bcvt_f32_to_sext_i32(float %a, float %b) nounwind {
+; RV64IF-LABEL: bcvt_f32_to_sext_i32:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    fadd.s ft0, ft1, ft0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
+  %1 = fadd float %a, %b
+  %2 = bitcast float %1 to i32
+  ret i32 %2
+}
+
+define zeroext i32 @bcvt_f32_to_zext_i32(float %a, float %b) nounwind {
+; RV64IF-LABEL: bcvt_f32_to_zext_i32:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    fadd.s ft0, ft1, ft0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    slli a0, a0, 32
+; RV64IF-NEXT:    srli a0, a0, 32
+; RV64IF-NEXT:    ret
+  %1 = fadd float %a, %b
+  %2 = bitcast float %1 to i32
+  ret i32 %2
+}
+
+define float @bcvt_i64_to_f32_via_i32(i64 %a, i64 %b) nounwind {
+; RV64IF-LABEL: bcvt_i64_to_f32_via_i32:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fmv.w.x ft0, a1
+; RV64IF-NEXT:    fmv.w.x ft1, a0
+; RV64IF-NEXT:    fadd.s ft0, ft1, ft0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
+  %1 = trunc i64 %a to i32
+  %2 = trunc i64 %b to i32
+  %3 = bitcast i32 %1 to float
+  %4 = bitcast i32 %2 to float
+  %5 = fadd float %3, %4
+  ret float %5
+}
+
+define float @uitofp_aext_i32_to_f32(i32 %a) nounwind {
+; RV64IF-LABEL: uitofp_aext_i32_to_f32:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fcvt.s.wu ft0, a0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
+  %1 = uitofp i32 %a to float
+  ret float %1
+}
+
+define float @uitofp_sext_i32_to_f32(i32 signext %a) nounwind {
+; RV64IF-LABEL: uitofp_sext_i32_to_f32:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fcvt.s.wu ft0, a0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
+  %1 = uitofp i32 %a to float
+  ret float %1
+}
+
+define float @uitofp_zext_i32_to_f32(i32 zeroext %a) nounwind {
+; RV64IF-LABEL: uitofp_zext_i32_to_f32:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fcvt.s.wu ft0, a0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
+  %1 = uitofp i32 %a to float
+  ret float %1
+}
+
+define float @sitofp_aext_i32_to_f32(i32 %a) nounwind {
+; RV64IF-LABEL: sitofp_aext_i32_to_f32:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fcvt.s.w ft0, a0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
+  %1 = sitofp i32 %a to float
+  ret float %1
+}
+
+define float @sitofp_sext_i32_to_f32(i32 signext %a) nounwind {
+; RV64IF-LABEL: sitofp_sext_i32_to_f32:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fcvt.s.l ft0, a0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
+  %1 = sitofp i32 %a to float
+  ret float %1
+}
+
+define float @sitofp_zext_i32_to_f32(i32 zeroext %a) nounwind {
+; RV64IF-LABEL: sitofp_zext_i32_to_f32:
+; RV64IF:       # %bb.0:
+; RV64IF-NEXT:    fcvt.s.w ft0, a0
+; RV64IF-NEXT:    fmv.x.w a0, ft0
+; RV64IF-NEXT:    ret
+  %1 = sitofp i32 %a to float
+  ret float %1
+}