[ARM] Set up infrastructure for MVE vector instructions.

[simon_tatham]

Summary

This commit prepares the way to start adding the main collection of
MVE instructions, which operate on the 128-bit vector registers.

The most obvious thing that's needed, and the simplest, is to add the
MQPR register class, which is like the existing QPR except that it has
fewer registers in it.

The more complicated part: MVE defines a system of vector predication,
in which instructions operating on 128-bit vector registers can be
constrained to operate on only a subset of the lanes, using a system
of prefix instructions similar to the existing Thumb IT, in that you
have one prefix instruction which designates up to 4 following
instructions as subject to predication, and within that sequence, the
predicate can be inverted by means of T/E suffixes ('Then' / 'Else').

To support instructions of this type, we've added two new Tablegen
classes vpred_n and vpred_r for standard clusters of MC operands
to add to a predicated instruction. Both include a flag indicating how
the instruction is predicated at all (options are T, E and 'not
predicated'), and an input register field for the register controlling
the set of active lanes. They differ from each other in that vpred_r
also includes an input operand for the previous value of the output
register, for instructions that leave inactive lanes unchanged.
vpred_n lacks that extra operand; it will be used for instructions
that don't preserve inactive lanes in their output register (either
because inactive lanes are zeroed, as the MVE load instructions do, or
because the output register isn't a vector at all).

This commit also adds the family of prefix instructions themselves
(VPT / VPST), and all the machinery needed to work with them in
assembly and disassembly (e.g. generating the 't' and 'e' mnemonic
suffixes on disassembled instructions within a predicated block)

I've added a couple of demo instructions that derive from the new
Tablegen base classes and use those two operand clusters. The bulk of
the vector instructions will come in followup commits small enough to
be manageable. (One exception is that I've added the full version of
isMnemonicVPTPredicable in the AsmParser, because it seemed
pointless to carefully split it up.)

[samparker]

llvm/lib/Target/ARM/ARMRegisterInfo.td

461 ↗

(On Diff #202204)

Is this really necessary..? Maybe it would be clearer to just have the MQPR class for MVE and not touch QPR?

[ostannard]

llvm/lib/Target/ARM/AsmParser/ARMAsmParser.cpp

504 ↗

(On Diff #202204)

mve.fp implies mve.int, so one of these functions is redundant.

2114 ↗

(On Diff #202204)

getCondCode already returns an ARMCC::CondCodes, so these casts aren't needed.

6317 ↗

(On Diff #202204)

MQPR is a subset of QPR, so there's no point in checking both.

6457 ↗

(On Diff #202204)

Why is this inverted compared to IT? Is the format of these masks written down anywhere?

6907 ↗

(On Diff #202204)

Could the bit-twiddling on the VPS state mask be split into a helper function, like currentITCond?

6925 ↗

(On Diff #202204)

Why do we need to check hasMVE here?

llvm/lib/Target/ARM/MCTargetDesc/ARMMCCodeEmitter.cpp

534 ↗

(On Diff #202204)

I think this needs a comment explaining what is different about MVE.

llvm/test/MC/ARM/mve-vpt.s

9 ↗

(On Diff #202204)

Are these instructions testing anything related to VPT blocks? If not, they should be moved to a separate file (and we should check the actual error which is emitted in the non-FP case).

66 ↗

(On Diff #202204)

We should also check the errors which are reported for invalid instructions.

[dmgreen]

llvm/lib/Target/ARM/AsmParser/ARMAsmParser.cpp

2183 ↗

(On Diff #202204)

I think this should be using MII.get(Inst.getOpcode()), if it can.

llvm/lib/Target/ARM/MCTargetDesc/ARMMCCodeEmitter.cpp

535 ↗

(On Diff #202204)

Does HasMVEFloatOps imply HasMVEIntegerOps?

[simon_tatham]

Remastered patch to apply cleanly against current trunk.

[miyuki]

llvm/lib/Target/ARM/MCTargetDesc/ARMMCCodeEmitter.cpp

410 ↗

(On Diff #204023)

GCC complains that this function is unused. It can probably be moved to D62678.

[simon_tatham]

llvm/lib/Target/ARM/AsmParser/ARMAsmParser.cpp

2183 ↗

(On Diff #202204)

Unfortunately, that's rather fiddly, because the context here is that we're in a method of ARMOperand, which doesn't have an MCInstrInfo conveniently accessible that I can find – the ARMAsmParser as a whole is given one at construction time, but it's not passed through to ARMOperand.

I wondered about changing the prototype of this addVPTPredROperands method so that it received the MCInstrDesc from its caller, and then that could look it up in the ambient MCInstrInfo. But the caller is the tablegenerated AsmMatcher, and that has hard-coded function call code for all the RenderMethods which expects them to conform to the same API, so that would be a huge piece of plumbing to change.

I think leaving it as it is is safe from a modules perspective. We're referring directly to the llvm::ARMInsts array, which is defined in the LLVMARMDesc library. But that library also defines the llvm::ARMMCRegisterClasses array, which we refer to constantly in this same source file. So anyone linking against this module must surely also have LLVMARMDesc available, or else they'd already have had a link error.

[dmgreen]

llvm/lib/Target/ARM/AsmParser/ARMAsmParser.cpp

2183 ↗

(On Diff #202204)

Ah, that makes sense. Thanks for checking.

[simon_tatham]

llvm/lib/Target/ARM/AsmParser/ARMAsmParser.cpp

6457 ↗

(On Diff #202204)

Recording the result of our investigation just now:

It's IT, not VPT, that is doing something strange. The ARMOperand coming out of the operand parsing stage has the IT mask in one format; it is translated directly into an MCOperand storing the mask in the same format; but then processInstruction translates it into an MCOperand in a different format. For example, if you feed the instruction itete eq into llvm-mc -debug -triple=thumbv8a -show-inst, you see

Parsed as: <MCInst #2921 t2IT <MCOperand Imm:0> <MCOperand Imm:5>>
        itete   eq                      @ <MCInst #2921 t2IT
                                        @  <MCOperand Imm:0>
                                        @  <MCOperand Imm:11>>

in which you can see that the mask is originally parsed into an MCOperand with value 5, and processInstruction later converts that into 11, without changing the type of structure it's stored in. So that MCOperand has different semantics depending on which part of the assembly pipeline you're in!

Apparently the reason for having two formats in the case of IT is because the transient initial format is what has to be passed to startExplicitITBlock in order to have the assembler statically track the IT block state and validate it against the suffixes on the next few instructions. So processInstruction receives the IT mask in the right format to pass to that function, and then transforms it into the second format.

The VPT code here is avoiding this complication, by using the same format for the mask throughout, in both the ARMOperand and MCOperand structures. So vpstete, for example, would have the mask value 11 from beginning to end of the parsing process. (The architectural encoding is different, but that's taken care of by getVPTMaskOpValue at the point of translation from MCOperand to an instruction encoding, which seems to me the sensible place to do it.)

I think it's very confusing for the immediate value in an MCOperand to have different semantics at different points in the code, so I propose to make IT work more like VPT rather than vice versa: I think all masks in operand structures should be stored in the same format, and processInstruction will have to do a local translation into the format that startExplicitITBlock expects.

[simon_tatham]

I think this revision addresses all the previous review comments. In
particular, IT and VPT mask handling is now identical, because
IT handling was simplified by D63219 (which also introduced that
helper function @ostannard asked for to do the mask bit extraction).

Also I've moved the tests that aren't specifically about VPT into test
files that further MVE assembly patches will add more things to. (Each
of the handful of example instructions in this patch is pulled forward
from one of those later patches, and now when those later patches
land, each one will end up being tested in the same test input files
as its siblings.)

[ostannard]

LGTM, just one minor nit.

llvm/lib/Target/ARM/Disassembler/ARMDisassembler.cpp

126 ↗

(On Diff #204479)

This could be a SmallVector<unsigned char, 4> to avoid some heap allocations.

llvm/trunk/lib/Target/ARM/ARMBaseInstrInfo.h

return MachineOperand::CreateReg(ARM::CPSR,
                                    /*Kill*/ false, isDead);
 }
 
 static inline
 bool isUncondBranchOpcode(int Opc) {
   return Opc == ARM::B || Opc == ARM::tB || Opc == ARM::t2B;
 }
 
+static inline bool isVPTOpcode(int Opc) {
+  return Opc == ARM::t2VPTv16i8 || Opc == ARM::t2VPTv16u8 ||
+         Opc == ARM::t2VPTv16s8 || Opc == ARM::t2VPTv8i16 ||
+         Opc == ARM::t2VPTv8u16 || Opc == ARM::t2VPTv8s16 ||
+         Opc == ARM::t2VPTv4i32 || Opc == ARM::t2VPTv4u32 ||
+         Opc == ARM::t2VPTv4s32 || Opc == ARM::t2VPTv4f32 ||
+         Opc == ARM::t2VPTv8f16 || Opc == ARM::t2VPTv16i8r ||
+         Opc == ARM::t2VPTv16u8r || Opc == ARM::t2VPTv16s8r ||
+         Opc == ARM::t2VPTv8i16r || Opc == ARM::t2VPTv8u16r ||
+         Opc == ARM::t2VPTv8s16r || Opc == ARM::t2VPTv4i32r ||
+         Opc == ARM::t2VPTv4u32r || Opc == ARM::t2VPTv4s32r ||
+         Opc == ARM::t2VPTv4f32r || Opc == ARM::t2VPTv8f16r ||
+         Opc == ARM::t2VPST;
+}
+
 static inline
 bool isCondBranchOpcode(int Opc) {
   return Opc == ARM::Bcc || Opc == ARM::tBcc || Opc == ARM::t2Bcc;
 }
 
 static inline bool isJumpTableBranchOpcode(int Opc) {
   return Opc == ARM::BR_JTr || Opc == ARM::BR_JTm_i12 ||
          Opc == ARM::BR_JTm_rs || Opc == ARM::BR_JTadd || Opc == ARM::tBR_JTr ||

llvm/trunk/lib/Target/ARM/ARMInstrFormats.td

 // Same as cc_out except it defaults to setting CPSR.
 def s_cc_out : OptionalDefOperand<OtherVT, (ops CCR), (ops (i32 CPSR))> {
   let EncoderMethod = "getCCOutOpValue";
   let PrintMethod = "printSBitModifierOperand";
   let ParserMatchClass = CCOutOperand;
   let DecoderMethod = "DecodeCCOutOperand";
 }
 
+// VPT predicate
+
+def VPTPredNOperand : AsmOperandClass {
+  let Name = "VPTPredN";
+  let PredicateMethod = "isVPTPred";
+}
+def VPTPredROperand : AsmOperandClass {
+  let Name = "VPTPredR";
+  let PredicateMethod = "isVPTPred";
+}
+def undef_tied_input;
+
+// Operand classes for the cluster of MC operands describing a
+// VPT-predicated MVE instruction.
+//
+// There are two of these classes. Both of them have the same first
+// two options:
+//
+// $cond (an integer) indicates the instruction's predication status:
+//   * ARMVCC::None means it's unpredicated
+//   * ARMVCC::Then means it's in a VPT block and appears with the T suffix
+//   * ARMVCC::Else means it's in a VPT block and appears with the E suffix.
+// During code generation, unpredicated and predicated instructions
+// are indicated by setting this parameter to 'None' or to 'Then'; the
+// third value 'Else' is only used for assembly and disassembly.
+//
+// $cond_reg (type VCCR) gives the input predicate register. This is
+// always either zero_reg or VPR, but needs to be modelled as an
+// explicit operand so that it can be register-allocated and spilled
+// when these operands are used in code generation).
+//
+// For 'vpred_r', there's an extra operand $inactive, which specifies
+// the vector register which will supply any lanes of the output
+// register that the predication mask prevents from being written by
+// this instruction. It's always tied to the actual output register
+// (i.e. must be allocated into the same physical reg), but again,
+// code generation will need to model it as a separate input value.
+//
+// 'vpred_n' doesn't have that extra operand: it only has $cond and
+// $cond_reg. This variant is used for any instruction that can't, or
+// doesn't want to, tie $inactive to the output register. Sometimes
+// that's because another input parameter is already tied to it (e.g.
+// instructions that both read and write their Qd register even when
+// unpredicated, either because they only partially overwrite it like
+// a narrowing integer conversion, or simply because the instruction
+// encoding doesn't have enough register fields to make the output
+// independent of all inputs). It can also be because the instruction
+// is defined to set disabled output lanes to zero rather than leaving
+// them unchanged (vector loads), or because it doesn't output a
+// vector register at all (stores, compares). In any of these
+// situations it's unnecessary to have an extra operand tied to the
+// output, and inconvenient to leave it there unused.
+
+// Base class for both kinds of vpred.
+class vpred_ops<dag extra_op, dag extra_mi> : OperandWithDefaultOps<OtherVT,
+            !con((ops (i32 0), (i32 zero_reg)), extra_op)> {
+  let PrintMethod = "printVPTPredicateOperand";
+  let OperandNamespace = "ARM";
+  let MIOperandInfo = !con((ops i32imm:$cond, VCCR:$cond_reg), extra_mi);
+
+  // For convenience, we provide a string value that can be appended
+  // to the constraints string. It's empty for vpred_n, and for
+  // vpred_r it ties the $inactive operand to the output q-register
+  // (which by convention will be called $Qd).
+  string vpred_constraint;
+}
+
+def vpred_r : vpred_ops<(ops (v4i32 undef_tied_input)), (ops MQPR:$inactive)> {
+  let ParserMatchClass = VPTPredROperand;
+  let OperandType = "OPERAND_VPRED_R";
+  let DecoderMethod = "DecodeVpredROperand";
+  let vpred_constraint = ",$Qd = $vp.inactive";
+}
+
+def vpred_n : vpred_ops<(ops), (ops)> {
+  let ParserMatchClass = VPTPredNOperand;
+  let OperandType = "OPERAND_VPRED_N";
+  let vpred_constraint = "";
+}
+
 // ARM special operands for disassembly only.
 //
 def SetEndAsmOperand : ImmAsmOperand<0,1> {
   let Name = "SetEndImm";
   let ParserMethod = "parseSetEndImm";
 }
 def setend_op : Operand<i32> {
   let PrintMethod = "printSetendOperand";

llvm/trunk/lib/Target/ARM/ARMInstrMVE.td

 //===-- ARMInstrMVE.td - MVE support for ARM ---------------*- 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 ARM MVE instruction set.
 //
 //===----------------------------------------------------------------------===//
 
+// VPT condition mask
+def vpt_mask : Operand<i32> {
+  let PrintMethod = "printVPTMask";
+  let ParserMatchClass = it_mask_asmoperand;
+  let EncoderMethod = "getVPTMaskOpValue";
+  let DecoderMethod = "DecodeVPTMaskOperand";
+}
+
+// VPT/VCMP restricted predicate for sign invariant types
+def pred_restricted_i_asmoperand : AsmOperandClass {
+  let Name = "CondCodeRestrictedI";
+  let RenderMethod = "addITCondCodeOperands";
+  let PredicateMethod = "isITCondCodeRestrictedI";
+  let ParserMethod = "parseITCondCode";
+}
+
+// VPT/VCMP restricted predicate for signed types
+def pred_restricted_s_asmoperand : AsmOperandClass {
+  let Name = "CondCodeRestrictedS";
+  let RenderMethod = "addITCondCodeOperands";
+  let PredicateMethod = "isITCondCodeRestrictedS";
+  let ParserMethod = "parseITCondCode";
+}
+
+// VPT/VCMP restricted predicate for unsigned types
+def pred_restricted_u_asmoperand : AsmOperandClass {
+  let Name = "CondCodeRestrictedU";
+  let RenderMethod = "addITCondCodeOperands";
+  let PredicateMethod = "isITCondCodeRestrictedU";
+  let ParserMethod = "parseITCondCode";
+}
+
+// VPT/VCMP restricted predicate for floating point
+def pred_restricted_fp_asmoperand : AsmOperandClass {
+  let Name = "CondCodeRestrictedFP";
+  let RenderMethod = "addITCondCodeOperands";
+  let PredicateMethod = "isITCondCodeRestrictedFP";
+  let ParserMethod = "parseITCondCode";
+}
+
+def pred_basic_i : Operand<i32> {
+  let PrintMethod = "printMandatoryRestrictedPredicateOperand";
+  let ParserMatchClass = pred_restricted_i_asmoperand;
+  let DecoderMethod = "DecodeRestrictedIPredicateOperand";
+  let EncoderMethod = "getRestrictedCondCodeOpValue";
+}
+
+def pred_basic_u : Operand<i32> {
+  let PrintMethod = "printMandatoryRestrictedPredicateOperand";
+  let ParserMatchClass = pred_restricted_u_asmoperand;
+  let DecoderMethod = "DecodeRestrictedUPredicateOperand";
+  let EncoderMethod = "getRestrictedCondCodeOpValue";
+}
+
+def pred_basic_s : Operand<i32> {
+  let PrintMethod = "printMandatoryRestrictedPredicateOperand";
+  let ParserMatchClass = pred_restricted_s_asmoperand;
+  let DecoderMethod = "DecodeRestrictedSPredicateOperand";
+  let EncoderMethod = "getRestrictedCondCodeOpValue";
+}
+
+def pred_basic_fp : Operand<i32> {
+  let PrintMethod = "printMandatoryRestrictedPredicateOperand";
+  let ParserMatchClass = pred_restricted_fp_asmoperand;
+  let DecoderMethod = "DecodeRestrictedFPPredicateOperand";
+  let EncoderMethod = "getRestrictedCondCodeOpValue";
+}
+
+class MVE_MI<dag oops, dag iops, InstrItinClass itin, string asm,
+             string ops, string cstr, list<dag> pattern>
+  : Thumb2XI<oops, iops, AddrModeNone, 4, itin, !strconcat(asm, "\t", ops), cstr,
+             pattern>,
+    Requires<[HasMVEInt]> {
+  let D = MVEDomain;
+  let DecoderNamespace = "MVE";
+}
+
+// MVE_p is used for most predicated instructions, to add the cluster
+// of input operands that provides the VPT suffix (none, T or E) and
+// the input predicate register.
+class MVE_p<dag oops, dag iops, InstrItinClass itin, string iname,
+            string suffix, string ops, vpred_ops vpred, string cstr,
+            list<dag> pattern=[]>
+  : MVE_MI<oops, !con(iops, (ins vpred:$vp)), itin,
+           // If the instruction has a suffix, like vadd.f32, then the
+           // VPT predication suffix goes before the dot, so the full
+           // name has to be "vadd${vp}.f32".
+           !strconcat(iname, "${vp}",
+                      !if(!eq(suffix, ""), "", !strconcat(".", suffix))),
+           ops, !strconcat(cstr, vpred.vpred_constraint), pattern> {
+  let Inst{31-29} = 0b111;
+  let Inst{27-26} = 0b11;
+}
+
 class MVE_MI_with_pred<dag oops, dag iops, InstrItinClass itin, string asm,
                        string ops, string cstr, list<dag> pattern>
   : Thumb2I<oops, iops, AddrModeNone, 4, itin, asm, !strconcat("\t", ops), cstr,
              pattern>,
     Requires<[HasV8_1MMainline, HasMVEInt]> {
   let D = MVEDomain;
   let DecoderNamespace = "MVE";
 }

 
 def t2SQRSHRL : t2MVEShiftDRegReg<"sqrshrl", 0b1,  0b1>;
 def t2SQSHLL  : t2MVEShiftDRegImm<"sqshll",  0b11, 0b1>;
 def t2SRSHRL  : t2MVEShiftDRegImm<"srshrl",  0b10, 0b1>;
 
 def t2UQRSHLL : t2MVEShiftDRegReg<"uqrshll", 0b0,  0b1>;
 def t2UQSHLL  : t2MVEShiftDRegImm<"uqshll",  0b00, 0b1>;
 def t2URSHRL  : t2MVEShiftDRegImm<"urshrl",  0b01, 0b1>;
+
+// start of mve_rDest instructions
+
+class MVE_rDest<dag oops, dag iops, InstrItinClass itin, string iname, string suffix,
+                string ops, string cstr, list<dag> pattern=[]>
+// Always use vpred_n and not vpred_r: with the output register being
+// a GPR and not a vector register, there can't be any question of
+// what to put in its inactive lanes.
+  : MVE_p<oops, iops, itin, iname, suffix, ops, vpred_n, cstr, pattern> {
+
+  let Inst{25-23} = 0b101;
+  let Inst{11-9} = 0b111;
+  let Inst{4} = 0b0;
+}
+
+class t2VABAV<string suffix, bit U, bits<2> size, list<dag> pattern=[]>
+  : MVE_rDest<(outs rGPR:$Rda), (ins rGPR:$Rda_src, MQPR:$Qn, MQPR:$Qm),
+              NoItinerary, "vabav", suffix, "$Rda, $Qn, $Qm", "$Rda = $Rda_src",
+              pattern> {
+  bits<4> Qm;
+  bits<4> Qn;
+  bits<4> Rda;
+
+  let Inst{28} = U;
+  let Inst{22} = 0b0;
+  let Inst{21-20} = size{1-0};
+  let Inst{19-17} = Qn{2-0};
+  let Inst{16} = 0b0;
+  let Inst{15-12} = Rda{3-0};
+  let Inst{8} = 0b1;
+  let Inst{7} = Qn{3};
+  let Inst{6} = 0b0;
+  let Inst{5} = Qm{3};
+  let Inst{3-1} = Qm{2-0};
+  let Inst{0} = 0b1;
+}
+
+def VABAVs8  : t2VABAV<"s8", 0b0, 0b00>;
+def VABAVs16 : t2VABAV<"s16", 0b0, 0b01>;
+def VABAVs32 : t2VABAV<"s32", 0b0, 0b10>;
+def VABAVu8  : t2VABAV<"u8", 0b1, 0b00>;
+def VABAVu16 : t2VABAV<"u16", 0b1, 0b01>;
+def VABAVu32 : t2VABAV<"u32", 0b1, 0b10>;
+
+// end of mve_rDest instructions
+
+// start of mve_comp instructions
+
+class MVE_comp<InstrItinClass itin, string iname, string suffix,
+               string cstr, list<dag> pattern=[]>
+  : MVE_p<(outs MQPR:$Qd), (ins MQPR:$Qn, MQPR:$Qm), itin, iname, suffix,
+           "$Qd, $Qn, $Qm", vpred_r, cstr, pattern> {
+  bits<4> Qd;
+  bits<4> Qn;
+  bits<4> Qm;
+
+  let Inst{22} = Qd{3};
+  let Inst{19-17} = Qn{2-0};
+  let Inst{16} = 0b0;
+  let Inst{15-13} = Qd{2-0};
+  let Inst{12} = 0b0;
+  let Inst{10-9} = 0b11;
+  let Inst{7} = Qn{3};
+  let Inst{5} = Qm{3};
+  let Inst{3-1} = Qm{2-0};
+  let Inst{0} = 0b0;
+}
+
+class VMINMAXNM<string iname, string suffix, bit sz, bit bit_21,
+                list<dag> pattern=[]>
+  : MVE_comp<NoItinerary, iname, suffix, "", pattern> {
+
+  let Inst{28} = 0b1;
+  let Inst{25-24} = 0b11;
+  let Inst{23} = 0b0;
+  let Inst{21} = bit_21;
+  let Inst{20} = sz;
+  let Inst{11} = 0b1;
+  let Inst{8} = 0b1;
+  let Inst{6} = 0b1;
+  let Inst{4} = 0b1;
+
+  let Predicates = [HasMVEFloat];
+}
+
+def VMAXNMf32 : VMINMAXNM<"vmaxnm", "f32", 0b0, 0b0>;
+def VMAXNMf16 : VMINMAXNM<"vmaxnm", "f16", 0b1, 0b0>;
+
+def VMINNMf32 : VMINMAXNM<"vminnm", "f32", 0b0, 0b1>;
+def VMINNMf16 : VMINMAXNM<"vminnm", "f16", 0b1, 0b1>;
+
+// end of mve_comp instructions
+
+class t2VPT<string suffix, bits<2> size, dag iops, string asm, list<dag> pattern=[]>
+  : MVE_MI<(outs ), iops, NoItinerary, !strconcat("vpt", "${Mk}", ".", suffix), asm, "", pattern> {
+  bits<3> fc;
+  bits<4> Mk;
+  bits<3> Qn;
+
+  let Inst{31-23} = 0b111111100;
+  let Inst{22} = Mk{3};
+  let Inst{21-20} = size;
+  let Inst{19-17} = Qn{2-0};
+  let Inst{16} = 0b1;
+  let Inst{15-13} = Mk{2-0};
+  let Inst{12} = fc{2};
+  let Inst{11-8} = 0b1111;
+  let Inst{7} = fc{0};
+  let Inst{4} = 0b0;
+
+  let Defs = [VPR, P0];
+}
+
+class t2VPTt1<string suffix, bits<2> size, dag iops>
+  : t2VPT<suffix, size, iops, "$fc, $Qn, $Qm"> {
+  bits<4> Qm;
+  bits<4> Mk;
+
+  let Inst{6} = 0b0;
+  let Inst{5} = Qm{3};
+  let Inst{3-1} = Qm{2-0};
+  let Inst{0} = fc{1};
+}
+
+class t2VPTt1i<string suffix, bits<2> size>
+ : t2VPTt1<suffix, size,
+           (ins vpt_mask:$Mk, pred_basic_i:$fc, MQPR:$Qn, MQPR:$Qm)> {
+  let Inst{12} = 0b0;
+  let Inst{0} = 0b0;
+}
+
+def t2VPTv4i32 : t2VPTt1i<"i32", 0b10>;
+def t2VPTv8i16 : t2VPTt1i<"i16", 0b01>;
+def t2VPTv16i8 : t2VPTt1i<"i8", 0b00>;
+
+class t2VPTt1u<string suffix, bits<2> size>
+ : t2VPTt1<suffix, size,
+           (ins vpt_mask:$Mk, pred_basic_u:$fc, MQPR:$Qn, MQPR:$Qm)> {
+  let Inst{12} = 0b0;
+  let Inst{0} = 0b1;
+}
+
+def t2VPTv4u32 : t2VPTt1u<"u32", 0b10>;
+def t2VPTv8u16 : t2VPTt1u<"u16", 0b01>;
+def t2VPTv16u8 : t2VPTt1u<"u8", 0b00>;
+
+class t2VPTt1s<string suffix, bits<2> size>
+ : t2VPTt1<suffix, size,
+           (ins vpt_mask:$Mk, pred_basic_s:$fc, MQPR:$Qn, MQPR:$Qm)> {
+  let Inst{12} = 0b1;
+}
+
+def t2VPTv4s32 : t2VPTt1s<"s32", 0b10>;
+def t2VPTv8s16 : t2VPTt1s<"s16", 0b01>;
+def t2VPTv16s8 : t2VPTt1s<"s8", 0b00>;
+
+class t2VPTt2<string suffix, bits<2> size, dag iops>
+  : t2VPT<suffix, size, iops,
+          "$fc, $Qn, $Rm"> {
+  bits<4> Rm;
+  bits<3> fc;
+  bits<4> Mk;
+
+  let Inst{6} = 0b1;
+  let Inst{5} = fc{1};
+  let Inst{3-0} = Rm{3-0};
+}
+
+class t2VPTt2i<string suffix, bits<2> size>
+  : t2VPTt2<suffix, size,
+            (ins vpt_mask:$Mk, pred_basic_i:$fc, MQPR:$Qn, GPRwithZR:$Rm)> {
+  let Inst{12} = 0b0;
+  let Inst{5} = 0b0;
+}
+
+def t2VPTv4i32r : t2VPTt2i<"i32", 0b10>;
+def t2VPTv8i16r : t2VPTt2i<"i16", 0b01>;
+def t2VPTv16i8r : t2VPTt2i<"i8", 0b00>;
+
+class t2VPTt2u<string suffix, bits<2> size>
+  : t2VPTt2<suffix, size,
+            (ins vpt_mask:$Mk, pred_basic_u:$fc, MQPR:$Qn, GPRwithZR:$Rm)> {
+  let Inst{12} = 0b0;
+  let Inst{5} = 0b1;
+}
+
+def t2VPTv4u32r : t2VPTt2u<"u32", 0b10>;
+def t2VPTv8u16r : t2VPTt2u<"u16", 0b01>;
+def t2VPTv16u8r : t2VPTt2u<"u8", 0b00>;
+
+class t2VPTt2s<string suffix, bits<2> size>
+  : t2VPTt2<suffix, size,
+            (ins vpt_mask:$Mk, pred_basic_s:$fc, MQPR:$Qn, GPRwithZR:$Rm)> {
+  let Inst{12} = 0b1;
+}
+
+def t2VPTv4s32r : t2VPTt2s<"s32", 0b10>;
+def t2VPTv8s16r : t2VPTt2s<"s16", 0b01>;
+def t2VPTv16s8r : t2VPTt2s<"s8", 0b00>;
+
+
+class t2VPTf<string suffix, bit size, dag iops, string asm, list<dag> pattern=[]>
+  : MVE_MI<(outs ), iops, NoItinerary, !strconcat("vpt", "${Mk}", ".", suffix), asm,
+            "", pattern> {
+  bits<3> fc;
+  bits<4> Mk;
+  bits<3> Qn;
+
+  let Inst{31-29} = 0b111;
+  let Inst{28} = size;
+  let Inst{27-23} = 0b11100;
+  let Inst{22} = Mk{3};
+  let Inst{21-20} = 0b11;
+  let Inst{19-17} = Qn{2-0};
+  let Inst{16} = 0b1;
+  let Inst{15-13} = Mk{2-0};
+  let Inst{12} = fc{2};
+  let Inst{11-8} = 0b1111;
+  let Inst{7} = fc{0};
+  let Inst{4} = 0b0;
+
+  let Defs = [P0];
+  let Predicates = [HasMVEFloat];
+}
+
+class t2VPTft1<string suffix, bit size>
+  : t2VPTf<suffix, size, (ins vpt_mask:$Mk, pred_basic_fp:$fc, MQPR:$Qn, MQPR:$Qm),
+          "$fc, $Qn, $Qm"> {
+  bits<3> fc;
+  bits<4> Qm;
+
+  let Inst{6} = 0b0;
+  let Inst{5} = Qm{3};
+  let Inst{3-1} = Qm{2-0};
+  let Inst{0} = fc{1};
+}
+
+def t2VPTv4f32         : t2VPTft1<"f32", 0b0>;
+def t2VPTv8f16         : t2VPTft1<"f16", 0b1>;
+
+class t2VPTft2<string suffix, bit size>
+  : t2VPTf<suffix, size, (ins vpt_mask:$Mk, pred_basic_fp:$fc, MQPR:$Qn, GPRwithZR:$Rm),
+          "$fc, $Qn, $Rm"> {
+  bits<3> fc;
+  bits<4> Rm;
+
+  let Inst{6} = 0b1;
+  let Inst{5} = fc{1};
+  let Inst{3-0} = Rm{3-0};
+}
+
+def t2VPTv4f32r        : t2VPTft2<"f32", 0b0>;
+def t2VPTv8f16r        : t2VPTft2<"f16", 0b1>;
+
+def t2VPST : MVE_MI<(outs ), (ins vpt_mask:$Mk), NoItinerary,
+       !strconcat("vpst", "${Mk}"), "", "", []> {
+  bits<4> Mk;
+
+  let Inst{31-23} = 0b111111100;
+  let Inst{22} = Mk{3};
+  let Inst{21-16} = 0b110001;
+  let Inst{15-13} = Mk{2-0};
+  let Inst{12-0} = 0b0111101001101;
+  let Unpredictable{12} = 0b1;
+  let Unpredictable{7} = 0b1;
+  let Unpredictable{5} = 0b1;
+
+  let Defs = [P0];
+}

llvm/trunk/lib/Target/ARM/ARMRegisterBankInfo.cpp

ARMRegisterBankInfo::ARMRegisterBankInfo(const TargetRegisterInfo &TRI)
   assert(RBGPR.covers(*TRI.getRegClass(ARM::rGPRRegClassID)) &&
          "Subclass not added?");
   assert(RBGPR.covers(*TRI.getRegClass(ARM::tGPRRegClassID)) &&
          "Subclass not added?");
   assert(RBGPR.covers(*TRI.getRegClass(ARM::tcGPRRegClassID)) &&
          "Subclass not added?");
   assert(RBGPR.covers(*TRI.getRegClass(ARM::tGPR_and_tcGPRRegClassID)) &&
          "Subclass not added?");
+  assert(RBGPR.covers(*TRI.getRegClass(ARM::tGPREven_and_tGPR_and_tcGPRRegClassID)) &&
+         "Subclass not added?");
+  assert(RBGPR.covers(*TRI.getRegClass(ARM::tGPROdd_and_tcGPRRegClassID)) &&
+         "Subclass not added?");
   assert(RBGPR.getSize() == 32 && "GPRs should hold up to 32-bit");
 
 #ifndef NDEBUG
   ARM::checkPartialMappings();
   ARM::checkValueMappings();
 #endif
 }
 

llvm/trunk/lib/Target/ARM/ARMRegisterInfo.td

def DPR_8 : RegisterClass<"ARM", [f64, v8i8, v4i16, v2i32, v1i64, v2f32, v4f16], 64,
                           (trunc DPR, 8)> {
   let DiagnosticString = "operand must be a register in range [d0, d7]";
 }
 
 // Generic 128-bit vector register class.
 def QPR : RegisterClass<"ARM", [v16i8, v8i16, v4i32, v2i64, v4f32, v2f64, v8f16], 128,
                         (sequence "Q%u", 0, 15)> {
   // Allocate non-VFP2 aliases Q8-Q15 first.
-  let AltOrders = [(rotl QPR, 8)];
-  let AltOrderSelect = [{ return 1; }];
+  let AltOrders = [(rotl QPR, 8), (trunc QPR, 8)];
+  let AltOrderSelect = [{
+    return 1 + MF.getSubtarget<ARMSubtarget>().hasMVEIntegerOps();
+  }];
   let DiagnosticString = "operand must be a register in range [q0, q15]";
 }
 
 // Subset of QPR that have 32-bit SPR subregs.
 def QPR_VFP2 : RegisterClass<"ARM", [v16i8, v8i16, v4i32, v2i64, v4f32, v2f64],
                              128, (trunc QPR, 8)> {
   let DiagnosticString = "operand must be a register in range [q0, q7]";
 }
 
 // Subset of QPR that have DPR_8 and SPR_8 subregs.
 def QPR_8 : RegisterClass<"ARM", [v16i8, v8i16, v4i32, v2i64, v4f32, v2f64],
                            128, (trunc QPR, 4)> {
   let DiagnosticString = "operand must be a register in range [q0, q3]";
 }
 
+// MVE 128-bit vector register class. This class is only really needed for
+// parsing assembly, since we still have to truncate the register set in the QPR
+// class anyway.
+def MQPR : RegisterClass<"ARM", [v16i8, v8i16, v4i32, v2i64, v4f32, v2f64, v8f16],
+                         128, (trunc QPR, 8)>;
+
 // Pseudo-registers representing odd-even pairs of D registers. The even-odd
 // pairs are already represented by the Q registers.
 // These are needed by NEON instructions requiring two consecutive D registers.
 // There is no D31_D0 register as that is always an UNPREDICTABLE encoding.
 def TuplesOE2D : RegisterTuples<[dsub_0, dsub_1],
                                 [(decimate (shl DPR, 1), 2),
                                  (decimate (shl DPR, 2), 2)]>;
 

llvm/trunk/lib/Target/ARM/AsmParser/ARMAsmParser.cpp

 #include <string>
 #include <utility>
 #include <vector>
 
 #define DEBUG_TYPE "asm-parser"
 
 using namespace llvm;
 
+namespace llvm {
+extern const MCInstrDesc ARMInsts[];
+} // end namespace llvm
+
 namespace {
 
 enum class ImplicitItModeTy { Always, Never, ARMOnly, ThumbOnly };
 
 static cl::opt<ImplicitItModeTy> ImplicitItMode(
     "arm-implicit-it", cl::init(ImplicitItModeTy::ARMOnly),
     cl::desc("Allow conditional instructions outdside of an IT block"),
     cl::values(clEnumValN(ImplicitItModeTy::Always, "always",
class ARMAsmParser : public MCTargetAsmParser {
   void startExplicitITBlock(ARMCC::CondCodes Cond, unsigned Mask) {
     assert(!inITBlock());
     ITState.Cond = Cond;
     ITState.Mask = Mask;
     ITState.CurPosition = 0;
     ITState.IsExplicit = true;
   }
 
+  struct {
+    unsigned Mask : 4;
+    unsigned CurPosition;
+  } VPTState;
+  bool inVPTBlock() { return VPTState.CurPosition != ~0U; }
+  void forwardVPTPosition() {
+    if (!inVPTBlock()) return;
+    unsigned TZ = countTrailingZeros(VPTState.Mask);
+    if (++VPTState.CurPosition == 5 - TZ)
+      VPTState.CurPosition = ~0U;
+  }
+
   void Note(SMLoc L, const Twine &Msg, SMRange Range = None) {
     return getParser().Note(L, Msg, Range);
   }
 
   bool Warning(SMLoc L, const Twine &Msg, SMRange Range = None) {
     return getParser().Warning(L, Msg, Range);
   }
 
class ARMAsmParser : public MCTargetAsmParser {
   bool parseDirectiveUnwindRaw(SMLoc L);
   bool parseDirectiveTLSDescSeq(SMLoc L);
   bool parseDirectiveMovSP(SMLoc L);
   bool parseDirectiveObjectArch(SMLoc L);
   bool parseDirectiveArchExtension(SMLoc L);
   bool parseDirectiveAlign(SMLoc L);
   bool parseDirectiveThumbSet(SMLoc L);
 
-  StringRef splitMnemonic(StringRef Mnemonic, unsigned &PredicationCode,
-                          bool &CarrySetting, unsigned &ProcessorIMod,
-                          StringRef &ITMask);
-  void getMnemonicAcceptInfo(StringRef Mnemonic, StringRef FullInst,
-                             bool &CanAcceptCarrySet,
-                             bool &CanAcceptPredicationCode);
+  bool isMnemonicVPTPredicable(StringRef Mnemonic, StringRef ExtraToken);
+  StringRef splitMnemonic(StringRef Mnemonic, StringRef ExtraToken,
+                          unsigned &PredicationCode,
+                          unsigned &VPTPredicationCode, bool &CarrySetting,
+                          unsigned &ProcessorIMod, StringRef &ITMask);
+  void getMnemonicAcceptInfo(StringRef Mnemonic, StringRef ExtraToken,
+                             StringRef FullInst, bool &CanAcceptCarrySet,
+                             bool &CanAcceptPredicationCode,
+                             bool &CanAcceptVPTPredicationCode);
 
   void tryConvertingToTwoOperandForm(StringRef Mnemonic, bool CarrySetting,
                                      OperandVector &Operands);
   bool isThumb() const {
     // FIXME: Can tablegen auto-generate this?
     return getSTI().getFeatureBits()[ARM::ModeThumb];
   }
 
class ARMAsmParser : public MCTargetAsmParser {
   }
 
   bool hasV8MMainline() const {
     return getSTI().getFeatureBits()[ARM::HasV8MMainlineOps];
   }
   bool hasV8_1MMainline() const {
     return getSTI().getFeatureBits()[ARM::HasV8_1MMainlineOps];
   }
+  bool hasMVE() const {
+    return getSTI().getFeatureBits()[ARM::HasMVEIntegerOps];
+  }
+  bool hasMVEFloat() const {
+    return getSTI().getFeatureBits()[ARM::HasMVEFloatOps];
+  }
   bool has8MSecExt() const {
     return getSTI().getFeatureBits()[ARM::Feature8MSecExt];
   }
 
   bool hasARM() const {
     return !getSTI().getFeatureBits()[ARM::FeatureNoARM];
   }
 
#include "ARMGenAsmMatcher.inc"
   // Asm Match Converter Methods
   void cvtThumbMultiply(MCInst &Inst, const OperandVector &);
   void cvtThumbBranches(MCInst &Inst, const OperandVector &);
 
   bool validateInstruction(MCInst &Inst, const OperandVector &Ops);
   bool processInstruction(MCInst &Inst, const OperandVector &Ops, MCStreamer &Out);
   bool shouldOmitCCOutOperand(StringRef Mnemonic, OperandVector &Operands);
   bool shouldOmitPredicateOperand(StringRef Mnemonic, OperandVector &Operands);
+  bool shouldOmitVectorPredicateOperand(StringRef Mnemonic, OperandVector &Operands);
   bool isITBlockTerminator(MCInst &Inst) const;
   void fixupGNULDRDAlias(StringRef Mnemonic, OperandVector &Operands);
   bool validateLDRDSTRD(MCInst &Inst, const OperandVector &Operands,
                         bool Load, bool ARMMode, bool Writeback);
 
 public:
   enum ARMMatchResultTy {
     Match_RequiresITBlock = FIRST_TARGET_MATCH_RESULT_TY,
ARMAsmParser(const MCSubtargetInfo &STI, MCAsmParser &Parser,
 
     // Add build attributes based on the selected target.
     if (AddBuildAttributes)
       getTargetStreamer().emitTargetAttributes(STI);
 
     // Not in an ITBlock to start with.
     ITState.CurPosition = ~0U;
 
+    VPTState.CurPosition = ~0U;
+
     NextSymbolIsThumb = false;
   }
 
   // Implementation of the MCTargetAsmParser interface:
   bool ParseRegister(unsigned &RegNo, SMLoc &StartLoc, SMLoc &EndLoc) override;
   bool ParseInstruction(ParseInstructionInfo &Info, StringRef Name,
                         SMLoc NameLoc, OperandVector &Operands) override;
   bool ParseDirective(AsmToken DirectiveID) override;
#include "ARMGenAsmMatcher.inc"
   void onLabelParsed(MCSymbol *Symbol) override;
 };
 
 /// ARMOperand - Instances of this class represent a parsed ARM machine
 /// operand.
 class ARMOperand : public MCParsedAsmOperand {
   enum KindTy {
     k_CondCode,
+    k_VPTPred,
     k_CCOut,
     k_ITCondMask,
     k_CoprocNum,
     k_CoprocReg,
     k_CoprocOption,
     k_Immediate,
     k_MemBarrierOpt,
     k_InstSyncBarrierOpt,
class ARMOperand : public MCParsedAsmOperand {
 
   SMLoc StartLoc, EndLoc, AlignmentLoc;
   SmallVector<unsigned, 8> Registers;
 
   struct CCOp {
     ARMCC::CondCodes Val;
   };
 
+  struct VCCOp {
+    ARMVCC::VPTCodes Val;
+  };
+
   struct CopOp {
     unsigned Val;
   };
 
   struct CoprocOptionOp {
     unsigned Val;
   };
 
class ARMOperand : public MCParsedAsmOperand {
 
   struct BitfieldOp {
     unsigned LSB;
     unsigned Width;
   };
 
   union {
     struct CCOp CC;
+    struct VCCOp VCC;
     struct CopOp Cop;
     struct CoprocOptionOp CoprocOption;
     struct MBOptOp MBOpt;
     struct ISBOptOp ISBOpt;
     struct TSBOptOp TSBOpt;
     struct ITMaskOp ITMask;
     struct IFlagsOp IFlags;
     struct MMaskOp MMask;
SMLoc getAlignmentLoc() const {
     return AlignmentLoc;
   }
 
   ARMCC::CondCodes getCondCode() const {
     assert(Kind == k_CondCode && "Invalid access!");
     return CC.Val;
   }
 
+  ARMVCC::VPTCodes getVPTPred() const {
+    assert(isVPTPred() && "Invalid access!");
+    return VCC.Val;
+  }
+
   unsigned getCoproc() const {
     assert((Kind == k_CoprocNum || Kind == k_CoprocReg) && "Invalid access!");
     return Cop.Val;
   }
 
   StringRef getToken() const {
     assert(Kind == k_Token && "Invalid access!");
     return StringRef(Tok.Data, Tok.Length);
unsigned getBankedReg() const {
     assert(Kind == k_BankedReg && "Invalid access!");
     return BankedReg.Val;
   }
 
   bool isCoprocNum() const { return Kind == k_CoprocNum; }
   bool isCoprocReg() const { return Kind == k_CoprocReg; }
   bool isCoprocOption() const { return Kind == k_CoprocOption; }
   bool isCondCode() const { return Kind == k_CondCode; }
+  bool isVPTPred() const { return Kind == k_VPTPred; }
   bool isCCOut() const { return Kind == k_CCOut; }
   bool isITMask() const { return Kind == k_ITCondMask; }
   bool isITCondCode() const { return Kind == k_CondCode; }
   bool isImm() const override {
     return Kind == k_Immediate;
   }
 
   bool isARMBranchTarget() const {
bool isVecListFourQHWordIndexed() const {
     return VectorList.Count == 4 && VectorList.LaneIndex <= 3;
   }
 
   bool isVecListFourDWordIndexed() const {
     if (!isSingleSpacedVectorIndexed()) return false;
     return VectorList.Count == 4 && VectorList.LaneIndex <= 1;
   }
 
+  bool isVectorIndex() const { return Kind == k_VectorIndex; }
+
   bool isVectorIndex8() const {
     if (Kind != k_VectorIndex) return false;
     return VectorIndex.Val < 8;
   }
 
   bool isVectorIndex16() const {
     if (Kind != k_VectorIndex) return false;
     return VectorIndex.Val < 4;
bool isMVELongShift() const {
     // Must be a constant.
     if (!CE) return false;
     uint64_t Value = CE->getValue();
     return Value >= 1 && Value <= 32;
   }
 
   bool isITCondCodeNoAL() const {
     if (!isITCondCode()) return false;
-    auto CC = (ARMCC::CondCodes) getCondCode();
+    ARMCC::CondCodes CC = getCondCode();
     return CC != ARMCC::AL;
   }
 
+  bool isITCondCodeRestrictedI() const {
+    if (!isITCondCode())
+      return false;
+    ARMCC::CondCodes CC = getCondCode();
+    return CC == ARMCC::EQ || CC == ARMCC::NE;
+  }
+
+  bool isITCondCodeRestrictedS() const {
+    if (!isITCondCode())
+      return false;
+    ARMCC::CondCodes CC = getCondCode();
+    return CC == ARMCC::LT || CC == ARMCC::GT || CC == ARMCC::LE ||
+           CC == ARMCC::GE;
+  }
+
+  bool isITCondCodeRestrictedU() const {
+    if (!isITCondCode())
+      return false;
+    ARMCC::CondCodes CC = getCondCode();
+    return CC == ARMCC::HS || CC == ARMCC::HI;
+  }
+
+  bool isITCondCodeRestrictedFP() const {
+    if (!isITCondCode())
+      return false;
+    ARMCC::CondCodes CC = getCondCode();
+    return CC == ARMCC::EQ || CC == ARMCC::NE || CC == ARMCC::LT ||
+           CC == ARMCC::GT || CC == ARMCC::LE || CC == ARMCC::GE;
+  }
+
   void addExpr(MCInst &Inst, const MCExpr *Expr) const {
     // Add as immediates when possible.  Null MCExpr = 0.
     if (!Expr)
       Inst.addOperand(MCOperand::createImm(0));
     else if (const MCConstantExpr *CE = dyn_cast<MCConstantExpr>(Expr))
       Inst.addOperand(MCOperand::createImm(CE->getValue()));
     else
       Inst.addOperand(MCOperand::createExpr(Expr));

 
   void addCondCodeOperands(MCInst &Inst, unsigned N) const {
     assert(N == 2 && "Invalid number of operands!");
     Inst.addOperand(MCOperand::createImm(unsigned(getCondCode())));
     unsigned RegNum = getCondCode() == ARMCC::AL ? 0: ARM::CPSR;
     Inst.addOperand(MCOperand::createReg(RegNum));
   }
 
+  void addVPTPredNOperands(MCInst &Inst, unsigned N) const {
+    assert(N == 2 && "Invalid number of operands!");
+    Inst.addOperand(MCOperand::createImm(unsigned(getVPTPred())));
+    unsigned RegNum = getVPTPred() == ARMVCC::None ? 0: ARM::P0;
+    Inst.addOperand(MCOperand::createReg(RegNum));
+  }
+
+  void addVPTPredROperands(MCInst &Inst, unsigned N) const {
+    assert(N == 3 && "Invalid number of operands!");
+    addVPTPredNOperands(Inst, N-1);
+    unsigned RegNum;
+    if (getVPTPred() == ARMVCC::None) {
+      RegNum = 0;
+    } else {
+      unsigned NextOpIndex = Inst.getNumOperands();
+      const MCInstrDesc &MCID = ARMInsts[Inst.getOpcode()];
+      int TiedOp = MCID.getOperandConstraint(NextOpIndex, MCOI::TIED_TO);
+      assert(TiedOp >= 0 &&
+             "Inactive register in vpred_r is not tied to an output!");
+      RegNum = Inst.getOperand(TiedOp).getReg();
+    }
+    Inst.addOperand(MCOperand::createReg(RegNum));
+  }
+
   void addCoprocNumOperands(MCInst &Inst, unsigned N) const {
     assert(N == 1 && "Invalid number of operands!");
     Inst.addOperand(MCOperand::createImm(getCoproc()));
   }
 
   void addCoprocRegOperands(MCInst &Inst, unsigned N) const {
     assert(N == 1 && "Invalid number of operands!");
     Inst.addOperand(MCOperand::createImm(getCoproc()));
void addPostIdxRegShiftedOperands(MCInst &Inst, unsigned N) const {
     // The sign, shift type, and shift amount are encoded in a single operand
     // using the AM2 encoding helpers.
     ARM_AM::AddrOpc opc = PostIdxReg.isAdd ? ARM_AM::add : ARM_AM::sub;
     unsigned Imm = ARM_AM::getAM2Opc(opc, PostIdxReg.ShiftImm,
                                      PostIdxReg.ShiftTy);
     Inst.addOperand(MCOperand::createImm(Imm));
   }
 
+  void addPowerTwoOperands(MCInst &Inst, unsigned N) const {
+    assert(N == 1 && "Invalid number of operands!");
+    const MCConstantExpr *CE = dyn_cast<MCConstantExpr>(getImm());
+    Inst.addOperand(MCOperand::createImm(CE->getValue()));
+  }
+
   void addMSRMaskOperands(MCInst &Inst, unsigned N) const {
     assert(N == 1 && "Invalid number of operands!");
     Inst.addOperand(MCOperand::createImm(unsigned(getMSRMask())));
   }
 
   void addBankedRegOperands(MCInst &Inst, unsigned N) const {
     assert(N == 1 && "Invalid number of operands!");
     Inst.addOperand(MCOperand::createImm(unsigned(getBankedReg())));
static std::unique_ptr<ARMOperand> CreateCondCode(ARMCC::CondCodes CC,
                                                     SMLoc S) {
     auto Op = make_unique<ARMOperand>(k_CondCode);
     Op->CC.Val = CC;
     Op->StartLoc = S;
     Op->EndLoc = S;
     return Op;
   }
 
+  static std::unique_ptr<ARMOperand> CreateVPTPred(ARMVCC::VPTCodes CC,
+                                                   SMLoc S) {
+    auto Op = make_unique<ARMOperand>(k_VPTPred);
+    Op->VCC.Val = CC;
+    Op->StartLoc = S;
+    Op->EndLoc = S;
+    return Op;
+  }
+
   static std::unique_ptr<ARMOperand> CreateCoprocNum(unsigned CopVal, SMLoc S) {
     auto Op = make_unique<ARMOperand>(k_CoprocNum);
     Op->Cop.Val = CopVal;
     Op->StartLoc = S;
     Op->EndLoc = S;
     return Op;
   }
 
auto RegName = [](unsigned Reg) {
     else
       return "noreg";
   };
 
   switch (Kind) {
   case k_CondCode:
     OS << "<ARMCC::" << ARMCondCodeToString(getCondCode()) << ">";
     break;
+  case k_VPTPred:
+    OS << "<ARMVCC::" << ARMVPTPredToString(getVPTPred()) << ">";
+    break;
   case k_CCOut:
     OS << "<ccout " << RegName(getReg()) << ">";
     break;
   case k_ITCondMask: {
     static const char *const MaskStr[] = {
       "(invalid)", "(tttt)", "(ttt)", "(ttte)",
       "(tt)",      "(ttet)", "(tte)", "(ttee)",
       "(t)",       "(tett)", "(tet)", "(tete)",

 }
 
 /// Given a mnemonic, split out possible predication code and carry
 /// setting letters to form a canonical mnemonic and flags.
 //
 // FIXME: Would be nice to autogen this.
 // FIXME: This is a bit of a maze of special cases.
 StringRef ARMAsmParser::splitMnemonic(StringRef Mnemonic,
+                                      StringRef ExtraToken,
                                       unsigned &PredicationCode,
+                                      unsigned &VPTPredicationCode,
                                       bool &CarrySetting,
                                       unsigned &ProcessorIMod,
                                       StringRef &ITMask) {
   PredicationCode = ARMCC::AL;
+  VPTPredicationCode = ARMVCC::None;
   CarrySetting = false;
   ProcessorIMod = 0;
 
   // Ignore some mnemonics we know aren't predicated forms.
   //
   // FIXME: Would be nice to autogen this.
   if ((Mnemonic == "movs" && isThumb()) ||
       Mnemonic == "teq"   || Mnemonic == "vceq"   || Mnemonic == "svc"   ||
unsigned IMod =
       .Case("id", ARM_PROC::ID)
       .Default(~0U);
     if (IMod != ~0U) {
       Mnemonic = Mnemonic.slice(0, Mnemonic.size()-2);
       ProcessorIMod = IMod;
     }
   }
 
+  if (isMnemonicVPTPredicable(Mnemonic, ExtraToken)) {
+    unsigned CC = ARMVectorCondCodeFromString(Mnemonic.substr(Mnemonic.size()-1));
+    if (CC != ~0U) {
+      Mnemonic = Mnemonic.slice(0, Mnemonic.size()-1);
+      VPTPredicationCode = CC;
+    }
+    return Mnemonic;
+  }
+
   // The "it" instruction has the condition mask on the end of the mnemonic.
   if (Mnemonic.startswith("it")) {
     ITMask = Mnemonic.slice(2, Mnemonic.size());
     Mnemonic = Mnemonic.slice(0, 2);
   }
 
+  if (Mnemonic.startswith("vpst")) {
+    ITMask = Mnemonic.slice(4, Mnemonic.size());
+    Mnemonic = Mnemonic.slice(0, 4);
+  }
+  else if (Mnemonic.startswith("vpt")) {
+    ITMask = Mnemonic.slice(3, Mnemonic.size());
+    Mnemonic = Mnemonic.slice(0, 3);
+  }
+
   return Mnemonic;
 }
 
 /// Given a canonical mnemonic, determine if the instruction ever allows
 /// inclusion of carry set or predication code operands.
 //
 // FIXME: It would be nice to autogen this.
-void ARMAsmParser::getMnemonicAcceptInfo(StringRef Mnemonic, StringRef FullInst,
+void ARMAsmParser::getMnemonicAcceptInfo(StringRef Mnemonic,
+                                         StringRef ExtraToken,
+                                         StringRef FullInst,
                                          bool &CanAcceptCarrySet,
-                                         bool &CanAcceptPredicationCode) {
+                                         bool &CanAcceptPredicationCode,
+                                         bool &CanAcceptVPTPredicationCode) {
+  CanAcceptVPTPredicationCode = isMnemonicVPTPredicable(Mnemonic, ExtraToken);
+
   CanAcceptCarrySet =
       Mnemonic == "and" || Mnemonic == "lsl" || Mnemonic == "lsr" ||
       Mnemonic == "rrx" || Mnemonic == "ror" || Mnemonic == "sub" ||
       Mnemonic == "add" || Mnemonic == "adc" || Mnemonic == "mul" ||
       Mnemonic == "bic" || Mnemonic == "asr" || Mnemonic == "orr" ||
       Mnemonic == "mvn" || Mnemonic == "rsb" || Mnemonic == "rsc" ||
       Mnemonic == "orn" || Mnemonic == "sbc" || Mnemonic == "eor" ||
       Mnemonic == "neg" || Mnemonic == "vfm" || Mnemonic == "vfnm" ||
if (Mnemonic == "bkpt" || Mnemonic == "cbnz" || Mnemonic == "setend" ||
       Mnemonic == "vcmla" || Mnemonic == "vcadd" ||
       Mnemonic == "vfmal" || Mnemonic == "vfmsl" ||
       Mnemonic == "sb"    || Mnemonic == "ssbb"  ||
       Mnemonic == "pssbb" ||
       Mnemonic == "bfcsel" || Mnemonic == "wls" ||
       Mnemonic == "dls" || Mnemonic == "le" || Mnemonic == "csel" ||
       Mnemonic == "csinc" || Mnemonic == "csinv" || Mnemonic == "csneg" ||
       Mnemonic == "cinc" || Mnemonic == "cinv" || Mnemonic == "cneg" ||
-      Mnemonic == "cset" || Mnemonic == "csetm") {
+      Mnemonic == "cset" || Mnemonic == "csetm" ||
+      Mnemonic.startswith("vpt") || Mnemonic.startswith("vpst")) {
     // These mnemonics are never predicable
     CanAcceptPredicationCode = false;
   } else if (!isThumb()) {
     // Some instructions are only predicable in Thumb mode
     CanAcceptPredicationCode =
         Mnemonic != "cdp2" && Mnemonic != "clrex" && Mnemonic != "mcr2" &&
         Mnemonic != "mcrr2" && Mnemonic != "mrc2" && Mnemonic != "mrrc2" &&
         Mnemonic != "dmb" && Mnemonic != "dfb" && Mnemonic != "dsb" &&
if (static_cast<ARMOperand &>(*Operands[RegIdx]).isReg() &&
              static_cast<ARMOperand &>(*Operands[RegIdx]).getReg()) ||
          ARMMCRegisterClasses[ARM::QPRRegClassID].contains(
              static_cast<ARMOperand &>(*Operands[RegIdx]).getReg())))
       return true;
   }
   return false;
 }
 
+bool ARMAsmParser::shouldOmitVectorPredicateOperand(StringRef Mnemonic,
+                                                    OperandVector &Operands) {
+  if (!hasMVE() || Operands.size() < 3)
+    return true;
+
+  for (auto &Operand : Operands) {
+    // We check the larger class QPR instead of just the legal class
+    // MQPR, to more accurately report errors when using Q registers
+    // outside of the allowed range.
+    if (static_cast<ARMOperand &>(*Operand).isVectorIndex() ||
+        (Operand->isReg() &&
+         (ARMMCRegisterClasses[ARM::QPRRegClassID].contains(
+             Operand->getReg()))))
+      return false;
+  }
+  return true;
+}
+
 static bool isDataTypeToken(StringRef Tok) {
   return Tok == ".8" || Tok == ".16" || Tok == ".32" || Tok == ".64" ||
     Tok == ".i8" || Tok == ".i16" || Tok == ".i32" || Tok == ".i64" ||
     Tok == ".u8" || Tok == ".u16" || Tok == ".u32" || Tok == ".u64" ||
     Tok == ".s8" || Tok == ".s16" || Tok == ".s32" || Tok == ".s64" ||
     Tok == ".p8" || Tok == ".p16" || Tok == ".f32" || Tok == ".f64" ||
     Tok == ".f" || Tok == ".d";
 }
if (Parser.getTok().is(AsmToken::Identifier) &&
     // We always return 'error' for this, as we're done with this
     // statement and don't need to match the 'instruction."
     return true;
   }
 
   // Create the leading tokens for the mnemonic, split by '.' characters.
   size_t Start = 0, Next = Name.find('.');
   StringRef Mnemonic = Name.slice(Start, Next);
+  StringRef ExtraToken = Name.slice(Next, Name.find(' ', Next + 1));
 
   // Split out the predication code and carry setting flag from the mnemonic.
   unsigned PredicationCode;
+  unsigned VPTPredicationCode;
   unsigned ProcessorIMod;
   bool CarrySetting;
   StringRef ITMask;
-  Mnemonic = splitMnemonic(Mnemonic, PredicationCode, CarrySetting,
-                           ProcessorIMod, ITMask);
+  Mnemonic = splitMnemonic(Mnemonic, ExtraToken, PredicationCode, VPTPredicationCode,
+                           CarrySetting, ProcessorIMod, ITMask);
 
   // In Thumb1, only the branch (B) instruction can be predicated.
   if (isThumbOne() && PredicationCode != ARMCC::AL && Mnemonic != "b") {
     return Error(NameLoc, "conditional execution not supported in Thumb1");
   }
 
   Operands.push_back(ARMOperand::CreateToken(Mnemonic, NameLoc));
 
-  // Handle the mask for IT instructions. In ARMOperand and MCOperand,
-  // this is stored in a format independent of the condition code: the
-  // lowest set bit indicates the end of the encoding, and above that,
-  // a 1 bit indicates 'else', and an 0 indicates 'then'. E.g.
+  // Handle the mask for IT and VPT instructions. In ARMOperand and
+  // MCOperand, this is stored in a format independent of the
+  // condition code: the lowest set bit indicates the end of the
+  // encoding, and above that, a 1 bit indicates 'else', and an 0
+  // indicates 'then'. E.g.
   //    IT    -> 1000
   //    ITx   -> x100    (ITT -> 0100, ITE -> 1100)
   //    ITxy  -> xy10    (e.g. ITET -> 1010)
   //    ITxyz -> xyz1    (e.g. ITEET -> 1101)
-  if (Mnemonic == "it") {
-    SMLoc Loc = SMLoc::getFromPointer(NameLoc.getPointer() + 2);
+  if (Mnemonic == "it" || Mnemonic.startswith("vpt") ||
+      Mnemonic.startswith("vpst")) {
+    SMLoc Loc = Mnemonic == "it"  ? SMLoc::getFromPointer(NameLoc.getPointer() + 2) :
+                Mnemonic == "vpt" ? SMLoc::getFromPointer(NameLoc.getPointer() + 3) :
+                                    SMLoc::getFromPointer(NameLoc.getPointer() + 4);
     if (ITMask.size() > 3) {
+      if (Mnemonic == "it")
         return Error(Loc, "too many conditions on IT instruction");
+      return Error(Loc, "too many conditions on VPT instruction");
     }
     unsigned Mask = 8;
     for (unsigned i = ITMask.size(); i != 0; --i) {
       char pos = ITMask[i - 1];
       if (pos != 't' && pos != 'e') {
         return Error(Loc, "illegal IT block condition mask '" + ITMask + "'");
       }
       Mask >>= 1;
       if (ITMask[i - 1] == 'e')
         Mask |= 8;
     }
     Operands.push_back(ARMOperand::CreateITMask(Mask, Loc));
   }
 
   // FIXME: This is all a pretty gross hack. We should automatically handle
   // optional operands like this via tblgen.
 
   // Next, add the CCOut and ConditionCode operands, if needed.
   //
   // For mnemonics which can ever incorporate a carry setting bit or predication
   // code, our matching model involves us always generating CCOut and
   // ConditionCode operands to match the mnemonic "as written" and then we let
   // the matcher deal with finding the right instruction or generating an
   // appropriate error.
-  bool CanAcceptCarrySet, CanAcceptPredicationCode;
-  getMnemonicAcceptInfo(Mnemonic, Name, CanAcceptCarrySet, CanAcceptPredicationCode);
+  bool CanAcceptCarrySet, CanAcceptPredicationCode, CanAcceptVPTPredicationCode;
+  getMnemonicAcceptInfo(Mnemonic, ExtraToken, Name, CanAcceptCarrySet,
+                        CanAcceptPredicationCode, CanAcceptVPTPredicationCode);
 
   // If we had a carry-set on an instruction that can't do that, issue an
   // error.
   if (!CanAcceptCarrySet && CarrySetting) {
     return Error(NameLoc, "instruction '" + Mnemonic +
                  "' can not set flags, but 's' suffix specified");
   }
   // If we had a predication code on an instruction that can't do that, issue an
   // error.
   if (!CanAcceptPredicationCode && PredicationCode != ARMCC::AL) {
     return Error(NameLoc, "instruction '" + Mnemonic +
                  "' is not predicable, but condition code specified");
   }
 
+  // If we had a VPT predication code on an instruction that can't do that, issue an
+  // error.
+  if (!CanAcceptVPTPredicationCode && VPTPredicationCode != ARMVCC::None) {
+    return Error(NameLoc, "instruction '" + Mnemonic +
+                 "' is not VPT predicable, but VPT code T/E is specified");
+  }
+
   // Add the carry setting operand, if necessary.
   if (CanAcceptCarrySet) {
     SMLoc Loc = SMLoc::getFromPointer(NameLoc.getPointer() + Mnemonic.size());
     Operands.push_back(ARMOperand::CreateCCOut(CarrySetting ? ARM::CPSR : 0,
                                                Loc));
   }
 
   // Add the predication code operand, if necessary.
   if (CanAcceptPredicationCode) {
     SMLoc Loc = SMLoc::getFromPointer(NameLoc.getPointer() + Mnemonic.size() +
                                       CarrySetting);
     Operands.push_back(ARMOperand::CreateCondCode(
-                         ARMCC::CondCodes(PredicationCode), Loc));
   }
 
+  // Add the VPT predication code operand, if necessary.
+  if (CanAcceptVPTPredicationCode) {
+    SMLoc Loc = SMLoc::getFromPointer(NameLoc.getPointer() + Mnemonic.size() +
+                                      CarrySetting);
+    Operands.push_back(ARMOperand::CreateVPTPred(
+                         ARMVCC::VPTCodes(VPTPredicationCode), Loc));
+  }
+
   // Add the processor imod operand, if necessary.
   if (ProcessorIMod) {
     Operands.push_back(ARMOperand::CreateImm(
           MCConstantExpr::create(ProcessorIMod, getContext()),
                                  NameLoc, NameLoc));
   } else if (Mnemonic == "cps" && isMClass()) {
     return Error(NameLoc, "instruction 'cps' requires effect for M-class");
   }
 
   // Add the remaining tokens in the mnemonic.
   while (Next != StringRef::npos) {
     Start = Next;
     Next = Name.find('.', Start + 1);
-    StringRef ExtraToken = Name.slice(Start, Next);
+    ExtraToken = Name.slice(Start, Next);
 
     // Some NEON instructions have an optional datatype suffix that is
     // completely ignored. Check for that.
     if (isDataTypeToken(ExtraToken) &&
         doesIgnoreDataTypeSuffix(Mnemonic, ExtraToken))
       continue;
 
     // For for ARM mode generate an error if the .n qualifier is used.
bool ARMAsmParser::ParseInstruction(ParseInstructionInfo &Info, StringRef Name,
   // try to remove a cc_out operand that was explicitly set on the
   // mnemonic, of course (CarrySetting == true). Reason number #317 the
   // table driven matcher doesn't fit well with the ARM instruction set.
   if (!CarrySetting && shouldOmitCCOutOperand(Mnemonic, Operands))
     Operands.erase(Operands.begin() + 1);
 
   // Some instructions have the same mnemonic, but don't always
   // have a predicate. Distinguish them here and delete the
-  // predicate if needed.
+  // appropriate predicate if needed.  This could be either the scalar
+  // predication code or the vector predication code.
   if (PredicationCode == ARMCC::AL &&
       shouldOmitPredicateOperand(Mnemonic, Operands))
     Operands.erase(Operands.begin() + 1);
 
+
+  if (hasMVE()) {
+    if (CanAcceptVPTPredicationCode) {
+      // For all other instructions, make sure only one of the two
+      // predication operands is left behind, depending on whether we should
+      // use the vector predication.
+      if (shouldOmitVectorPredicateOperand(Mnemonic, Operands)) {
+        if (CanAcceptPredicationCode)
+          Operands.erase(Operands.begin() + 2);
+        else
+          Operands.erase(Operands.begin() + 1);
+      } else if (CanAcceptPredicationCode && PredicationCode == ARMCC::AL) {
+        Operands.erase(Operands.begin() + 1);
+      }
+    }
+  }
+
+  if (VPTPredicationCode != ARMVCC::None) {
+    bool usedVPTPredicationCode = false;
+    for (unsigned I = 1; I < Operands.size(); ++I)
+      if (static_cast<ARMOperand &>(*Operands[I]).isVPTPred())
+        usedVPTPredicationCode = true;
+    if (!usedVPTPredicationCode) {
+      // If we have a VPT predication code and we haven't just turned it
+      // into an operand, then it was a mistake for splitMnemonic to
+      // separate it from the rest of the mnemonic in the first place,
+      // and this may lead to wrong disassembly (e.g. scalar floating
+      // point VCMPE is actually a different instruction from VCMP, so
+      // we mustn't treat them the same). In that situation, glue it
+      // back on.
+      Mnemonic = Name.slice(0, Mnemonic.size() + 1);
+      Operands.erase(Operands.begin());
+      Operands.insert(Operands.begin(),
+                      ARMOperand::CreateToken(Mnemonic, NameLoc));
+    }
+  }
+
     // ARM mode 'blx' need special handling, as the register operand version
     // is predicable, but the label operand version is not. So, we can't rely
     // on the Mnemonic based checking to correctly figure out when to put
     // a k_CondCode operand in the list. If we're trying to match the label
     // version, remove the k_CondCode operand here.
     if (!isThumb() && Mnemonic == "blx" && Operands.size() == 3 &&
         static_cast<ARMOperand &>(*Operands[2]).isImm())
       Operands.erase(Operands.begin() + 1);
 
     // Adjust operands of ldrexd/strexd to MCK_GPRPair.
     // ldrexd/strexd require even/odd GPR pair. To enforce this constraint,
     // a single GPRPair reg operand is used in the .td file to replace the two
-  // GPRs. However, when parsing from asm, the two GRPs cannot be automatically
+    // GPRs. However, when parsing from asm, the two GRPs cannot be
+    // automatically
     // expressed as a GPRPair, so we have to manually merge them.
     // FIXME: We would really like to be able to tablegen'erate this.
     if (!isThumb() && Operands.size() > 4 &&
         (Mnemonic == "ldrexd" || Mnemonic == "strexd" || Mnemonic == "ldaexd" ||
          Mnemonic == "stlexd")) {
       bool isLoad = (Mnemonic == "ldrexd" || Mnemonic == "ldaexd");
       unsigned Idx = isLoad ? 2 : 3;
       ARMOperand &Op1 = static_cast<ARMOperand &>(*Operands[Idx]);
       ARMOperand &Op2 = static_cast<ARMOperand &>(*Operands[Idx + 1]);
 
-    const MCRegisterClass& MRC = MRI->getRegClass(ARM::GPRRegClassID);
+      const MCRegisterClass &MRC = MRI->getRegClass(ARM::GPRRegClassID);
       // Adjust only if Op1 and Op2 are GPRs.
       if (Op1.isReg() && Op2.isReg() && MRC.contains(Op1.getReg()) &&
           MRC.contains(Op2.getReg())) {
         unsigned Reg1 = Op1.getReg();
         unsigned Reg2 = Op2.getReg();
         unsigned Rt = MRI->getEncodingValue(Reg1);
         unsigned Rt2 = MRI->getEncodingValue(Reg2);
 
         // Rt2 must be Rt + 1 and Rt must be even.
         if (Rt + 1 != Rt2 || (Rt & 1)) {
           return Error(Op2.getStartLoc(),
                        isLoad ? "destination operands must be sequential"
                               : "source operands must be sequential");
         }
-      unsigned NewReg = MRI->getMatchingSuperReg(Reg1, ARM::gsub_0,
-          &(MRI->getRegClass(ARM::GPRPairRegClassID)));
+        unsigned NewReg = MRI->getMatchingSuperReg(
+            Reg1, ARM::gsub_0, &(MRI->getRegClass(ARM::GPRPairRegClassID)));
         Operands[Idx] =
             ARMOperand::CreateReg(NewReg, Op1.getStartLoc(), Op2.getEndLoc());
         Operands.erase(Operands.begin() + Idx + 1);
       }
   }
 
   // GNU Assembler extension (compatibility).
   fixupGNULDRDAlias(Mnemonic, Operands);
 
   // FIXME: As said above, this is all a pretty gross hack.  This instruction
   // does not fit with other "subs" and tblgen.
   // Adjust operands of B9.3.19 SUBS PC, LR, #imm (Thumb2) system instruction
if (Writeback) {
 
     // FIXME: Diagnose ldrd/strd with writeback and n == 15.
     // (Except the immediate form of ldrd?)
   }
 
   return false;
 }
 
+static int findFirstVectorPredOperandIdx(const MCInstrDesc &MCID) {
+  for (unsigned i = 0; i < MCID.NumOperands; ++i) {
+    if (ARM::isVpred(MCID.OpInfo[i].OperandType))
+      return i;
+  }
+  return -1;
+}
+
+static bool isVectorPredicable(const MCInstrDesc &MCID) {
+  return findFirstVectorPredOperandIdx(MCID) != -1;
+}
 
 // FIXME: We would really like to be able to tablegen'erate this.
 bool ARMAsmParser::validateInstruction(MCInst &Inst,
                                        const OperandVector &Operands) {
   const MCInstrDesc &MCID = MII.get(Inst.getOpcode());
   SMLoc Loc = Operands[0]->getStartLoc();
 
   // Check the IT block state first.
bool ARMAsmParser::validateInstruction(MCInst &Inst,
   }
 
   // PC-setting instructions in an IT block, but not the last instruction of
   // the block, are UNPREDICTABLE.
   if (inExplicitITBlock() && !lastInITBlock() && isITBlockTerminator(Inst)) {
     return Error(Loc, "instruction must be outside of IT block or the last instruction in an IT block");
   }
 
+  if (inVPTBlock() && !instIsBreakpoint(Inst)) {
+    unsigned Bit = extractITMaskBit(VPTState.Mask, VPTState.CurPosition);
+    if (!isVectorPredicable(MCID))
+      return Error(Loc, "instruction in VPT block must be predicable");
+    unsigned Pred = Inst.getOperand(findFirstVectorPredOperandIdx(MCID)).getImm();
+    unsigned VPTPred = Bit ? ARMVCC::Else : ARMVCC::Then;
+    if (Pred != VPTPred) {
+      SMLoc PredLoc;
+      for (unsigned I = 1; I < Operands.size(); ++I)
+        if (static_cast<ARMOperand &>(*Operands[I]).isVPTPred())
+          PredLoc = Operands[I]->getStartLoc();
+      return Error(PredLoc, "incorrect predication in VPT block; got '" +
+                   StringRef(ARMVPTPredToString(ARMVCC::VPTCodes(Pred))) +
+                   "', but expected '" +
+                   ARMVPTPredToString(ARMVCC::VPTCodes(VPTPred)) + "'");
+    }
+  }
+  else if (isVectorPredicable(MCID) &&
+           Inst.getOperand(findFirstVectorPredOperandIdx(MCID)).getImm() !=
+           ARMVCC::None)
+    return Error(Loc, "VPT predicated instructions must be in VPT block");
+
   const unsigned Opcode = Inst.getOpcode();
   switch (Opcode) {
   case ARM::t2IT: {
     // Encoding is unpredictable if it ever results in a notional 'NV'
     // predicate. Since we don't parse 'NV' directly this means an 'AL'
     // predicate with an "else" mask bit.
     unsigned Cond = Inst.getOperand(0).getImm();
     unsigned Mask = Inst.getOperand(1).getImm();
if ((isARMLowRegister(Inst.getOperand(1).getReg()) &&
         TmpInst.addOperand(Inst.getOperand(1));
       }
       TmpInst.addOperand(Inst.getOperand(3));
       TmpInst.addOperand(Inst.getOperand(4));
       Inst = TmpInst;
       return true;
     }
     return false;
+  case ARM::t2VPST:
+  case ARM::t2VPTv16i8:
+  case ARM::t2VPTv8i16:
+  case ARM::t2VPTv4i32:
+  case ARM::t2VPTv16u8:
+  case ARM::t2VPTv8u16:
+  case ARM::t2VPTv4u32:
+  case ARM::t2VPTv16s8:
+  case ARM::t2VPTv8s16:
+  case ARM::t2VPTv4s32:
+  case ARM::t2VPTv4f32:
+  case ARM::t2VPTv8f16:
+  case ARM::t2VPTv16i8r:
+  case ARM::t2VPTv8i16r:
+  case ARM::t2VPTv4i32r:
+  case ARM::t2VPTv16u8r:
+  case ARM::t2VPTv8u16r:
+  case ARM::t2VPTv4u32r:
+  case ARM::t2VPTv16s8r:
+  case ARM::t2VPTv8s16r:
+  case ARM::t2VPTv4s32r:
+  case ARM::t2VPTv4f32r:
+  case ARM::t2VPTv8f16r: {
+    assert(!inVPTBlock() && "Nested VPT blocks are not allowed");
+    MCOperand &MO = Inst.getOperand(0);
+    VPTState.Mask = MO.getImm();
+    VPTState.CurPosition = 0;
+    break;
+  }
   }
   return false;
 }
 
 unsigned ARMAsmParser::checkTargetMatchPredicate(MCInst &Inst) {
   // 16-bit thumb arithmetic instructions either require or preclude the 'S'
   // suffix depending on whether they're in an IT block or not.
   unsigned Opc = Inst.getOpcode();
LLVM_DEBUG(dbgs() << "Parsed as: ";
                dbgs() << "\n");
 
     // Context sensitive operand constraints aren't handled by the matcher,
     // so check them here.
     if (validateInstruction(Inst, Operands)) {
       // Still progress the IT block, otherwise one wrong condition causes
       // nasty cascading errors.
       forwardITPosition();
+      forwardVPTPosition();
       return true;
     }
 
     { // processInstruction() updates inITBlock state, we need to save it away
       bool wasInITBlock = inITBlock();
 
       // Some instructions need post-processing to, for example, tweak which
       // encoding is selected. Loop on it while changes happen so the
if (validateInstruction(Inst, Operands)) {
         Warning(IDLoc, "deprecated instruction in IT block");
       }
     }
 
     // Only move forward at the very end so that everything in validate
     // and process gets a consistent answer about whether we're in an IT
     // block.
     forwardITPosition();
+    forwardVPTPosition();
 
     // ITasm is an ARM mode pseudo-instruction that just sets the ITblock and
     // doesn't actually encode.
     if (Inst.getOpcode() == ARM::ITasm)
       return false;
 
     Inst.setLoc(IDLoc);
     if (PendConditionalInstruction) {
unsigned ARMAsmParser::validateTargetOperandClass(MCParsedAsmOperand &AsmOp,
   case MCK_GPRPair:
     if (Op.isReg() &&
         MRI->getRegClass(ARM::GPRRegClassID).contains(Op.getReg()))
       return Match_Success;
     break;
   }
   return Match_InvalidOperand;
 }
+
+bool ARMAsmParser::isMnemonicVPTPredicable(StringRef Mnemonic,
+                                           StringRef ExtraToken) {
+  if (!hasMVE())
+    return false;
+
+  return Mnemonic.startswith("vabav") || Mnemonic.startswith("vaddv") ||
+         Mnemonic.startswith("vaddlv") || Mnemonic.startswith("vminnmv") ||
+         Mnemonic.startswith("vminnmav") || Mnemonic.startswith("vminv") ||
+         Mnemonic.startswith("vminav") || Mnemonic.startswith("vmaxnmv") ||
+         Mnemonic.startswith("vmaxnmav") || Mnemonic.startswith("vmaxv") ||
+         Mnemonic.startswith("vmaxav") || Mnemonic.startswith("vmladav") ||
+         Mnemonic.startswith("vrmlaldavh") || Mnemonic.startswith("vrmlalvh") ||
+         Mnemonic.startswith("vmlsdav") || Mnemonic.startswith("vmlav") ||
+         Mnemonic.startswith("vmlaldav") || Mnemonic.startswith("vmlalv") ||
+         Mnemonic.startswith("vmaxnm") || Mnemonic.startswith("vminnm") ||
+         Mnemonic.startswith("vmax") || Mnemonic.startswith("vmin") ||
+         Mnemonic.startswith("vshlc") || Mnemonic.startswith("vmovlt") ||
+         Mnemonic.startswith("vmovlb") || Mnemonic.startswith("vshll") ||
+         Mnemonic.startswith("vrshrn") || Mnemonic.startswith("vshrn") ||
+         Mnemonic.startswith("vqrshrun") || Mnemonic.startswith("vqshrun") ||
+         Mnemonic.startswith("vqrshrn") || Mnemonic.startswith("vqshrn") ||
+         Mnemonic.startswith("vbic") || Mnemonic.startswith("vrev64") ||
+         Mnemonic.startswith("vrev32") || Mnemonic.startswith("vrev16") ||
+         Mnemonic.startswith("vmvn") || Mnemonic.startswith("veor") ||
+         Mnemonic.startswith("vorn") || Mnemonic.startswith("vorr") ||
+         Mnemonic.startswith("vand") || Mnemonic.startswith("vmul") ||
+         Mnemonic.startswith("vqrdmulh") || Mnemonic.startswith("vqdmulh") ||
+         Mnemonic.startswith("vsub") || Mnemonic.startswith("vadd") ||
+         Mnemonic.startswith("vqsub") || Mnemonic.startswith("vqadd") ||
+         Mnemonic.startswith("vabd") || Mnemonic.startswith("vrhadd") ||
+         Mnemonic.startswith("vhsub") || Mnemonic.startswith("vhadd") ||
+         Mnemonic.startswith("vdup") || Mnemonic.startswith("vcls") ||
+         Mnemonic.startswith("vclz") || Mnemonic.startswith("vneg") ||
+         Mnemonic.startswith("vabs") || Mnemonic.startswith("vqneg") ||
+         Mnemonic.startswith("vqabs") ||
+         (Mnemonic.startswith("vrint") && Mnemonic != "vrintr") ||
+         Mnemonic.startswith("vcmla") || Mnemonic.startswith("vfma") ||
+         Mnemonic.startswith("vfms") || Mnemonic.startswith("vcadd") ||
+         Mnemonic.startswith("vadd") || Mnemonic.startswith("vsub") ||
+         Mnemonic.startswith("vshl") || Mnemonic.startswith("vqshl") ||
+         Mnemonic.startswith("vqrshl") || Mnemonic.startswith("vrshl") ||
+         Mnemonic.startswith("vsri") || Mnemonic.startswith("vsli") ||
+         Mnemonic.startswith("vrshr") || Mnemonic.startswith("vshr") ||
+         Mnemonic.startswith("vpsel") || Mnemonic.startswith("vcmp") ||
+         Mnemonic.startswith("vqdmladh") || Mnemonic.startswith("vqrdmladh") ||
+         Mnemonic.startswith("vqdmlsdh") || Mnemonic.startswith("vqrdmlsdh") ||
+         Mnemonic.startswith("vcmul") || Mnemonic.startswith("vrmulh") ||
+         Mnemonic.startswith("vqmovn") || Mnemonic.startswith("vqmovun") ||
+         Mnemonic.startswith("vmovnt") || Mnemonic.startswith("vmovnb") ||
+         Mnemonic.startswith("vmaxa") || Mnemonic.startswith("vmaxnma") ||
+         Mnemonic.startswith("vhcadd") || Mnemonic.startswith("vadc") ||
+         Mnemonic.startswith("vsbc") || Mnemonic.startswith("vrshr") ||
+         Mnemonic.startswith("vshr") || Mnemonic.startswith("vstrb") ||
+         Mnemonic.startswith("vldrb") ||
+         (Mnemonic.startswith("vstrh") && Mnemonic != "vstrhi") ||
+         (Mnemonic.startswith("vldrh") && Mnemonic != "vldrhi") ||
+         Mnemonic.startswith("vstrw") || Mnemonic.startswith("vldrw") ||
+         Mnemonic.startswith("vldrd") || Mnemonic.startswith("vstrd") ||
+         Mnemonic.startswith("vqdmull") || Mnemonic.startswith("vbrsr") ||
+         Mnemonic.startswith("vfmas") || Mnemonic.startswith("vmlas") ||
+         Mnemonic.startswith("vmla") || Mnemonic.startswith("vqdmlash") ||
+         Mnemonic.startswith("vqdmlah") || Mnemonic.startswith("vqrdmlash") ||
+         Mnemonic.startswith("vqrdmlah") || Mnemonic.startswith("viwdup") ||
+         Mnemonic.startswith("vdwdup") || Mnemonic.startswith("vidup") ||
+         Mnemonic.startswith("vddup") || Mnemonic.startswith("vctp") ||
+         Mnemonic.startswith("vpnot") || Mnemonic.startswith("vbic") ||
+         Mnemonic.startswith("vrmlsldavh") || Mnemonic.startswith("vmlsldav") ||
+         Mnemonic.startswith("vcvt") ||
+         (Mnemonic.startswith("vmov") &&
+          !(ExtraToken == ".f16" || ExtraToken == ".32" ||
+            ExtraToken == ".16" || ExtraToken == ".8"));
+}

llvm/trunk/lib/Target/ARM/Disassembler/ARMDisassembler.cpp

 //===- ARMDisassembler.cpp - Disassembler for ARM/Thumb ISA ---------------===//
 //
 // 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
 //
 //===----------------------------------------------------------------------===//
 
+#include "ARMBaseInstrInfo.h"
 #include "MCTargetDesc/ARMAddressingModes.h"
 #include "MCTargetDesc/ARMBaseInfo.h"
 #include "MCTargetDesc/ARMMCTargetDesc.h"
 #include "TargetInfo/ARMTargetInfo.h"
 #include "Utils/ARMBaseInfo.h"
 #include "llvm/MC/MCContext.h"
 #include "llvm/MC/MCDisassembler/MCDisassembler.h"
 #include "llvm/MC/MCFixedLenDisassembler.h"
public:
         }
         ITStates.push_back(CCBits);
       }
 
     private:
       std::vector<unsigned char> ITStates;
   };
 
+  class VPTStatus
+  {
+    public:
+      unsigned getVPTPred() {
+        unsigned Pred = ARMVCC::None;
+        if (instrInVPTBlock())
+          Pred = VPTStates.back();
+        return Pred;
+      }
+
+      void advanceVPTState() {
+        VPTStates.pop_back();
+      }
+
+      bool instrInVPTBlock() {
+        return !VPTStates.empty();
+      }
+
+      bool instrLastInVPTBlock() {
+        return VPTStates.size() == 1;
+      }
+
+      void setVPTState(char Mask) {
+        // (3 - the number of trailing zeros) is the number of then / else.
+        unsigned NumTZ = countTrailingZeros<uint8_t>(Mask);
+        assert(NumTZ <= 3 && "Invalid VPT mask!");
+        // push predicates onto the stack the correct order for the pops
+        for (unsigned Pos = NumTZ+1; Pos <= 3; ++Pos) {
+          bool T = ((Mask >> Pos) & 1) == 0;
+          if (T)
+            VPTStates.push_back(ARMVCC::Then);
+          else
+            VPTStates.push_back(ARMVCC::Else);
+        }
+        VPTStates.push_back(ARMVCC::Then);
+      }
+
+    private:
+      SmallVector<unsigned char, 4> VPTStates;
+  };
+
 /// ARM disassembler for all ARM platforms.
 class ARMDisassembler : public MCDisassembler {
 public:
   ARMDisassembler(const MCSubtargetInfo &STI, MCContext &Ctx) :
     MCDisassembler(STI, Ctx) {
   }
 
   ~ARMDisassembler() override = default;
public:
 
   DecodeStatus getInstruction(MCInst &Instr, uint64_t &Size,
                               ArrayRef<uint8_t> Bytes, uint64_t Address,
                               raw_ostream &VStream,
                               raw_ostream &CStream) const override;
 
 private:
   mutable ITStatus ITBlock;
+  mutable VPTStatus VPTBlock;
 
   DecodeStatus AddThumbPredicate(MCInst&) const;
   void UpdateThumbVFPPredicate(DecodeStatus &, MCInst&) const;
 };
 
 } // end anonymous namespace
 
 static bool Check(DecodeStatus &Out, DecodeStatus In) {

 static DecodeStatus DecodeSPR_8RegisterClass(MCInst &Inst, unsigned RegNo,
                                    uint64_t Address, const void *Decoder);
 static DecodeStatus DecodeDPR_VFP2RegisterClass(MCInst &Inst,
                                                 unsigned RegNo,
                                                 uint64_t Address,
                                                 const void *Decoder);
 static DecodeStatus DecodeQPRRegisterClass(MCInst &Inst, unsigned RegNo,
                                    uint64_t Address, const void *Decoder);
+static DecodeStatus DecodeMQPRRegisterClass(MCInst &Inst, unsigned RegNo,
+                                   uint64_t Address, const void *Decoder);
 static DecodeStatus DecodeDPairRegisterClass(MCInst &Inst, unsigned RegNo,
                                    uint64_t Address, const void *Decoder);
 static DecodeStatus DecodeDPairSpacedRegisterClass(MCInst &Inst,
                                unsigned RegNo, uint64_t Address,
                                const void *Decoder);
 
 static DecodeStatus DecodePredicateOperand(MCInst &Inst, unsigned Val,
                                uint64_t Address, const void *Decoder);
static DecodeStatus DecodePredNoALOperand(MCInst &Inst, unsigned Val,
                                           const void *Decoder);
 static DecodeStatus DecodeLOLoop(MCInst &Inst, unsigned Insn, uint64_t Address,
                                  const void *Decoder);
 static DecodeStatus DecodeLongShiftOperand(MCInst &Inst, unsigned Val,
                                            uint64_t Address,
                                            const void *Decoder);
 static DecodeStatus DecodeVSCCLRM(MCInst &Inst, unsigned Insn, uint64_t Address,
                                   const void *Decoder);
+static DecodeStatus DecodeVPTMaskOperand(MCInst &Inst, unsigned Val,
+                                         uint64_t Address, const void *Decoder);
+static DecodeStatus DecodeVpredROperand(MCInst &Inst, unsigned Val,
+                                        uint64_t Address, const void *Decoder);
+static DecodeStatus DecodeRestrictedIPredicateOperand(MCInst &Inst, unsigned Val,
+                                                     uint64_t Address,
+                                                     const void *Decoder);
+static DecodeStatus DecodeRestrictedSPredicateOperand(MCInst &Inst, unsigned Val,
+                                                     uint64_t Address,
+                                                     const void *Decoder);
+static DecodeStatus DecodeRestrictedUPredicateOperand(MCInst &Inst, unsigned Val,
+                                                     uint64_t Address,
+                                                     const void *Decoder);
+static DecodeStatus DecodeRestrictedFPPredicateOperand(MCInst &Inst,
+                                                       unsigned Val,
+                                                       uint64_t Address,
+                                                       const void *Decoder);
 template<bool Writeback>
 static DecodeStatus DecodeVSTRVLDR_SYSREG(MCInst &Inst, unsigned Insn,
                                           uint64_t Address,
                                           const void *Decoder);
 static DecodeStatus DecodeMVEOverlappingLongShift(MCInst &Inst, unsigned Insn,
                                                   uint64_t Address,
                                                   const void *Decoder);
 #include "ARMGenDisassemblerTables.inc"
if (OpInfo[i].isOptionalDef() && OpInfo[i].RegClass == ARM::CCRRegClassID) {
       MI.insert(I, MCOperand::createReg(InITBlock ? 0 : ARM::CPSR));
       return;
     }
   }
 
   MI.insert(I, MCOperand::createReg(InITBlock ? 0 : ARM::CPSR));
 }
 
+static bool isVectorPredicable(unsigned Opcode) {
+  const MCOperandInfo *OpInfo = ARMInsts[Opcode].OpInfo;
+  unsigned short NumOps = ARMInsts[Opcode].NumOperands;
+  for (unsigned i = 0; i < NumOps; ++i) {
+    if (ARM::isVpred(OpInfo[i].OperandType))
+      return true;
+  }
+  return false;
+}
+
 // Most Thumb instructions don't have explicit predicates in the
 // encoding, but rather get their predicates from IT context.  We need
 // to fix up the predicate operands using this context information as a
 // post-pass.
 MCDisassembler::DecodeStatus
 ThumbDisassembler::AddThumbPredicate(MCInst &MI) const {
   MCDisassembler::DecodeStatus S = Success;
 
case ARM::t2TBH:
       // only appears at the end of, or outside of, an IT.
       if (ITBlock.instrInITBlock() && !ITBlock.instrLastInITBlock())
         S = SoftFail;
       break;
     default:
       break;
   }
 
-  // If we're in an IT block, base the predicate on that.  Otherwise,
+  // Warn on non-VPT predicable instruction in a VPT block and a VPT
+  // predicable instruction in an IT block
+  if ((!isVectorPredicable(MI.getOpcode()) && VPTBlock.instrInVPTBlock()) ||
+       (isVectorPredicable(MI.getOpcode()) && ITBlock.instrInITBlock()))
+    S = SoftFail;
+
+  // If we're in an IT/VPT block, base the predicate on that.  Otherwise,
   // assume a predicate of AL.
-  unsigned CC;
+  unsigned CC = ARMCC::AL;
+  unsigned VCC = ARMVCC::None;
+  if (ITBlock.instrInITBlock()) {
     CC = ITBlock.getITCC();
-  if (CC == 0xF)
-    CC = ARMCC::AL;
-  if (ITBlock.instrInITBlock())
     ITBlock.advanceITState();
+  } else if (VPTBlock.instrInVPTBlock()) {
+    VCC = VPTBlock.getVPTPred();
+    VPTBlock.advanceVPTState();
+  }
 
   const MCOperandInfo *OpInfo = ARMInsts[MI.getOpcode()].OpInfo;
   unsigned short NumOps = ARMInsts[MI.getOpcode()].NumOperands;
-  MCInst::iterator I = MI.begin();
-  for (unsigned i = 0; i < NumOps; ++i, ++I) {
-    if (I == MI.end()) break;
-    if (OpInfo[i].isPredicate()) {
-      if (CC != ARMCC::AL && !ARMInsts[MI.getOpcode()].isPredicable())
-        Check(S, SoftFail);
-      I = MI.insert(I, MCOperand::createImm(CC));
-      ++I;
-      if (CC == ARMCC::AL)
-        MI.insert(I, MCOperand::createReg(0));
-      else
-        MI.insert(I, MCOperand::createReg(ARM::CPSR));
-      return S;
-    }
+
+  MCInst::iterator CCI = MI.begin();
+  for (unsigned i = 0; i < NumOps; ++i, ++CCI) {
+    if (OpInfo[i].isPredicate() || CCI == MI.end()) break;
+  }
+
+  if (ARMInsts[MI.getOpcode()].isPredicable()) {
+    CCI = MI.insert(CCI, MCOperand::createImm(CC));
+    ++CCI;
+      MI.insert(CCI, MCOperand::createReg(0));
+      MI.insert(CCI, MCOperand::createReg(ARM::CPSR));
+  } else if (CC != ARMCC::AL) {
+    Check(S, SoftFail);
+
+  MCInst::iterator VCCI = MI.begin();
+  unsigned VCCPos;
+  for (VCCPos = 0; VCCPos < NumOps; ++VCCPos, ++VCCI) {
+    if (ARM::isVpred(OpInfo[VCCPos].OperandType) || VCCI == MI.end()) break;
   }
 
-  I = MI.insert(I, MCOperand::createImm(CC));
-  ++I;
-  if (CC == ARMCC::AL)
-    MI.insert(I, MCOperand::createReg(0));
+  if (isVectorPredicable(MI.getOpcode())) {
+    VCCI = MI.insert(VCCI, MCOperand::createImm(VCC));
+    ++VCCI;
+    if (VCC == ARMVCC::None)
+      MI.insert(VCCI, MCOperand::createReg(0));
     else
-    MI.insert(I, MCOperand::createReg(ARM::CPSR));
+      MI.insert(VCCI, MCOperand::createReg(ARM::P0));
+    if (OpInfo[VCCPos].OperandType == ARM::OPERAND_VPRED_R) {
+      int TiedOp = ARMInsts[MI.getOpcode()].getOperandConstraint(
+        VCCPos + 2, MCOI::TIED_TO);
+      assert(TiedOp >= 0 &&
+             "Inactive register in vpred_r is not tied to an output!");
+      MI.insert(VCCI, MI.getOperand(TiedOp));
+    }
+  } else if (VCC != ARMVCC::None) {
+    Check(S, SoftFail);
+  }
 
   return S;
 }
 
 // Thumb VFP instructions are a special case.  Because we share their
 // encodings between ARM and Thumb modes, and they are predicable in ARM
 // mode, the auto-generated decoder will give them an (incorrect)
 // predicate operand.  We need to rewrite these operands based on the IT
 // context as a post-pass.
 void ThumbDisassembler::UpdateThumbVFPPredicate(
   DecodeStatus &S, MCInst &MI) const {
   unsigned CC;
   CC = ITBlock.getITCC();
   if (CC == 0xF)
     CC = ARMCC::AL;
   if (ITBlock.instrInITBlock())
     ITBlock.advanceITState();
+  else if (VPTBlock.instrInVPTBlock()) {
+    CC = VPTBlock.getVPTPred();
+    VPTBlock.advanceVPTState();
+  }
 
   const MCOperandInfo *OpInfo = ARMInsts[MI.getOpcode()].OpInfo;
   MCInst::iterator I = MI.begin();
   unsigned short NumOps = ARMInsts[MI.getOpcode()].NumOperands;
   for (unsigned i = 0; i < NumOps; ++i, ++I) {
     if (OpInfo[i].isPredicate() ) {
       if (CC != ARMCC::AL && !ARMInsts[MI.getOpcode()].isPredicable())
         Check(S, SoftFail);
DecodeStatus ThumbDisassembler::getInstruction(MCInst &MI, uint64_t &Size,
 
   uint32_t Insn32 =
       (Bytes[3] << 8) | (Bytes[2] << 0) | (Bytes[1] << 24) | (Bytes[0] << 16);
 
   Result =
       decodeInstruction(DecoderTableMVE32, MI, Insn32, Address, this, STI);
   if (Result != MCDisassembler::Fail) {
     Size = 4;
+
+    // Nested VPT blocks are UNPREDICTABLE. Must be checked before we add
+    // the VPT predicate.
+    if (isVPTOpcode(MI.getOpcode()) && VPTBlock.instrInVPTBlock())
+      Result = MCDisassembler::SoftFail;
+
     Check(Result, AddThumbPredicate(MI));
+
+    if (isVPTOpcode(MI.getOpcode())) {
+      unsigned Mask = MI.getOperand(0).getImm();
+      VPTBlock.setVPTState(Mask);
+    }
+
     return Result;
   }
 
   Result =
       decodeInstruction(DecoderTableThumb32, MI, Insn32, Address, this, STI);
   if (Result != MCDisassembler::Fail) {
     Size = 4;
     bool InITBlock = ITBlock.instrInITBlock();
if (!Check(S, DecodeSPRRegListOperand(Inst, reglist, Address, Decoder))) {
       return MCDisassembler::Fail;
     }
   }
   Inst.addOperand(MCOperand::createReg(ARM::VPR));
 
   return S;
 }
 
+static DecodeStatus DecodeMQPRRegisterClass(MCInst &Inst, unsigned RegNo,
+                              uint64_t Address,
+                              const void *Decoder) {
+  if (RegNo > 7)
+    return MCDisassembler::Fail;
+
+  unsigned Register = QPRDecoderTable[RegNo];
+  Inst.addOperand(MCOperand::createReg(Register));
+  return MCDisassembler::Success;
+}
+
+static DecodeStatus DecodeVPTMaskOperand(MCInst &Inst, unsigned Val,
+                                         uint64_t Address,
+                                         const void *Decoder) {
+  DecodeStatus S = MCDisassembler::Success;
+
+  // Parse VPT mask and encode it in the MCInst as an immediate with the same
+  // format as the it_mask.  That is, from the second 'e|t' encode 'e' as 1 and
+  // 't' as 0 and finish with a 1.
+  unsigned Imm = 0;
+  // We always start with a 't'.
+  unsigned CurBit = 0;
+  for (int i = 3; i >= 0; --i) {
+    // If the bit we are looking at is not the same as last one, invert the
+    // CurBit, if it is the same leave it as is.
+    CurBit ^= (Val >> i) & 1U;
+
+    // Encode the CurBit at the right place in the immediate.
+    Imm |= (CurBit << i);
+
+    // If we are done, finish the encoding with a 1.
+    if ((Val & ~(~0U << i)) == 0) {
+      Imm |= 1U << i;
+      break;
+    }
+  }
+
+  Inst.addOperand(MCOperand::createImm(Imm));
+
+  return S;
+}
+
+static DecodeStatus DecodeVpredROperand(MCInst &Inst, unsigned RegNo,
+                                        uint64_t Address, const void *Decoder) {
+  // The vpred_r operand type includes an MQPR register field derived
+  // from the encoding. But we don't actually want to add an operand
+  // to the MCInst at this stage, because AddThumbPredicate will do it
+  // later, and will infer the register number from the TIED_TO
+  // constraint. So this is a deliberately empty decoder method that
+  // will inhibit the auto-generated disassembly code from adding an
+  // operand at all.
+  return MCDisassembler::Success;
+}
+
+static DecodeStatus DecodeRestrictedIPredicateOperand(MCInst &Inst,
+                                                      unsigned Val,
+                                                      uint64_t Address,
+                                                      const void *Decoder) {
+  Inst.addOperand(MCOperand::createImm((Val & 0x1) == 0 ? ARMCC::EQ : ARMCC::NE));
+  return MCDisassembler::Success;
+}
+
+static DecodeStatus DecodeRestrictedSPredicateOperand(MCInst &Inst,
+                                                      unsigned Val,
+                                                      uint64_t Address,
+                                                      const void *Decoder) {
+  unsigned Code;
+  switch (Val & 0x3) {
+  case 0:
+    Code = ARMCC::GE;
+    break;
+  case 1:
+    Code = ARMCC::LT;
+    break;
+  case 2:
+    Code = ARMCC::GT;
+    break;
+  case 3:
+    Code = ARMCC::LE;
+    break;
+  }
+  Inst.addOperand(MCOperand::createImm(Code));
+  return MCDisassembler::Success;
+}
+
+static DecodeStatus DecodeRestrictedUPredicateOperand(MCInst &Inst,
+                                                      unsigned Val,
+                                                      uint64_t Address,
+                                                      const void *Decoder) {
+  Inst.addOperand(MCOperand::createImm((Val & 0x1) == 0 ? ARMCC::HS : ARMCC::HI));
+  return MCDisassembler::Success;
+}
+
+static DecodeStatus DecodeRestrictedFPPredicateOperand(MCInst &Inst, unsigned Val,
+                                                     uint64_t Address,
+                                                     const void *Decoder) {
+  unsigned Code;
+  switch (Val) {
+  default:
+    return MCDisassembler::Fail;
+  case 0:
+    Code = ARMCC::EQ;
+    break;
+  case 1:
+    Code = ARMCC::NE;
+    break;
+  case 4:
+    Code = ARMCC::GE;
+    break;
+  case 5:
+    Code = ARMCC::LT;
+    break;
+  case 6:
+    Code = ARMCC::GT;
+    break;
+  case 7:
+    Code = ARMCC::LE;
+    break;
+  }
+
+  Inst.addOperand(MCOperand::createImm(Code));
+  return MCDisassembler::Success;
+}
+
 static unsigned FixedRegForVSTRVLDR_SYSREG(unsigned Opcode) {
   switch (Opcode) {
   case ARM::VSTR_P0_off:
   case ARM::VSTR_P0_pre:
   case ARM::VSTR_P0_post:
   case ARM::VLDR_P0_off:
   case ARM::VLDR_P0_pre:
   case ARM::VLDR_P0_post:

llvm/trunk/lib/Target/ARM/MCTargetDesc/ARMInstPrinter.h

void printVectorListFourSpacedAllLanes(const MCInst *MI, unsigned OpNum,
                                          raw_ostream &O);
   void printVectorListThreeSpaced(const MCInst *MI, unsigned OpNum,
                                   const MCSubtargetInfo &STI, raw_ostream &O);
   void printVectorListFourSpaced(const MCInst *MI, unsigned OpNum,
                                  const MCSubtargetInfo &STI, raw_ostream &O);
   template<int64_t Angle, int64_t Remainder>
   void printComplexRotationOp(const MCInst *MI, unsigned OpNum,
                               const MCSubtargetInfo &STI, raw_ostream &O);
+  // MVE
+  void printVPTPredicateOperand(const MCInst *MI, unsigned OpNum,
+                                const MCSubtargetInfo &STI,
+                                raw_ostream &O);
+  void printVPTMask(const MCInst *MI, unsigned OpNum,
+                    const MCSubtargetInfo &STI, raw_ostream &O);
 
 private:
   unsigned DefaultAltIdx = ARM::NoRegAltName;
 };
 
 } // end namespace llvm
 
 #endif // LLVM_LIB_TARGET_ARM_MCTARGETDESC_ARMINSTPRINTER_H

llvm/trunk/lib/Target/ARM/MCTargetDesc/ARMInstPrinter.cpp

 template<int64_t Angle, int64_t Remainder>
 void ARMInstPrinter::printComplexRotationOp(const MCInst *MI, unsigned OpNo,
                                             const MCSubtargetInfo &STI,
                                             raw_ostream &O) {
   unsigned Val = MI->getOperand(OpNo).getImm();
   O << "#" << (Val * Angle) + Remainder;
 }
 
+void ARMInstPrinter::printVPTPredicateOperand(const MCInst *MI, unsigned OpNum,
+                                              const MCSubtargetInfo &STI,
+                                              raw_ostream &O) {
+  ARMVCC::VPTCodes CC = (ARMVCC::VPTCodes)MI->getOperand(OpNum).getImm();
+  if (CC != ARMVCC::None)
+    O << ARMVPTPredToString(CC);
+}
+
+void ARMInstPrinter::printVPTMask(const MCInst *MI, unsigned OpNum,
+                                  const MCSubtargetInfo &STI,
+                                  raw_ostream &O) {
+  // (3 - the number of trailing zeroes) is the number of them / else.
+  unsigned Mask = MI->getOperand(OpNum).getImm();
+  unsigned NumTZ = countTrailingZeros(Mask);
+  assert(NumTZ <= 3 && "Invalid VPT mask!");
+  for (unsigned Pos = 3, e = NumTZ; Pos > e; --Pos) {
+    bool T = ((Mask >> Pos) & 1) == 0;
+    if (T)
+      O << 't';
+    else
+      O << 'e';
+  }
+}
+

llvm/trunk/lib/Target/ARM/MCTargetDesc/ARMMCCodeEmitter.cpp

public:
   template <bool isNeg, ARM::Fixups fixup>
   uint32_t getBFTargetOpValue(const MCInst &MI, unsigned OpIdx,
                               SmallVectorImpl<MCFixup> &Fixups,
                               const MCSubtargetInfo &STI) const;
 
   uint32_t getBFAfterTargetOpValue(const MCInst &MI, unsigned OpIdx,
                                    SmallVectorImpl<MCFixup> &Fixups,
                                    const MCSubtargetInfo &STI) const;
+
+  uint32_t getVPTMaskOpValue(const MCInst &MI, unsigned OpIdx,
+                             SmallVectorImpl<MCFixup> &Fixups,
+                             const MCSubtargetInfo &STI) const;
+  uint32_t getRestrictedCondCodeOpValue(const MCInst &MI, unsigned OpIdx,
+                                        SmallVectorImpl<MCFixup> &Fixups,
+                                        const MCSubtargetInfo &STI) const;
 };
 
 } // end anonymous namespace
 
 /// NEONThumb2DataIPostEncoder - Post-process encoded NEON data-processing
 /// instructions, and rewrite them to their Thumb2 form if we are currently in
 /// Thumb2 mode.
 unsigned ARMMCCodeEmitter::NEONThumb2DataIPostEncoder(const MCInst &MI,

 unsigned ARMMCCodeEmitter::
 getMachineOpValue(const MCInst &MI, const MCOperand &MO,
                   SmallVectorImpl<MCFixup> &Fixups,
                   const MCSubtargetInfo &STI) const {
   if (MO.isReg()) {
     unsigned Reg = MO.getReg();
     unsigned RegNo = CTX.getRegisterInfo()->getEncodingValue(Reg);
 
-    // Q registers are encoded as 2x their register number.
+    // In NEON, Q registers are encoded as 2x their register number,
+    // because they're using the same indices as the D registers they
+    // overlap. In MVE, there are no 64-bit vector instructions, so
+    // the encodings all refer to Q-registers by their literal
+    // register number.
+
+    if (STI.getFeatureBits()[ARM::HasMVEIntegerOps])
+      return RegNo;
+
     switch (Reg) {
     default:
       return RegNo;
     case ARM::Q0:  case ARM::Q1:  case ARM::Q2:  case ARM::Q3:
     case ARM::Q4:  case ARM::Q5:  case ARM::Q6:  case ARM::Q7:
     case ARM::Q8:  case ARM::Q9:  case ARM::Q10: case ARM::Q11:
     case ARM::Q12: case ARM::Q13: case ARM::Q14: case ARM::Q15:
       return 2 * RegNo;
ARMMCCodeEmitter::getBFAfterTargetOpValue(const MCInst &MI, unsigned OpIdx,
   }
 
   assert(MO.isImm() && BranchMO.isImm());
   int Diff = MO.getImm() - BranchMO.getImm();
   assert(Diff == 4 || Diff == 2);
 
   return Diff == 4;
 }
+
+uint32_t ARMMCCodeEmitter::getVPTMaskOpValue(const MCInst &MI, unsigned OpIdx,
+                                             SmallVectorImpl<MCFixup> &Fixups,
+                                             const MCSubtargetInfo &STI)const {
+  const MCOperand MO = MI.getOperand(OpIdx);
+  assert(MO.isImm() && "Unexpected operand type!");
+
+  int Value = MO.getImm();
+  int Imm = 0;
+
+  // VPT Masks are actually encoded as a series of invert/don't invert bits,
+  // rather than true/false bits.
+  unsigned PrevBit = 0;
+  for (int i = 3; i >= 0; --i) {
+    unsigned Bit = (Value >> i) & 1;
+
+    // Check if we are at the end of the mask.
+    if ((Value & ~(~0U << i)) == 0) {
+      Imm |= (1 << i);
+      break;
+    }
+
+    // Convert the bit in the mask based on the previous bit.
+    if (Bit != PrevBit)
+      Imm |= (1 << i);
+
+    PrevBit = Bit;
+  }
+
+  return Imm;
+}
+
+uint32_t ARMMCCodeEmitter::getRestrictedCondCodeOpValue(
+    const MCInst &MI, unsigned OpIdx, SmallVectorImpl<MCFixup> &Fixups,
+    const MCSubtargetInfo &STI) const {
+
+  const MCOperand MO = MI.getOperand(OpIdx);
+  assert(MO.isImm() && "Unexpected operand type!");
+
+  switch (MO.getImm()) {
+  default:
+    assert(0 && "Unexpected Condition!");
+    return 0;
+  case ARMCC::HS:
+  case ARMCC::EQ:
+    return 0;
+  case ARMCC::HI:
+  case ARMCC::NE:
+    return 1;
+  case ARMCC::GE:
+    return 4;
+  case ARMCC::LT:
+    return 5;
+  case ARMCC::GT:
+    return 6;
+  case ARMCC::LE:
+    return 7;
+  }
+}
+
 #include "ARMGenMCCodeEmitter.inc"
 
 MCCodeEmitter *llvm::createARMLEMCCodeEmitter(const MCInstrInfo &MCII,
                                               const MCRegisterInfo &MRI,
                                               MCContext &Ctx) {
   return new ARMMCCodeEmitter(MCII, Ctx, true);
 }
 
 MCCodeEmitter *llvm::createARMBEMCCodeEmitter(const MCInstrInfo &MCII,
                                               const MCRegisterInfo &MRI,
                                               MCContext &Ctx) {
   return new ARMMCCodeEmitter(MCII, Ctx, false);
 }

llvm/trunk/lib/Target/ARM/MCTargetDesc/ARMMCTargetDesc.h

 //===-- ARMMCTargetDesc.h - ARM Target Descriptions -------------*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
 // This file provides ARM specific target descriptions.
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef LLVM_LIB_TARGET_ARM_MCTARGETDESC_ARMMCTARGETDESC_H
 #define LLVM_LIB_TARGET_ARM_MCTARGETDESC_ARMMCTARGETDESC_H
 
 #include "llvm/Support/DataTypes.h"
+#include "llvm/MC/MCInstrDesc.h"
 #include <memory>
 #include <string>
 
 namespace llvm {
 class formatted_raw_ostream;
 class MCAsmBackend;
 class MCCodeEmitter;
 class MCContext;
createARMMachObjectWriter(bool Is64Bit, uint32_t CPUType,
                           uint32_t CPUSubtype);
 
 /// Construct an ARM PE/COFF object writer.
 std::unique_ptr<MCObjectTargetWriter>
 createARMWinCOFFObjectWriter(bool Is64Bit);
 
 /// Construct ARM Mach-O relocation info.
 MCRelocationInfo *createARMMachORelocationInfo(MCContext &Ctx);
+
+namespace ARM {
+enum OperandType {
+  OPERAND_VPRED_R = MCOI::OPERAND_FIRST_TARGET,
+  OPERAND_VPRED_N,
+};
+inline bool isVpred(OperandType op) {
+  return op == OPERAND_VPRED_R || op == OPERAND_VPRED_N;
+}
+inline bool isVpred(uint8_t op) {
+  return isVpred(static_cast<OperandType>(op));
+}
+} // end namespace ARM
+
 } // End llvm namespace
 
 // Defines symbolic names for ARM registers.  This defines a mapping from
 // register name to register number.
 //
 #define GET_REGINFO_ENUM
 #include "ARMGenRegisterInfo.inc"
 

llvm/trunk/lib/Target/ARM/Utils/ARMBaseInfo.h

inline static CondCodes getOppositeCondition(CondCodes CC) {
   case GE: return LT;
   case LT: return GE;
   case GT: return LE;
   case LE: return GT;
   }
 }
 } // end namespace ARMCC
 
+namespace ARMVCC {
+  enum VPTCodes {
+    None = 0,
+    Then,
+    Else
+  };
+}
+
+inline static const char *ARMVPTPredToString(ARMVCC::VPTCodes CC) {
+  switch (CC) {
+  case ARMVCC::None:  return "none";
+  case ARMVCC::Then:  return "t";
+  case ARMVCC::Else:  return "e";
+  }
+  llvm_unreachable("Unknown VPT code");
+}
+
+inline static unsigned ARMVectorCondCodeFromString(StringRef CC) {
+  return StringSwitch<unsigned>(CC.lower())
+    .Case("t", ARMVCC::Then)
+    .Case("e", ARMVCC::Else)
+    .Default(~0U);
+}
+
 inline static const char *ARMCondCodeToString(ARMCC::CondCodes CC) {
   switch (CC) {
   case ARMCC::EQ:  return "eq";
   case ARMCC::NE:  return "ne";
   case ARMCC::HS:  return "hs";
   case ARMCC::LO:  return "lo";
   case ARMCC::MI:  return "mi";
   case ARMCC::PL:  return "pl";

llvm/trunk/test/MC/ARM/mve-minmax.s

+# RUN: not llvm-mc -triple=thumbv8.1m.main-none-eabi -mattr=+mve -show-encoding  < %s 2> %t \
+# RUN:   | FileCheck --check-prefix=CHECK-NOFP %s
+# RUN: FileCheck --check-prefix=ERROR-NOFP %s < %t
+# RUN: llvm-mc -triple=thumbv8.1m.main-none-eabi -mattr=+mve.fp,+fp64 -show-encoding  < %s \
+# RUN:   | FileCheck --check-prefix=CHECK %s
+
+# CHECK: vmaxnm.f32 q0, q1, q4  @ encoding: [0x02,0xff,0x58,0x0f]
+# CHECK-NOFP-NOT: vmaxnm.f32 q0, q1, q4  @ encoding: [0x02,0xff,0x58,0x0f]
+# ERROR-NOFP: [[@LINE+1]]:{{[0-9]+}}: error: instruction requires: mve.fp
+vmaxnm.f32 q0, q1, q4
+
+# CHECK: vminnm.f16 q3, q0, q1  @ encoding: [0x30,0xff,0x52,0x6f]
+# CHECK-NOFP-NOT: vminnm.f16 q3, q0, q1  @ encoding: [0x30,0xff,0x52,0x6f]
+# ERROR-NOFP: [[@LINE+1]]:{{[0-9]+}}: error: instruction requires: mve.fp
+vminnm.f16 q3, q0, q1

llvm/trunk/test/MC/ARM/mve-reductions.s

+# RUN: llvm-mc -triple=thumbv8.1m.main-none-eabi -mattr=+mve -show-encoding  < %s \
+# RUN:   | FileCheck --check-prefix=CHECK-NOFP %s
+# RUN: llvm-mc -triple=thumbv8.1m.main-none-eabi -mattr=+mve.fp,+fp64 -show-encoding  < %s \
+# RUN:   | FileCheck --check-prefix=CHECK %s
+
+# CHECK: vabav.s8  r0, q1, q3 @ encoding: [0x82,0xee,0x07,0x0f]
+# CHECK-NOFP: vabav.s8  r0, q1, q3 @ encoding: [0x82,0xee,0x07,0x0f]
+vabav.s8 r0, q1, q3
+
+# CHECK: vabav.s16  r0, q1, q3 @ encoding: [0x92,0xee,0x07,0x0f]
+# CHECK-NOFP: vabav.s16  r0, q1, q3 @ encoding: [0x92,0xee,0x07,0x0f]
+vabav.s16 r0, q1, q3
+
+# CHECK: vabav.s32  r0, q1, q3 @ encoding: [0xa2,0xee,0x07,0x0f]
+# CHECK-NOFP: vabav.s32  r0, q1, q3 @ encoding: [0xa2,0xee,0x07,0x0f]
+vabav.s32 r0, q1, q3
+
+# CHECK: vabav.u8  r0, q1, q3 @ encoding: [0x82,0xfe,0x07,0x0f]
+# CHECK-NOFP: vabav.u8  r0, q1, q3 @ encoding: [0x82,0xfe,0x07,0x0f]
+vabav.u8 r0, q1, q3
+
+# CHECK: vabav.u16  r0, q1, q3 @ encoding: [0x92,0xfe,0x07,0x0f]
+# CHECK-NOFP: vabav.u16  r0, q1, q3 @ encoding: [0x92,0xfe,0x07,0x0f]
+vabav.u16 r0, q1, q3
+
+# CHECK: vabav.u32  r0, q1, q3 @ encoding: [0xa2,0xfe,0x07,0x0f]
+# CHECK-NOFP: vabav.u32  r0, q1, q3 @ encoding: [0xa2,0xfe,0x07,0x0f]
+vabav.u32 r0, q1, q3

llvm/trunk/test/MC/ARM/mve-vpt.s

+# RUN: not llvm-mc -triple=thumbv8.1m.main-none-eabi -mattr=+mve -show-encoding  < %s \
+# RUN:   | FileCheck --check-prefix=CHECK-NOFP %s
+# RUN: not llvm-mc -triple=thumbv8.1m.main-none-eabi -mattr=+mve.fp,+fp64 -show-encoding  < %s 2>%t \
+# RUN:   | FileCheck --check-prefix=CHECK %s
+# RUN:     FileCheck --check-prefix=ERROR < %t %s
+
+# ERROR: [[@LINE+1]]:{{[0-9]+}}: {{error|note}}: VPT predicated instructions must be in VPT block
+vabavt.s32 lr, q1, q3
+
+# ERROR: [[@LINE+1]]:{{[0-9]+}}: {{error|note}}: VPT predicated instructions must be in VPT block
+vabave.u8 r12, q1, q3
+
+# ERROR: [[@LINE+2]]:{{[0-9]+}}: {{error|note}}: instructions in IT block must be predicable
+it eq
+vabav.s16 lr, q1, q3
+
+# CHECK: vpteee.i8 eq, q0, q1 @ encoding: [0x41,0xfe,0x02,0x2f]
+# CHECK-NOFP: vpteee.i8 eq, q0, q1 @ encoding: [0x41,0xfe,0x02,0x2f]
+vpteee.i8 eq, q0, q1
+vabavt.s32 lr, q1, q3
+vabave.s32 lr, q1, q3
+vabave.s32 lr, q1, q3
+vabave.s32 lr, q1, q3
+
+# CHECK: vptttt.s32 gt, q0, q1 @ encoding: [0x21,0xfe,0x03,0x3f]
+# CHECK-NOFP: vptttt.s32 gt, q0, q1 @ encoding: [0x21,0xfe,0x03,0x3f]
+vptttt.s32 gt, q0, q1
+vabavt.u32 lr, q1, q3
+vabavt.s32 lr, q1, q3
+vabavt.s16 lr, q1, q3
+vabavt.s8 lr, q1, q3
+
+# ERROR: [[@LINE+2]]:{{[0-9]+}}: {{error|note}}: instruction in VPT block must be predicable
+vpt.s8 le, q0, q1
+cinc lr, r2, lo
+
+# ----------------------------------------------------------------------
+# The following tests have to go last because of the NOFP-NOT checks inside the
+# VPT block.
+
+# CHECK: vptete.f16 ne, q0, q1 @ encoding: [0x71,0xfe,0x82,0xef]
+# CHECK-NOFP-NOT: vptete.f16 ne, q0, q1 @ encoding: [0x71,0xfe,0x82,0xef]
+vptete.f16 ne, q0, q1
+vabavt.s32 lr, q1, q3
+vabave.u32 lr, q1, q3
+vabavt.s32 lr, q1, q3
+vabave.s16 lr, q1, q3
+# ERROR: [[@LINE+1]]:{{[0-9]+}}: {{error|note}}: VPT predicated instructions must be in VPT block
+vabavt.s32 lr, q1, q3
+
+# CHECK: vpte.i8 eq, q0, q0
+# CHECK: vmaxnmt.f16 q1, q6, q2  @ encoding: [0x1c,0xff,0x54,0x2f]
+# CHECK-NOFP-NOT: vmaxnmt.f16 q1, q6, q2  @ encoding: [0x1c,0xff,0x54,0x2f]
+# CHECK-NOFP-NOT: vmaxnme.f16 q1, q6, q2  @ encoding: [0x1c,0xff,0x54,0x2f]
+vpte.i8 eq, q0, q0
+vmaxnmt.f16 q1, q6, q2
+vmaxnme.f16 q1, q6, q2

llvm/trunk/test/MC/Disassembler/ARM/mve-minmax.txt

+# RUN: llvm-mc -disassemble -triple=thumbv8.1m.main-none-eabi -mattr=+mve.fp,+fp64 -show-encoding %s | FileCheck %s
+# RUN: not llvm-mc -disassemble -triple=thumbv8.1m.main-none-eabi -show-encoding %s &> %t
+# RUN: FileCheck --check-prefix=CHECK-NOMVE < %t %s
+
+# CHECK: vmaxnm.f32 q0, q1, q4  @ encoding: [0x02,0xff,0x58,0x0f]
+# CHECK-NOMVE: [[@LINE+1]]:2: warning: invalid instruction encoding
+[0x02,0xff,0x58,0x0f]
+
+# CHECK: vminnm.f16 q3, q0, q1  @ encoding: [0x30,0xff,0x52,0x6f]
+# CHECK-NOMVE: [[@LINE+1]]:2: warning: invalid instruction encoding
+[0x30,0xff,0x52,0x6f]

llvm/trunk/test/MC/Disassembler/ARM/mve-reductions.txt

+# RUN: llvm-mc -disassemble -triple=thumbv8.1m.main-none-eabi -mattr=+mve.fp,+fp64 -show-encoding %s | FileCheck %s
+# RUN: not llvm-mc -disassemble -triple=thumbv8.1m.main-none-eabi -show-encoding %s &> %t
+# RUN: FileCheck --check-prefix=CHECK-NOMVE < %t %s
+
+[0x82 0xee 0x07 0x0f]
+# CHECK: vabav.s8  r0, q1, q3
+# CHECK-NOMVE: [[@LINE-2]]:2: warning: invalid instruction encoding
+
+[0x92 0xee 0x07 0x0f]
+# CHECK: vabav.s16 r0, q1, q3
+# CHECK-NOMVE: [[@LINE-2]]:2: warning: invalid instruction encoding
+
+[0xa2 0xee 0x07 0x0f]
+# CHECK: vabav.s32 r0, q1, q3
+# CHECK-NOMVE: [[@LINE-2]]:2: warning: invalid instruction encoding
+
+[0x82 0xfe 0x07 0x0f]
+# CHECK: vabav.u8 r0, q1, q3
+# CHECK-NOMVE: [[@LINE-2]]:2: warning: invalid instruction encoding
+
+[0x92 0xfe 0x07 0x0f]
+# CHECK: vabav.u16  r0, q1, q3
+# CHECK-NOMVE: [[@LINE-2]]:2: warning: invalid instruction encoding
+
+[0xa2 0xfe 0x07 0x0f]
+# CHECK: vabav.u32  r0, q1, q3
+# CHECK-NOMVE: [[@LINE-2]]:2: warning: invalid instruction encoding

llvm/trunk/test/MC/Disassembler/ARM/mve-vpt.txt

+# RUN: llvm-mc -disassemble -triple=thumbv8.1m.main-none-eabi -mattr=+mve.fp,+fp64 -show-encoding %s | FileCheck %s
+# RUN: not llvm-mc -disassemble -triple=thumbv8.1m.main-none-eabi -show-encoding %s &> %t
+# RUN: FileCheck --check-prefix=CHECK-NOMVE < %t %s
+
+# CHECK: vpte.i8 eq, q0, q0 @ encoding: [0x41,0xfe,0x00,0x8f]
+# CHECK-NOMVE: [[@LINE+5]]:2: warning: invalid instruction encoding
+# CHECK: vabavt.s16 lr, q3, q4  @ encoding: [0x96,0xee,0x09,0xef]
+# CHECK-NOMVE: [[@LINE+4]]:2: warning: invalid instruction encoding
+# CHECK: vabave.s16 lr, q3, q4  @ encoding: [0x96,0xee,0x09,0xef]
+# CHECK-NOMVE: [[@LINE+3]]:2: warning: invalid instruction encoding
+[0x41,0xfe,0x00,0x8f]
+[0x96,0xee,0x09,0xef]
+[0x96,0xee,0x09,0xef]