[AMDGPU] Allow register tuples to set asm names

[rampitec]

Summary

This change reverts most of the previous register name generation.
The real problem is that RegisterTuple does not generate asm names.
Added optional operand to RegisterTuple. This way we can simplify
register name access and dramatically reduce the size of static
tables for the backend.

[rampitec]

Note, with changes to tablegen we need to clean and rebuild llvm if optimized tablegen is used with debug build. There seems to be a missing dependency in this case.

[rampitec]

That also shall be possible to build register dags in a single parent class based on the same RegSeq. Everything with stride 1 is trunc/shl and everything with higher stride is decimate/shl of the same stride. They can even be combined into a single expression.

[arsenm]

LGTM

include/llvm/Target/Target.td

 //
 // The generated pseudo-registers inherit super-classes and fields from their
 // first sub-register. Most fields from the Register class are inferred, and
 // the AsmName and Dwarf numbers are cleared.
 //
 // RegisterTuples instances can be used in other set operations to form
 // register classes and so on. This is the only way of using the generated
 // registers.
-class RegisterTuples<list<SubRegIndex> Indices, list<dag> Regs> {
+//
+// RegNames may be specified to supply asm names for the generated tuples.
+// If used must have the same size as the list of produced registers.
+class RegisterTuples<list<SubRegIndex> Indices, list<dag> Regs,
+                     list<string> RegNames = []> {
   // SubRegs - N lists of registers to be zipped up. Super-registers are
   // synthesized from the first element of each SubRegs list, the second
   // element and so on.
   list<dag> SubRegs = Regs;
 
   // SubRegIndices - N SubRegIndex instances. This provides the names of the
   // sub-registers in the synthesized super-registers.
   list<SubRegIndex> SubRegIndices = Indices;
+
+  // List of asm names for the generated tuple registers.
+  list<string> RegAsmNames = RegNames;
 }
 
 
 //===----------------------------------------------------------------------===//
 // DwarfRegNum - This class provides a mapping of the llvm register enumeration
 // to the register numbering used by gcc and gdb.  These values are used by a
 // debug information writer to describe where values may be located during
 // execution.

lib/Target/AMDGPU/MCTargetDesc/AMDGPUInstPrinter.h

 //===-- AMDGPUInstPrinter.h - AMDGPU MC Inst -> ASM interface ---*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
 /// \file
 //===----------------------------------------------------------------------===//
 
 #ifndef LLVM_LIB_TARGET_AMDGPU_MCTARGETDESC_AMDGPUINSTPRINTER_H
 #define LLVM_LIB_TARGET_AMDGPU_MCTARGETDESC_AMDGPUINSTPRINTER_H
 
-#include "AMDGPUMCTargetDesc.h"
 #include "llvm/MC/MCInstPrinter.h"
 
 namespace llvm {
 
 class AMDGPUInstPrinter : public MCInstPrinter {
 public:
   AMDGPUInstPrinter(const MCAsmInfo &MAI,
                     const MCInstrInfo &MII, const MCRegisterInfo &MRI)
     : MCInstPrinter(MAI, MII, MRI) {}
 
   //Autogenerated by tblgen
   void printInstruction(const MCInst *MI, const MCSubtargetInfo &STI,
                         raw_ostream &O);
-  static const char *getRegisterName(unsigned RegNo,
-                                     unsigned AltIdx = AMDGPU::NoRegAltName);
+  static const char *getRegisterName(unsigned RegNo);
 
   void printInst(const MCInst *MI, raw_ostream &O, StringRef Annot,
                  const MCSubtargetInfo &STI) override;
   static void printRegOperand(unsigned RegNo, raw_ostream &O,
                               const MCRegisterInfo &MRI);
 
 private:
   void printU4ImmOperand(const MCInst *MI, unsigned OpNo,

lib/Target/AMDGPU/MCTargetDesc/AMDGPUInstPrinter.cpp

case AMDGPU::PRIVATE_RSRC_REG:
     llvm_unreachable("pseudo-register should not ever be emitted");
   case AMDGPU::SCC:
     llvm_unreachable("pseudo scc should not ever be emitted");
   default:
     break;
   }
 #endif
 
-  unsigned AltName = AMDGPU::NoRegAltName;
-
-  if (MRI.getRegClass(AMDGPU::VReg_64RegClassID).contains(RegNo) ||
-      MRI.getRegClass(AMDGPU::SGPR_64RegClassID).contains(RegNo) ||
-      MRI.getRegClass(AMDGPU::AReg_64RegClassID).contains(RegNo))
-    AltName = AMDGPU::Reg64;
-  else if (MRI.getRegClass(AMDGPU::VReg_128RegClassID).contains(RegNo) ||
-           MRI.getRegClass(AMDGPU::SGPR_128RegClassID).contains(RegNo) ||
-           MRI.getRegClass(AMDGPU::AReg_128RegClassID).contains(RegNo))
-    AltName = AMDGPU::Reg128;
-  else if (MRI.getRegClass(AMDGPU::VReg_96RegClassID).contains(RegNo) ||
-           MRI.getRegClass(AMDGPU::SReg_96RegClassID).contains(RegNo))
-    AltName = AMDGPU::Reg96;
-  else if (MRI.getRegClass(AMDGPU::VReg_160RegClassID).contains(RegNo) ||
-           MRI.getRegClass(AMDGPU::SReg_160RegClassID).contains(RegNo))
-    AltName = AMDGPU::Reg160;
-  else if (MRI.getRegClass(AMDGPU::VReg_256RegClassID).contains(RegNo) ||
-           MRI.getRegClass(AMDGPU::SGPR_256RegClassID).contains(RegNo))
-    AltName = AMDGPU::Reg256;
-  else if (MRI.getRegClass(AMDGPU::VReg_512RegClassID).contains(RegNo) ||
-           MRI.getRegClass(AMDGPU::SGPR_512RegClassID).contains(RegNo) ||
-           MRI.getRegClass(AMDGPU::AReg_512RegClassID).contains(RegNo))
-    AltName = AMDGPU::Reg512;
-  else if (MRI.getRegClass(AMDGPU::VReg_1024RegClassID).contains(RegNo) ||
-           MRI.getRegClass(AMDGPU::SReg_1024RegClassID).contains(RegNo) ||
-           MRI.getRegClass(AMDGPU::AReg_1024RegClassID).contains(RegNo))
-    AltName = AMDGPU::Reg1024;
-
-  O << getRegisterName(RegNo, AltName);
+  O << getRegisterName(RegNo);
 }
 
 void AMDGPUInstPrinter::printVOPDst(const MCInst *MI, unsigned OpNo,
                                     const MCSubtargetInfo &STI, raw_ostream &O) {
   if (OpNo == 0) {
     if (MII.get(MI->getOpcode()).TSFlags & SIInstrFlags::VOP3)
       O << "_e64 ";
     else if (MII.get(MI->getOpcode()).TSFlags & SIInstrFlags::DPP)

lib/Target/AMDGPU/SIRegisterInfo.cpp

default: {
           .addImm(Offset);
         FIOp.ChangeToRegister(TmpReg, false, false, true);
       }
     }
   }
 }
 
 StringRef SIRegisterInfo::getRegAsmName(unsigned Reg) const {
-  const TargetRegisterClass *RC = getMinimalPhysRegClass(Reg);
-  unsigned Size = getRegSizeInBits(*RC);
-  unsigned AltName = AMDGPU::NoRegAltName;
-
-  switch (Size) {
-  case 64:   AltName = AMDGPU::Reg64; break;
-  case 96:   AltName = AMDGPU::Reg96; break;
-  case 128:  AltName = AMDGPU::Reg128; break;
-  case 160:  AltName = AMDGPU::Reg160; break;
-  case 256:  AltName = AMDGPU::Reg256; break;
-  case 512:  AltName = AMDGPU::Reg512; break;
-  case 1024: AltName = AMDGPU::Reg1024; break;
-  }
-  return AMDGPUInstPrinter::getRegisterName(Reg, AltName);
+  return AMDGPUInstPrinter::getRegisterName(Reg);
 }
 
 // FIXME: This is very slow. It might be worth creating a map from physreg to
 // register class.
 const TargetRegisterClass *SIRegisterInfo::getPhysRegClass(unsigned Reg) const {
   assert(!TargetRegisterInfo::isVirtualRegister(Reg));
 
   static const TargetRegisterClass *const BaseClasses[] = {

lib/Target/AMDGPU/SIRegisterInfo.td

class getSubRegs<int size> {
   list<SubRegIndex> ret = !if(!eq(size, 2), ret2,
                               !if(!eq(size, 3), ret3,
                                   !if(!eq(size, 4), ret4,
                                       !if(!eq(size, 5), ret5,
                                           !if(!eq(size, 8), ret8,
                                               !if(!eq(size, 16), ret16, ret32))))));
 }
 
-let Namespace = "AMDGPU" in {
-defset list<RegAltNameIndex> AllRegAltNameIndices = {
-  def Reg64   : RegAltNameIndex;
-  def Reg96   : RegAltNameIndex;
-  def Reg128  : RegAltNameIndex;
-  def Reg160  : RegAltNameIndex;
-  def Reg256  : RegAltNameIndex;
-  def Reg512  : RegAltNameIndex;
-  def Reg1024 : RegAltNameIndex;
-}
+// Generates list of sequential register tuple names.
+// E.g. RegSeq<3,2,2,"s">.ret -> [ "s[0:1]", "s[2:3]" ]
+class RegSeq<int last_reg, int stride, int size, string prefix, int start = 0> {
+  int next = !add(start, stride);
+  int end_reg = !add(!add(start, size), -1);
+  list<string> ret =
+    !if(!le(end_reg, last_reg),
+        !listconcat([prefix # "[" # start # ":" # end_reg # "]"],
+                    RegSeq<last_reg, stride, size, prefix, next>.ret),
+                    []);
 }
 
 //===----------------------------------------------------------------------===//
 //  Declarations that describe the SI registers
 //===----------------------------------------------------------------------===//
-class SIReg <string n, bits<16> regIdx = 0, string prefix = "",
-             int regNo = !cast<int>(regIdx)> :
-  Register<n, !if(!eq(prefix, ""),
-                [ n, n, n, n, n, n, n ],
-                [ prefix # "[" # regNo # ":" # !and(!add(regNo, 1), 255) # "]",
-                  prefix # "[" # regNo # ":" # !and(!add(regNo, 2), 255) # "]",
-                  prefix # "[" # regNo # ":" # !and(!add(regNo, 3), 255) # "]",
-                  prefix # "[" # regNo # ":" # !and(!add(regNo, 4), 255) # "]",
-                  prefix # "[" # regNo # ":" # !and(!add(regNo, 7), 255) # "]",
-                  prefix # "[" # regNo # ":" # !and(!add(regNo, 15), 255) # "]",
-                  prefix # "[" # regNo # ":" # !and(!add(regNo, 31), 255) # "]",
-                ])>,
+class SIReg <string n, bits<16> regIdx = 0> :
+  Register<n>,
   DwarfRegNum<[!cast<int>(HWEncoding)]> {
   let Namespace = "AMDGPU";
-  let RegAltNameIndices = AllRegAltNameIndices;
 
   // This is the not yet the complete register encoding. An additional
   // bit is set for VGPRs.
   let HWEncoding = regIdx;
 }
 
-class SIRegisterWithSubRegs<string n, list<Register> subregs> :
-  RegisterWithSubRegs<n, subregs> {
-  let RegAltNameIndices = AllRegAltNameIndices;
-  let AltNames = [ n, n, n, n, n, n, n ];
-}
-
 // Special Registers
 def VCC_LO : SIReg<"vcc_lo", 106>;
 def VCC_HI : SIReg<"vcc_hi", 107>;
 
 // Pseudo-registers: Used as placeholders during isel and immediately
 // replaced, never seeing the verifier.
 def PRIVATE_RSRC_REG : SIReg<"private_rsrc", 0>;
 def FP_REG : SIReg<"fp", 0>;
 def SP_REG : SIReg<"sp", 0>;
 def SCRATCH_WAVE_OFFSET_REG : SIReg<"scratch_wave_offset", 0>;
 
 // VCC for 64-bit instructions
-def VCC : SIRegisterWithSubRegs<"vcc", [VCC_LO, VCC_HI]>,
+def VCC : RegisterWithSubRegs<"vcc", [VCC_LO, VCC_HI]>,
           DwarfRegAlias<VCC_LO> {
   let Namespace = "AMDGPU";
   let SubRegIndices = [sub0, sub1];
   let HWEncoding = 106;
 }
 
 def EXEC_LO : SIReg<"exec_lo", 126>;
 def EXEC_HI : SIReg<"exec_hi", 127>;
 
-def EXEC : SIRegisterWithSubRegs<"exec", [EXEC_LO, EXEC_HI]>,
+def EXEC : RegisterWithSubRegs<"exec", [EXEC_LO, EXEC_HI]>,
            DwarfRegAlias<EXEC_LO> {
   let Namespace = "AMDGPU";
   let SubRegIndices = [sub0, sub1];
   let HWEncoding = 126;
 }
 
 // 32-bit real registers, for MC only.
 // May be used with both 32-bit and 64-bit operands.

 def SRC_PRIVATE_LIMIT : SIReg<"src_private_limit", 238>;
 def SRC_POPS_EXITING_WAVE_ID : SIReg<"src_pops_exiting_wave_id", 239>;
 
 def LDS_DIRECT : SIReg <"src_lds_direct", 254>;
 
 def XNACK_MASK_LO : SIReg<"xnack_mask_lo", 104>;
 def XNACK_MASK_HI : SIReg<"xnack_mask_hi", 105>;
 
-def XNACK_MASK : SIRegisterWithSubRegs<"xnack_mask", [XNACK_MASK_LO, XNACK_MASK_HI]>,
+def XNACK_MASK : RegisterWithSubRegs<"xnack_mask", [XNACK_MASK_LO, XNACK_MASK_HI]>,
                  DwarfRegAlias<XNACK_MASK_LO> {
   let Namespace = "AMDGPU";
   let SubRegIndices = [sub0, sub1];
   let HWEncoding = 104;
 }
 
 // Trap handler registers
 def TBA_LO : SIReg<"tba_lo", 108>;
 def TBA_HI : SIReg<"tba_hi", 109>;
 
-def TBA : SIRegisterWithSubRegs<"tba", [TBA_LO, TBA_HI]>,
+def TBA : RegisterWithSubRegs<"tba", [TBA_LO, TBA_HI]>,
           DwarfRegAlias<TBA_LO> {
   let Namespace = "AMDGPU";
   let SubRegIndices = [sub0, sub1];
   let HWEncoding = 108;
 }
 
 def TMA_LO : SIReg<"tma_lo", 110>;
 def TMA_HI : SIReg<"tma_hi", 111>;
 
-def TMA : SIRegisterWithSubRegs<"tma", [TMA_LO, TMA_HI]>,
+def TMA : RegisterWithSubRegs<"tma", [TMA_LO, TMA_HI]>,
           DwarfRegAlias<TMA_LO> {
   let Namespace = "AMDGPU";
   let SubRegIndices = [sub0, sub1];
   let HWEncoding = 110;
 }
 
 foreach Index = 0-15 in {
   def TTMP#Index#_vi         : SIReg<"ttmp"#Index, !add(112, Index)>;
   def TTMP#Index#_gfx9_gfx10 : SIReg<"ttmp"#Index, !add(108, Index)>;
   def TTMP#Index             : SIReg<"ttmp"#Index, 0>;
 }
 
 multiclass FLAT_SCR_LOHI_m <string n, bits<16> ci_e, bits<16> vi_e> {
   def _ci : SIReg<n, ci_e>;
   def _vi : SIReg<n, vi_e>;
   def "" : SIReg<n, 0>;
 }
 
 class FlatReg <Register lo, Register hi, bits<16> encoding> :
-    SIRegisterWithSubRegs<"flat_scratch", [lo, hi]>,
+    RegisterWithSubRegs<"flat_scratch", [lo, hi]>,
     DwarfRegAlias<lo> {
   let Namespace = "AMDGPU";
   let SubRegIndices = [sub0, sub1];
   let HWEncoding = encoding;
 }
 
 defm FLAT_SCR_LO : FLAT_SCR_LOHI_m<"flat_scratch_lo", 104, 102>; // Offset in units of 256-bytes.
 defm FLAT_SCR_HI : FLAT_SCR_LOHI_m<"flat_scratch_hi", 105, 103>; // Size is the per-thread scratch size, in bytes.
 
 def FLAT_SCR_ci : FlatReg<FLAT_SCR_LO_ci, FLAT_SCR_HI_ci, 104>;
 def FLAT_SCR_vi : FlatReg<FLAT_SCR_LO_vi, FLAT_SCR_HI_vi, 102>;
 def FLAT_SCR : FlatReg<FLAT_SCR_LO, FLAT_SCR_HI, 0>;
 
 // SGPR registers
 foreach Index = 0-105 in {
-  def SGPR#Index : SIReg <"s"#Index, Index, "s">;
+  def SGPR#Index : SIReg <"s"#Index, Index>;
 }
 
 // VGPR registers
 foreach Index = 0-255 in {
-  def VGPR#Index : SIReg <"v"#Index, Index, "v"> {
+  def VGPR#Index : SIReg <"v"#Index, Index> {
     let HWEncoding{8} = 1;
   }
 }
 
 // AccVGPR registers
 foreach Index = 0-255 in {
-  def AGPR#Index : SIReg <"a"#Index, Index, "a"> {
+  def AGPR#Index : SIReg <"a"#Index, Index> {
     let HWEncoding{8} = 1;
   }
 }
 
 //===----------------------------------------------------------------------===//
 //  Groupings using register classes and tuples
 //===----------------------------------------------------------------------===//
 
def SGPR_32 : RegisterClass<"AMDGPU", [i32, f32, i16, f16, v2i16, v2f16], 32,
   // Give all SGPR classes higher priority than VGPR classes, because
   // we want to spill SGPRs to VGPRs.
   let AllocationPriority = 9;
 }
 
 // SGPR 64-bit registers
 def SGPR_64Regs : RegisterTuples<getSubRegs<2>.ret,
                              [(add (decimate SGPR_32, 2)),
-                              (add (decimate (shl SGPR_32, 1), 2))]>;
+                              (add (decimate (shl SGPR_32, 1), 2))],
+                             RegSeq<105, 2, 2, "s">.ret>;
 
 // SGPR 96-bit registers. No operations use these, but for symmetry with 96-bit VGPRs.
 def SGPR_96Regs : RegisterTuples<getSubRegs<3>.ret,
                             [(add (decimate SGPR_32, 3)),
                              (add (decimate (shl SGPR_32, 1), 3)),
-                             (add (decimate (shl SGPR_32, 2), 3))]>;
+                             (add (decimate (shl SGPR_32, 2), 3))],
+                            RegSeq<105, 3, 3, "s">.ret>;
 
 // SGPR 128-bit registers
 def SGPR_128Regs : RegisterTuples<getSubRegs<4>.ret,
                               [(add (decimate SGPR_32, 4)),
                                (add (decimate (shl SGPR_32, 1), 4)),
                                (add (decimate (shl SGPR_32, 2), 4)),
-                               (add (decimate (shl SGPR_32, 3), 4))]>;
+                               (add (decimate (shl SGPR_32, 3), 4))],
+                              RegSeq<105, 4, 4, "s">.ret>;
 
 // SGPR 160-bit registers. No operations use these, but for symmetry with 160-bit VGPRs.
 def SGPR_160Regs : RegisterTuples<getSubRegs<5>.ret,
                             [(add (decimate SGPR_32, 4)),
                              (add (decimate (shl SGPR_32, 1), 4)),
                              (add (decimate (shl SGPR_32, 2), 4)),
                              (add (decimate (shl SGPR_32, 3), 4)),
-                             (add (decimate (shl SGPR_32, 4), 4))]>;
+                             (add (decimate (shl SGPR_32, 4), 4))],
+                            RegSeq<105, 4, 5, "s">.ret>;
 
 // SGPR 256-bit registers
 def SGPR_256Regs : RegisterTuples<getSubRegs<8>.ret,
                               [(add (decimate SGPR_32, 4)),
                                (add (decimate (shl SGPR_32, 1), 4)),
                                (add (decimate (shl SGPR_32, 2), 4)),
                                (add (decimate (shl SGPR_32, 3), 4)),
                                (add (decimate (shl SGPR_32, 4), 4)),
                                (add (decimate (shl SGPR_32, 5), 4)),
                                (add (decimate (shl SGPR_32, 6), 4)),
-                               (add (decimate (shl SGPR_32, 7), 4))]>;
+                               (add (decimate (shl SGPR_32, 7), 4))],
+                              RegSeq<105, 4, 8, "s">.ret>;
 
 // SGPR 512-bit registers
 def SGPR_512Regs : RegisterTuples<getSubRegs<16>.ret,
                               [(add (decimate SGPR_32, 4)),
                                (add (decimate (shl SGPR_32, 1), 4)),
                                (add (decimate (shl SGPR_32, 2), 4)),
                                (add (decimate (shl SGPR_32, 3), 4)),
                                (add (decimate (shl SGPR_32, 4), 4)),
                                (add (decimate (shl SGPR_32, 5), 4)),
                                (add (decimate (shl SGPR_32, 6), 4)),
                                (add (decimate (shl SGPR_32, 7), 4)),
                                (add (decimate (shl SGPR_32, 8), 4)),
                                (add (decimate (shl SGPR_32, 9), 4)),
                                (add (decimate (shl SGPR_32, 10), 4)),
                                (add (decimate (shl SGPR_32, 11), 4)),
                                (add (decimate (shl SGPR_32, 12), 4)),
                                (add (decimate (shl SGPR_32, 13), 4)),
                                (add (decimate (shl SGPR_32, 14), 4)),
-                               (add (decimate (shl SGPR_32, 15), 4))]>;
+                               (add (decimate (shl SGPR_32, 15), 4))],
+                              RegSeq<105, 4, 16, "s">.ret>;
 
 // SGPR 1024-bit registers
 def SGPR_1024Regs : RegisterTuples<getSubRegs<32>.ret,
                               [(add (decimate SGPR_32, 4)),
                                (add (decimate (shl SGPR_32, 1), 4)),
                                (add (decimate (shl SGPR_32, 2), 4)),
                                (add (decimate (shl SGPR_32, 3), 4)),
                                (add (decimate (shl SGPR_32, 4), 4)),
def SGPR_1024Regs : RegisterTuples<getSubRegs<32>.ret,
                                (add (decimate (shl SGPR_32, 23), 4)),
                                (add (decimate (shl SGPR_32, 24), 4)),
                                (add (decimate (shl SGPR_32, 25), 4)),
                                (add (decimate (shl SGPR_32, 26), 4)),
                                (add (decimate (shl SGPR_32, 27), 4)),
                                (add (decimate (shl SGPR_32, 28), 4)),
                                (add (decimate (shl SGPR_32, 29), 4)),
                                (add (decimate (shl SGPR_32, 30), 4)),
-                               (add (decimate (shl SGPR_32, 31), 4))]>;
+                               (add (decimate (shl SGPR_32, 31), 4))],
+                              RegSeq<105, 4, 32, "s">.ret>;
 
 // Trap handler TMP 32-bit registers
 def TTMP_32 : RegisterClass<"AMDGPU", [i32, f32, v2i16, v2f16], 32,
                             (add (sequence "TTMP%u", 0, 15))> {
   let isAllocatable = 0;
 }
 
 // Trap handler TMP 64-bit registers
 def TTMP_64Regs : RegisterTuples<getSubRegs<2>.ret,
                              [(add (decimate TTMP_32, 2)),
-                              (add (decimate (shl TTMP_32, 1), 2))]>;
+                              (add (decimate (shl TTMP_32, 1), 2))],
+                             RegSeq<15, 2, 2, "ttmp">.ret>;
 
 // Trap handler TMP 128-bit registers
 def TTMP_128Regs : RegisterTuples<getSubRegs<4>.ret,
                               [(add (decimate TTMP_32, 4)),
                                (add (decimate (shl TTMP_32, 1), 4)),
                                (add (decimate (shl TTMP_32, 2), 4)),
-                               (add (decimate (shl TTMP_32, 3), 4))]>;
+                               (add (decimate (shl TTMP_32, 3), 4))],
+                              RegSeq<15, 4, 4, "ttmp">.ret>;
 
 def TTMP_256Regs : RegisterTuples<getSubRegs<8>.ret,
                               [(add (decimate TTMP_32, 4)),
                                (add (decimate (shl TTMP_32, 1), 4)),
                                (add (decimate (shl TTMP_32, 2), 4)),
                                (add (decimate (shl TTMP_32, 3), 4)),
                                (add (decimate (shl TTMP_32, 4), 4)),
                                (add (decimate (shl TTMP_32, 5), 4)),
                                (add (decimate (shl TTMP_32, 6), 4)),
-                               (add (decimate (shl TTMP_32, 7), 4))]>;
+                               (add (decimate (shl TTMP_32, 7), 4))],
+                              RegSeq<15, 4, 8, "ttmp">.ret>;
 
 def TTMP_512Regs : RegisterTuples<getSubRegs<16>.ret,
                               [(add (decimate TTMP_32, 4)),
                                (add (decimate (shl TTMP_32, 1), 4)),
                                (add (decimate (shl TTMP_32, 2), 4)),
                                (add (decimate (shl TTMP_32, 3), 4)),
                                (add (decimate (shl TTMP_32, 4), 4)),
                                (add (decimate (shl TTMP_32, 5), 4)),
                                (add (decimate (shl TTMP_32, 6), 4)),
                                (add (decimate (shl TTMP_32, 7), 4)),
                                (add (decimate (shl TTMP_32, 8), 4)),
                                (add (decimate (shl TTMP_32, 9), 4)),
                                (add (decimate (shl TTMP_32, 10), 4)),
                                (add (decimate (shl TTMP_32, 11), 4)),
                                (add (decimate (shl TTMP_32, 12), 4)),
                                (add (decimate (shl TTMP_32, 13), 4)),
                                (add (decimate (shl TTMP_32, 14), 4)),
-                               (add (decimate (shl TTMP_32, 15), 4))]>;
+                               (add (decimate (shl TTMP_32, 15), 4))],
+                              RegSeq<15, 4, 16, "ttmp">.ret>;
 
 class TmpRegTuplesBase<int index, int size,
                        list<Register> subRegs,
                        list<SubRegIndex> indices = getSubRegs<size>.ret,
                        int index1 = !add(index, !add(size, -1)),
                        string name = "ttmp["#index#":"#index1#"]"> :
-  SIRegisterWithSubRegs<name, subRegs> {
+  RegisterWithSubRegs<name, subRegs> {
   let HWEncoding = subRegs[0].HWEncoding;
   let SubRegIndices = indices;
 }
 
 class TmpRegTuples<string tgt,
                    int size,
                    int index0,
                    int index1 = !add(index0, 1),
def VGPR_32 : RegisterClass<"AMDGPU", [i32, f32, i16, f16, v2i16, v2f16], 32,
                             (add (sequence "VGPR%u", 0, 255))> {
   let AllocationPriority = 1;
   let Size = 32;
 }
 
 // VGPR 64-bit registers
 def VGPR_64 : RegisterTuples<getSubRegs<2>.ret,
                              [(add (trunc VGPR_32, 255)),
-                              (add (shl VGPR_32, 1))]>;
+                              (add (shl VGPR_32, 1))],
+                             RegSeq<255, 1, 2, "v">.ret>;
 
 // VGPR 96-bit registers
 def VGPR_96 : RegisterTuples<getSubRegs<3>.ret,
                              [(add (trunc VGPR_32, 254)),
                               (add (shl VGPR_32, 1)),
-                              (add (shl VGPR_32, 2))]>;
+                              (add (shl VGPR_32, 2))],
+                             RegSeq<255, 1, 3, "v">.ret>;
 
 // VGPR 128-bit registers
 def VGPR_128 : RegisterTuples<getSubRegs<4>.ret,
                               [(add (trunc VGPR_32, 253)),
                                (add (shl VGPR_32, 1)),
                                (add (shl VGPR_32, 2)),
-                               (add (shl VGPR_32, 3))]>;
+                               (add (shl VGPR_32, 3))],
+                              RegSeq<255, 1, 4, "v">.ret>;
 
 // VGPR 160-bit registers
 def VGPR_160 : RegisterTuples<getSubRegs<5>.ret,
                              [(add (trunc VGPR_32, 252)),
                               (add (shl VGPR_32, 1)),
                               (add (shl VGPR_32, 2)),
                               (add (shl VGPR_32, 3)),
-                              (add (shl VGPR_32, 4))]>;
+                              (add (shl VGPR_32, 4))],
+                             RegSeq<255, 1, 5, "v">.ret>;
 
 // VGPR 256-bit registers
 def VGPR_256 : RegisterTuples<getSubRegs<8>.ret,
                               [(add (trunc VGPR_32, 249)),
                                (add (shl VGPR_32, 1)),
                                (add (shl VGPR_32, 2)),
                                (add (shl VGPR_32, 3)),
                                (add (shl VGPR_32, 4)),
                                (add (shl VGPR_32, 5)),
                                (add (shl VGPR_32, 6)),
-                               (add (shl VGPR_32, 7))]>;
+                               (add (shl VGPR_32, 7))],
+                              RegSeq<255, 1, 8, "v">.ret>;
 
 // VGPR 512-bit registers
 def VGPR_512 : RegisterTuples<getSubRegs<16>.ret,
                               [(add (trunc VGPR_32, 241)),
                                (add (shl VGPR_32, 1)),
                                (add (shl VGPR_32, 2)),
                                (add (shl VGPR_32, 3)),
                                (add (shl VGPR_32, 4)),
                                (add (shl VGPR_32, 5)),
                                (add (shl VGPR_32, 6)),
                                (add (shl VGPR_32, 7)),
                                (add (shl VGPR_32, 8)),
                                (add (shl VGPR_32, 9)),
                                (add (shl VGPR_32, 10)),
                                (add (shl VGPR_32, 11)),
                                (add (shl VGPR_32, 12)),
                                (add (shl VGPR_32, 13)),
                                (add (shl VGPR_32, 14)),
-                               (add (shl VGPR_32, 15))]>;
+                               (add (shl VGPR_32, 15))],
+                              RegSeq<255, 1, 16, "v">.ret>;
 
 // VGPR 1024-bit registers
 def VGPR_1024 : RegisterTuples<getSubRegs<32>.ret,
                               [(add (trunc VGPR_32, 225)),
                                (add (shl VGPR_32, 1)),
                                (add (shl VGPR_32, 2)),
                                (add (shl VGPR_32, 3)),
                                (add (shl VGPR_32, 4)),
def VGPR_1024 : RegisterTuples<getSubRegs<32>.ret,
                                (add (shl VGPR_32, 23)),
                                (add (shl VGPR_32, 24)),
                                (add (shl VGPR_32, 25)),
                                (add (shl VGPR_32, 26)),
                                (add (shl VGPR_32, 27)),
                                (add (shl VGPR_32, 28)),
                                (add (shl VGPR_32, 29)),
                                (add (shl VGPR_32, 30)),
-                               (add (shl VGPR_32, 31))]>;
+                               (add (shl VGPR_32, 31))],
+                              RegSeq<255, 1, 32, "v">.ret>;
 
 // AccVGPR 32-bit registers
 def AGPR_32 : RegisterClass<"AMDGPU", [i32, f32, i16, f16, v2i16, v2f16], 32,
                             (add (sequence "AGPR%u", 0, 255))> {
   let AllocationPriority = 1;
   let Size = 32;
 }
 
 // AGPR 64-bit registers
 def AGPR_64 : RegisterTuples<getSubRegs<2>.ret,
                              [(add (trunc AGPR_32, 255)),
-                              (add (shl AGPR_32, 1))]>;
+                              (add (shl AGPR_32, 1))],
+                             RegSeq<255, 1, 2, "a">.ret>;
 
 // AGPR 128-bit registers
 def AGPR_128 : RegisterTuples<getSubRegs<4>.ret,
                               [(add (trunc AGPR_32, 253)),
                                (add (shl AGPR_32, 1)),
                                (add (shl AGPR_32, 2)),
-                               (add (shl AGPR_32, 3))]>;
+                               (add (shl AGPR_32, 3))],
+                              RegSeq<255, 1, 4, "a">.ret>;
 
 // AGPR 512-bit registers
 def AGPR_512 : RegisterTuples<getSubRegs<16>.ret,
                               [(add (trunc AGPR_32, 241)),
                                (add (shl AGPR_32, 1)),
                                (add (shl AGPR_32, 2)),
                                (add (shl AGPR_32, 3)),
                                (add (shl AGPR_32, 4)),
                                (add (shl AGPR_32, 5)),
                                (add (shl AGPR_32, 6)),
                                (add (shl AGPR_32, 7)),
                                (add (shl AGPR_32, 8)),
                                (add (shl AGPR_32, 9)),
                                (add (shl AGPR_32, 10)),
                                (add (shl AGPR_32, 11)),
                                (add (shl AGPR_32, 12)),
                                (add (shl AGPR_32, 13)),
                                (add (shl AGPR_32, 14)),
-                               (add (shl AGPR_32, 15))]>;
+                               (add (shl AGPR_32, 15))],
+                              RegSeq<255, 1, 16, "a">.ret>;
 
 // AGPR 1024-bit registers
 def AGPR_1024 : RegisterTuples<getSubRegs<32>.ret,
                               [(add (trunc AGPR_32, 225)),
                                (add (shl AGPR_32, 1)),
                                (add (shl AGPR_32, 2)),
                                (add (shl AGPR_32, 3)),
                                (add (shl AGPR_32, 4)),
def AGPR_1024 : RegisterTuples<getSubRegs<32>.ret,
                                (add (shl AGPR_32, 23)),
                                (add (shl AGPR_32, 24)),
                                (add (shl AGPR_32, 25)),
                                (add (shl AGPR_32, 26)),
                                (add (shl AGPR_32, 27)),
                                (add (shl AGPR_32, 28)),
                                (add (shl AGPR_32, 29)),
                                (add (shl AGPR_32, 30)),
-                               (add (shl AGPR_32, 31))]>;
+                               (add (shl AGPR_32, 31))],
+                              RegSeq<255, 1, 32, "a">.ret>;
 
 //===----------------------------------------------------------------------===//
 //  Register classes used as source and destination
 //===----------------------------------------------------------------------===//
 
 def Pseudo_SReg_32 : RegisterClass<"AMDGPU", [i32, f32, i16, f16, v2i16, v2f16], 32,
   (add FP_REG, SP_REG, SCRATCH_WAVE_OFFSET_REG)> {
   let isAllocatable = 0;
   let CopyCost = -1;
 }
 
 def Pseudo_SReg_128 : RegisterClass<"AMDGPU", [v4i32, v2i64, v2f64], 32,
-  (add PRIVATE_RSRC_REG), Reg128> {
+  (add PRIVATE_RSRC_REG)> {
   let isAllocatable = 0;
   let CopyCost = -1;
 }
 
 def LDS_DIRECT_CLASS : RegisterClass<"AMDGPU", [i32, f32, i16, f16, v2i16, v2f16], 32,
   (add LDS_DIRECT)> {
   let isAllocatable = 0;
   let CopyCost = -1;

 }
 
 def SRegOrLds_32 : RegisterClass<"AMDGPU", [i32, f32, i16, f16, v2i16, v2f16, i1], 32,
   (add SReg_32_XM0, M0_CLASS, EXEC_LO, EXEC_HI, SReg_32_XEXEC_HI, LDS_DIRECT_CLASS)> {
   let isAllocatable = 0;
 }
 
 def SGPR_64 : RegisterClass<"AMDGPU", [v2i32, i64, v2f32, f64, v4i16, v4f16], 32,
-                            (add SGPR_64Regs), Reg64> {
+                            (add SGPR_64Regs)> {
   let CopyCost = 1;
   let AllocationPriority = 11;
 }
 
 // CCR (call clobbered registers) SGPR 64-bit registers
 def CCR_SGPR_64 : RegisterClass<"AMDGPU", SGPR_64.RegTypes, 32,
-                                (add (trunc SGPR_64, 16)), Reg64> {
+                                (add (trunc SGPR_64, 16))> {
   let CopyCost = SGPR_64.CopyCost;
   let AllocationPriority = SGPR_64.AllocationPriority;
 }
 
 def TTMP_64 : RegisterClass<"AMDGPU", [v2i32, i64, f64, v4i16, v4f16], 32,
                             (add TTMP_64Regs)> {
   let isAllocatable = 0;
 }
 
 def SReg_64_XEXEC : RegisterClass<"AMDGPU", [v2i32, i64, v2f32, f64, i1, v4i16, v4f16], 32,
-  (add SGPR_64, VCC, FLAT_SCR, XNACK_MASK, TTMP_64, TBA, TMA), Reg64> {
+  (add SGPR_64, VCC, FLAT_SCR, XNACK_MASK, TTMP_64, TBA, TMA)> {
   let CopyCost = 1;
   let AllocationPriority = 13;
 }
 
 def SReg_64 : RegisterClass<"AMDGPU", [v2i32, i64, v2f32, f64, i1, v4i16, v4f16], 32,
-  (add SReg_64_XEXEC, EXEC), Reg64> {
+  (add SReg_64_XEXEC, EXEC)> {
   let CopyCost = 1;
   let AllocationPriority = 13;
 }
 
 def SReg_1_XEXEC : RegisterClass<"AMDGPU", [i1], 32,
   (add SReg_64_XEXEC, SReg_32_XM0_XEXEC)> {
   let CopyCost = 1;
   let isAllocatable = 0;
 }
 
 def SReg_1 : RegisterClass<"AMDGPU", [i1], 32,
   (add SReg_1_XEXEC, EXEC, EXEC_LO)> {
   let CopyCost = 1;
   let isAllocatable = 0;
 }
 
 // Requires 2 s_mov_b64 to copy
 let CopyCost = 2 in {
 
 // There are no 3-component scalar instructions, but this is needed
 // for symmetry with VGPRs.
 def SGPR_96 : RegisterClass<"AMDGPU", [v3i32, v3f32], 32,
-  (add SGPR_96Regs), Reg96> {
+  (add SGPR_96Regs)> {
   let AllocationPriority = 14;
 }
 
 def SReg_96 : RegisterClass<"AMDGPU", [v3i32, v3f32], 32,
-  (add SGPR_96), Reg96> {
+  (add SGPR_96)> {
   let AllocationPriority = 14;
 }
 
 def SGPR_128 : RegisterClass<"AMDGPU", [v4i32, v4f32, v2i64], 32,
-                             (add SGPR_128Regs), Reg128> {
+                             (add SGPR_128Regs)> {
   let AllocationPriority = 15;
 }
 
 def TTMP_128 : RegisterClass<"AMDGPU", [v4i32, v4f32, v2i64], 32,
                              (add TTMP_128Regs)> {
   let isAllocatable = 0;
 }
 
 def SReg_128 : RegisterClass<"AMDGPU", [v4i32, v4f32, v2i64, v2f64], 32,
-                             (add SGPR_128, TTMP_128), Reg128> {
+                             (add SGPR_128, TTMP_128)> {
   let AllocationPriority = 15;
 }
 
 } // End CopyCost = 2
 
 // There are no 5-component scalar instructions, but this is needed
 // for symmetry with VGPRs.
 def SGPR_160 : RegisterClass<"AMDGPU", [v5i32, v5f32], 32,
-                             (add SGPR_160Regs), Reg160> {
+                             (add SGPR_160Regs)> {
   let AllocationPriority = 16;
 }
 
 def SReg_160 : RegisterClass<"AMDGPU", [v5i32, v5f32], 32,
-                             (add SGPR_160), Reg160> {
+                             (add SGPR_160)> {
   let AllocationPriority = 16;
 }
 
-def SGPR_256 : RegisterClass<"AMDGPU", [v8i32, v8f32], 32, (add SGPR_256Regs),
-                             Reg256> {
+def SGPR_256 : RegisterClass<"AMDGPU", [v8i32, v8f32], 32, (add SGPR_256Regs)> {
   let AllocationPriority = 17;
 }
 
 def TTMP_256 : RegisterClass<"AMDGPU", [v8i32, v8f32], 32, (add TTMP_256Regs)> {
   let isAllocatable = 0;
 }
 
 def SReg_256 : RegisterClass<"AMDGPU", [v8i32, v8f32], 32,
-                             (add SGPR_256, TTMP_256), Reg256> {
+                             (add SGPR_256, TTMP_256)> {
   // Requires 4 s_mov_b64 to copy
   let CopyCost = 4;
   let AllocationPriority = 17;
 }
 
 def SGPR_512 : RegisterClass<"AMDGPU", [v16i32, v16f32], 32,
-                             (add SGPR_512Regs), Reg512> {
+                             (add SGPR_512Regs)> {
   let AllocationPriority = 18;
 }
 
 def TTMP_512 : RegisterClass<"AMDGPU", [v16i32, v16f32], 32,
                              (add TTMP_512Regs)> {
   let isAllocatable = 0;
 }
 
 def SReg_512 : RegisterClass<"AMDGPU", [v16i32, v16f32], 32,
-                             (add SGPR_512, TTMP_512), Reg512> {
+                             (add SGPR_512, TTMP_512)> {
   // Requires 8 s_mov_b64 to copy
   let CopyCost = 8;
   let AllocationPriority = 18;
 }
 
 def VRegOrLds_32 : RegisterClass<"AMDGPU", [i32, f32, i16, f16, v2i16, v2f16], 32,
                                  (add VGPR_32, LDS_DIRECT_CLASS)> {
   let isAllocatable = 0;
 }
 
 def SGPR_1024 : RegisterClass<"AMDGPU", [v32i32, v32f32], 32,
-                              (add SGPR_1024Regs), Reg1024> {
+                              (add SGPR_1024Regs)> {
   let AllocationPriority = 19;
 }
 
 def SReg_1024 : RegisterClass<"AMDGPU", [v32i32, v32f32], 32,
-                              (add SGPR_1024), Reg1024> {
+                              (add SGPR_1024)> {
   let CopyCost = 16;
   let AllocationPriority = 19;
 }
 
 // Register class for all vector registers (VGPRs + Interploation Registers)
 def VReg_64 : RegisterClass<"AMDGPU", [i64, f64, v2i32, v2f32, v4f16, v4i16], 32,
-                            (add VGPR_64), Reg64> {
+                            (add VGPR_64)> {
   let Size = 64;
 
   // Requires 2 v_mov_b32 to copy
   let CopyCost = 2;
   let AllocationPriority = 2;
 }
 
-def VReg_96 : RegisterClass<"AMDGPU", [v3i32, v3f32], 32, (add VGPR_96), Reg96> {
+def VReg_96 : RegisterClass<"AMDGPU", [v3i32, v3f32], 32, (add VGPR_96)> {
   let Size = 96;
 
   // Requires 3 v_mov_b32 to copy
   let CopyCost = 3;
   let AllocationPriority = 3;
 }
 
 def VReg_128 : RegisterClass<"AMDGPU", [v4i32, v4f32, v2i64, v2f64], 32,
-                             (add VGPR_128), Reg128> {
+                             (add VGPR_128)> {
   let Size = 128;
 
   // Requires 4 v_mov_b32 to copy
   let CopyCost = 4;
   let AllocationPriority = 4;
 }
 
 def VReg_160 : RegisterClass<"AMDGPU", [v5i32, v5f32], 32,
-                             (add VGPR_160), Reg160> {
+                             (add VGPR_160)> {
   let Size = 160;
 
   // Requires 5 v_mov_b32 to copy
   let CopyCost = 5;
   let AllocationPriority = 5;
 }
 
 def VReg_256 : RegisterClass<"AMDGPU", [v8i32, v8f32], 32,
-                             (add VGPR_256), Reg256> {
+                             (add VGPR_256)> {
   let Size = 256;
   let CopyCost = 8;
   let AllocationPriority = 6;
 }
 
 def VReg_512 : RegisterClass<"AMDGPU", [v16i32, v16f32], 32,
-                             (add VGPR_512), Reg512> {
+                             (add VGPR_512)> {
   let Size = 512;
   let CopyCost = 16;
   let AllocationPriority = 7;
 }
 
 def VReg_1024 : RegisterClass<"AMDGPU", [v32i32, v32f32], 32,
-                              (add VGPR_1024), Reg1024> {
+                              (add VGPR_1024)> {
   let Size = 1024;
   let CopyCost = 32;
   let AllocationPriority = 8;
 }
 
 def AReg_64 : RegisterClass<"AMDGPU", [i64, f64, v2i32, v2f32, v4f16, v4i16], 32,
-                            (add AGPR_64), Reg64> {
+                            (add AGPR_64)> {
   let Size = 64;
 
   let CopyCost = 5;
   let AllocationPriority = 2;
 }
 
 def AReg_128 : RegisterClass<"AMDGPU", [v4i32, v4f32, v2i64, v2f64], 32,
-                             (add AGPR_128), Reg128> {
+                             (add AGPR_128)> {
   let Size = 128;
 
   // Requires 4 v_accvgpr_write and 4 v_accvgpr_read to copy + burn 1 vgpr
   let CopyCost = 9;
   let AllocationPriority = 4;
 }
 
 def AReg_512 : RegisterClass<"AMDGPU", [v16i32, v16f32], 32,
-                             (add AGPR_512), Reg512> {
+                             (add AGPR_512)> {
   let Size = 512;
   let CopyCost = 33;
   let AllocationPriority = 7;
 }
 
 def AReg_1024 : RegisterClass<"AMDGPU", [v32i32, v32f32], 32,
-                              (add AGPR_1024), Reg1024> {
+                              (add AGPR_1024)> {
   let Size = 1024;
   let CopyCost = 65;
   let AllocationPriority = 8;
 }
 
 def VReg_1 : RegisterClass<"AMDGPU", [i1], 32, (add VGPR_32)> {
   let Size = 32;
 }
 
 def VS_32 : RegisterClass<"AMDGPU", [i32, f32, i16, f16, v2i16, v2f16], 32,
                           (add VGPR_32, SReg_32, LDS_DIRECT_CLASS)> {
   let isAllocatable = 0;
 }
 
-def VS_64 : RegisterClass<"AMDGPU", [i64, f64], 32, (add VReg_64, SReg_64),
-                          Reg64> {
+def VS_64 : RegisterClass<"AMDGPU", [i64, f64], 32, (add VReg_64, SReg_64)> {
   let isAllocatable = 0;
 }
 
 def AV_32 : RegisterClass<"AMDGPU", [i32, f32, i16, f16, v2i16, v2f16], 32,
                           (add AGPR_32, VGPR_32)> {
   let isAllocatable = 0;
 }
 
 def AV_64 : RegisterClass<"AMDGPU", [i64, f64, v4f16], 32,
-                          (add AReg_64, VReg_64), Reg64> {
+                          (add AReg_64, VReg_64)> {
   let isAllocatable = 0;
 }
 
 //===----------------------------------------------------------------------===//
 //  Register operands
 //===----------------------------------------------------------------------===//
 
 class RegImmMatcher<string name> : AsmOperandClass {

utils/TableGen/CodeGenRegisters.cpp

void expand(SetTheory &ST, Record *Def, SetTheory::RecSet &Elts) override {
     }
 
     if (Length == 0)
       return;
 
     // Precompute some types.
     Record *RegisterCl = Def->getRecords().getClass("Register");
     RecTy *RegisterRecTy = RecordRecTy::get(RegisterCl);
-    StringInit *BlankName = StringInit::get("");
+    std::vector<StringRef> RegNames =
+      Def->getValueAsListOfStrings("RegAsmNames");
 
     // Zip them up.
     for (unsigned n = 0; n != Length; ++n) {
       std::string Name;
       Record *Proto = Lists[0][n];
       std::vector<Init*> Tuple;
       unsigned CostPerUse = 0;
       for (unsigned i = 0; i != Dim; ++i) {
         Record *Reg = Lists[i][n];
         if (i) Name += '_';
         Name += Reg->getName();
         Tuple.push_back(DefInit::get(Reg));
         CostPerUse = std::max(CostPerUse,
                               unsigned(Reg->getValueAsInt("CostPerUse")));
       }
 
+      StringInit *AsmName = StringInit::get("");
+      if (!RegNames.empty()) {
+        if (RegNames.size() <= n)
+          PrintFatalError(Def->getLoc(),
+                          "Register tuple definition missing name for '" +
+                            Name + "'.");
+        AsmName = StringInit::get(RegNames[n]);
+      }
+
       // Create a new Record representing the synthesized register. This record
       // is only for consumption by CodeGenRegister, it is not added to the
       // RecordKeeper.
       SynthDefs.emplace_back(
           llvm::make_unique<Record>(Name, Def->getLoc(), Def->getRecords()));
       Record *NewReg = SynthDefs.back().get();
       Elts.insert(NewReg);
 
for (unsigned n = 0; n != Length; ++n) {
           continue;
 
         StringRef Field = RV.getName();
 
         // Replace the sub-register list with Tuple.
         if (Field == "SubRegs")
           RV.setValue(ListInit::get(Tuple, RegisterRecTy));
 
-        // Provide a blank AsmName. MC hacks are required anyway.
         if (Field == "AsmName")
-          RV.setValue(BlankName);
+          RV.setValue(AsmName);
 
         // CostPerUse is aggregated from all Tuple members.
         if (Field == "CostPerUse")
           RV.setValue(IntInit::get(CostPerUse));
 
         // Composite registers are always covered by sub-registers.
         if (Field == "CoveredBySubRegs")
           RV.setValue(BitInit::get(true));