[RISCV] Remove post-decoding instruction adjustments (#156360)

Some instructions implicitly define/use X2 (SP) register, but instead of
being present in the Defs/Uses lists, it is sometimes modeled as an
explicit operand with SP register class.
Since the operand is not encoded into the instruction, it cannot be
disassembled, and we have `RISCVDisassembler::addSPOperands()` that
addresses the issue by mutating the (incompletely) decoded instruction.

This change makes the operand decodable by adding `bits<0>` field for
that operand to relevant instruction encodings and removes
`RISCVDisassembler::addSPOperands()`.

Author: s-barannikov

llvm/lib/Target/RISCV/CMakeLists.txt

@@ -8,8 +8,7 @@ tablegen(LLVM RISCVGenCompressInstEmitter.inc -gen-compress-inst-emitter)
 tablegen(LLVM RISCVGenMacroFusion.inc -gen-macro-fusion-pred)
 tablegen(LLVM RISCVGenDAGISel.inc -gen-dag-isel)
 tablegen(LLVM RISCVGenDisassemblerTables.inc -gen-disassembler
-              --specialize-decoders-per-bitwidth
-              -ignore-non-decodable-operands)
+              --specialize-decoders-per-bitwidth)
 tablegen(LLVM RISCVGenInstrInfo.inc -gen-instr-info)
 tablegen(LLVM RISCVGenMCCodeEmitter.inc -gen-emitter)
 tablegen(LLVM RISCVGenMCPseudoLowering.inc -gen-pseudo-lowering)

llvm/lib/Target/RISCV/Disassembler/RISCVDisassembler.cpp

@@ -46,8 +46,6 @@ class RISCVDisassembler : public MCDisassembler {
                               raw_ostream &CStream) const override;
 
 private:
-  void addSPOperands(MCInst &MI) const;
-
   DecodeStatus getInstruction48(MCInst &Instr, uint64_t &Size,
                                 ArrayRef<uint8_t> Bytes, uint64_t Address,
                                 raw_ostream &CStream) const;
@@ -196,6 +194,12 @@ static DecodeStatus DecodeFPR128RegisterClass(MCInst &Inst, uint32_t RegNo,
   return MCDisassembler::Success;
 }
 
+static DecodeStatus DecodeSPRegisterClass(MCInst &Inst,
+                                          const MCDisassembler *Decoder) {
+  Inst.addOperand(MCOperand::createReg(RISCV::X2));
+  return MCDisassembler::Success;
+}
+
 static DecodeStatus DecodeGPRNoX0RegisterClass(MCInst &Inst, uint32_t RegNo,
                                                uint64_t Address,
                                                const MCDisassembler *Decoder) {
@@ -600,15 +604,6 @@ static DecodeStatus decodeXTHeadMemPair(MCInst &Inst, uint32_t Insn,
 
 #include "RISCVGenDisassemblerTables.inc"
 
-// Add implied SP operand for C.*SP compressed instructions. The SP operand
-// isn't explicitly encoded in the instruction.
-void RISCVDisassembler::addSPOperands(MCInst &MI) const {
-  const MCInstrDesc &MCID = MCII->get(MI.getOpcode());
-  for (unsigned i = 0; i < MCID.getNumOperands(); i++)
-    if (MCID.operands()[i].RegClass == RISCV::SPRegClassID)
-      MI.insert(MI.begin() + i, MCOperand::createReg(RISCV::X2));
-}
-
 namespace {
 
 struct DecoderListEntry {
@@ -774,12 +769,8 @@ DecodeStatus RISCVDisassembler::getInstruction16(MCInst &MI, uint64_t &Size,
     LLVM_DEBUG(dbgs() << "Trying " << Entry.Desc << " table:\n");
     DecodeStatus Result =
         decodeInstruction(Entry.Table, MI, Insn, Address, this, STI);
-    if (Result == MCDisassembler::Fail)
-      continue;
-
-    addSPOperands(MI);
-
-    return Result;
+    if (Result != MCDisassembler::Fail)
+      return Result;
   }
 
   return MCDisassembler::Fail;

llvm/lib/Target/RISCV/RISCVInstrFormatsC.td

@@ -54,7 +54,6 @@ class RVInst16CSS<bits<3> funct3, bits<2> opcode, dag outs, dag ins,
     : RVInst16<outs, ins, opcodestr, argstr, [], InstFormatCSS> {
   bits<10> imm;
   bits<5> rs2;
-  bits<5> rs1;
 
   let Inst{15-13} = funct3;
   let Inst{12-7} = imm{5-0};

llvm/lib/Target/RISCV/RISCVInstrInfoC.td

@@ -230,13 +230,17 @@ let hasSideEffects = 0, mayLoad = 1, mayStore = 0 in
 class CStackLoad<bits<3> funct3, string OpcodeStr,
                  DAGOperand cls, DAGOperand opnd>
     : RVInst16CI<funct3, 0b10, (outs cls:$rd), (ins SPMem:$rs1, opnd:$imm),
-                 OpcodeStr, "$rd, ${imm}(${rs1})">;
+                 OpcodeStr, "$rd, ${imm}(${rs1})"> {
+  bits<0> rs1;
+}
 
 let hasSideEffects = 0, mayLoad = 0, mayStore = 1 in
 class CStackStore<bits<3> funct3, string OpcodeStr,
                   DAGOperand cls, DAGOperand opnd>
     : RVInst16CSS<funct3, 0b10, (outs), (ins cls:$rs2, SPMem:$rs1, opnd:$imm),
-                  OpcodeStr, "$rs2, ${imm}(${rs1})">;
+                  OpcodeStr, "$rs2, ${imm}(${rs1})"> {
+  bits<0> rs1;
+}
 
 let hasSideEffects = 0, mayLoad = 1, mayStore = 0 in
 class CLoad_ri<bits<3> funct3, string OpcodeStr,

llvm/lib/Target/RISCV/RISCVInstrInfoXwch.td

@@ -106,6 +106,7 @@ def QK_C_LBUSP : QKStackInst<0b00, (outs GPRC:$rd_rs2),
                              (ins SPMem:$rs1, uimm4:$imm),
                              "qk.c.lbusp", "$rd_rs2, ${imm}(${rs1})">,
                  Sched<[WriteLDB, ReadMemBase]> {
+  bits<0> rs1;
   bits<4> imm;
   let Inst{10-7} = imm;
 }
@@ -115,6 +116,7 @@ def QK_C_SBSP : QKStackInst<0b10, (outs),
                                  uimm4:$imm),
                             "qk.c.sbsp", "$rd_rs2, ${imm}(${rs1})">,
                 Sched<[WriteSTB, ReadStoreData, ReadMemBase]> {
+  bits<0> rs1;
   bits<4> imm;
   let Inst{10-7} = imm;
 }
@@ -124,6 +126,7 @@ def QK_C_LHUSP : QKStackInst<0b01, (outs GPRC:$rd_rs2),
                              (ins SPMem:$rs1, uimm5_lsb0:$imm),
                              "qk.c.lhusp", "$rd_rs2, ${imm}(${rs1})">,
                  Sched<[WriteLDH, ReadMemBase]> {
+  bits<0> rs1;
   bits<5> imm;
   let Inst{10-8} = imm{3-1};
   let Inst{7} = imm{4};
@@ -133,6 +136,7 @@ def QK_C_SHSP : QKStackInst<0b11, (outs),
                             (ins GPRC:$rd_rs2, SPMem:$rs1, uimm5_lsb0:$imm),
                             "qk.c.shsp", "$rd_rs2, ${imm}(${rs1})">,
                 Sched<[WriteSTH, ReadStoreData, ReadMemBase]> {
+  bits<0> rs1;
   bits<5> imm;
   let Inst{10-8} = imm{3-1};
   let Inst{7} = imm{4};