[llvm] [Xtensa] Implement Code Density Option. (PR #119639)

[Andrei Safronov via llvm-commits]

================
@@ -243,9 +263,37 @@ static DecodeStatus decodeMem32Operand(MCInst &Inst, uint64_t Imm,
   return MCDisassembler::Success;
 }
 
+static DecodeStatus decodeMem32nOperand(MCInst &Inst, uint64_t Imm,
+                                        int64_t Address, const void *Decoder) {
+  assert(isUInt<8>(Imm) && "Invalid immediate");
+  DecodeARRegisterClass(Inst, Imm & 0xf, Address, Decoder);
+  Inst.addOperand(MCOperand::createImm((Imm >> 2) & 0x3c));
+  return MCDisassembler::Success;
+}
+
+/// Read two bytes from the ArrayRef and return 16 bit data sorted
+/// according to the given endianness.
+static DecodeStatus readInstruction16(ArrayRef<uint8_t> Bytes, uint64_t Address,
+                                      uint64_t &Size, uint64_t &Insn,
+                                      bool IsLittleEndian) {
+  // We want to read exactly 2 Bytes of data.
+  if (Bytes.size() < 2) {
+    Size = 0;
+    return MCDisassembler::Fail;
+  }
+
+  if (!IsLittleEndian) {
+    llvm_unreachable("Big-endian mode currently is not supported!");
----------------
andreisfr wrote:

I changed "llvm_unreachable" in readInstruction16 to "report_fatal_error", like it is done in readInstruction24.
Hypothetically, the Xtensa architecture could have a big endian implementation, but in practice it is hard to find such hardware (at least I don't see such an example). So we don't expect such a triple variant at this time.

https://github.com/llvm/llvm-project/pull/119639