[llvm] [TableGen][DecoderEmitter] Fix broken AdditionalEncoding support (PR #155057)

[Sergei Barannikov via llvm-commits]

https://github.com/s-barannikov created https://github.com/llvm/llvm-project/pull/155057

We didn't have tests for AdditionalEncoding and none of the in-tree targets use this functionality, so I inadvertently broke it in #154288.

>From 90004c1fb17af6f5813c7a5adf861aae9db2ff67 Mon Sep 17 00:00:00 2001
From: Sergei Barannikov <barannikov88 at gmail.com>
Date: Sat, 23 Aug 2025 05:05:02 +0300
Subject: [PATCH] [TableGen][DecoderEmitter] Fix broken AdditionalEncoding
 support

We didn't have tests for AdditionalEncoding and none of the in-tree
targets use this functionality, so I inadvertently broke it in #154288.
---
 .../additional-encoding.td                    | 67 +++++++++++++++++++
 llvm/utils/TableGen/DecoderEmitter.cpp        |  2 +
 2 files changed, 69 insertions(+)
 create mode 100644 llvm/test/TableGen/FixedLenDecoderEmitter/additional-encoding.td

diff --git a/llvm/test/TableGen/FixedLenDecoderEmitter/additional-encoding.td b/llvm/test/TableGen/FixedLenDecoderEmitter/additional-encoding.td
new file mode 100644
index 0000000000000..192cb3c3f5858
--- /dev/null
+++ b/llvm/test/TableGen/FixedLenDecoderEmitter/additional-encoding.td
@@ -0,0 +1,67 @@
+// RUN: llvm-tblgen -gen-disassembler -I %p/../../../include %s | FileCheck %s
+
+include "llvm/Target/Target.td"
+
+class Enc {
+  int Size = 2;
+  bits<16> Inst;
+}
+
+class EncSHIFT<bits<2> opc> : Enc {
+  bits<6> shamt;
+  let Inst{15...14} = {0, 0};
+  let Inst{13...12} = opc;
+  let Inst{11...6} = shamt;
+}
+
+class EncNOP<bits<2> opc> : Enc {
+  let Inst{15...14} = {0, 0};
+  let Inst{13...12} = opc;
+  let Inst{11...6} = {0, 0, 0, 0, 0, 0};
+}
+
+def ShAmtOp : Operand<i32> {
+  let DecoderMethod = "decodeShAmt";
+  let hasCompleteDecoder = false;
+}
+
+class I<dag out_ops, dag in_ops> : Instruction {
+  let InOperandList = in_ops;
+  let OutOperandList = out_ops;
+}
+
+// CHECK:      /* 0 */  MCD::OPC_ExtractField, 12, 4,                // Inst{15-12} ...
+// CHECK-NEXT: /* 3 */  MCD::OPC_FilterValue, 0, 14, 0,              // Skip to: 21
+// CHECK-NEXT: /* 7 */  MCD::OPC_CheckField, 6, 6, 0, 4, 0,          // Skip to: 17
+// CHECK-NEXT: /* 13 */ MCD::OPC_Decode, {{[0-9]+}}, 2, 0,           // Opcode: {{.*}}:NOP
+// CHECK-NEXT: /* 17 */ MCD::OPC_TryDecodeOrFail, {{[0-9]+}}, 2, 1,
+// CHECK-NEXT: /* 21 */ MCD::OPC_FilterValue, 1, 14, 0,              // Skip to: 39
+// CHECK-NEXT: /* 25 */ MCD::OPC_CheckField, 6, 6, 0, 4, 0,          // Skip to: 35
+// CHECK-NEXT: /* 31 */ MCD::OPC_Decode, {{[0-9]+}}, 2, 0,           // Opcode: {{.*}}:NOP
+// CHECK-NEXT: /* 35 */ MCD::OPC_TryDecodeOrFail, {{[0-9]+}}, 2, 1,
+// CHECK-NEXT: /* 39 */ MCD::OPC_FilterValue, 2, 14, 0,              // Skip to: 57
+// CHECK-NEXT: /* 43 */ MCD::OPC_CheckField, 6, 6, 0, 4, 0,          // Skip to: 53
+// CHECK-NEXT: /* 49 */ MCD::OPC_Decode, {{[0-9]+}}, 2, 0,           // Opcode: {{.*}}:NOP
+// CHECK-NEXT: /* 53 */ MCD::OPC_TryDecodeOrFail, {{[0-9]+}}, 2, 1,
+// CHECK-NEXT: /* 57 */ MCD::OPC_FilterValueOrFail, 3,
+// CHECK-NEXT: /* 59 */ MCD::OPC_CheckField, 6, 6, 0, 4, 0,          // Skip to: 69
+// CHECK-NEXT: /* 65 */ MCD::OPC_Decode, {{[0-9]+}}, 2, 0,           // Opcode: {{.*}}:NOP
+// CHECK-NEXT: /* 69 */ MCD::OPC_TryDecodeOrFail, {{[0-9]+}}, 2, 1,
+// CHECK-NEXT: /* 73 */ MCD::OPC_Fail,
+
+class SHIFT<bits<2> opc> : I<(outs), (ins ShAmtOp:$shamt)>, EncSHIFT<opc>;
+def SHIFT0 : SHIFT<0>;
+def SHIFT1 : SHIFT<1>;
+def SHIFT2 : SHIFT<2>;
+def SHIFT3 : SHIFT<3>;
+
+def NOP : I<(outs), (ins)>, EncNOP<0>;
+def : AdditionalEncoding<NOP>, EncNOP<1>;
+def : AdditionalEncoding<NOP>, EncNOP<2>;
+def : AdditionalEncoding<NOP>, EncNOP<3>;
+
+def II : InstrInfo;
+
+def MyTarget : Target {
+  let InstructionSet = II;
+}
diff --git a/llvm/utils/TableGen/DecoderEmitter.cpp b/llvm/utils/TableGen/DecoderEmitter.cpp
index ece92c7b76ec0..25f1d824ffcd9 100644
--- a/llvm/utils/TableGen/DecoderEmitter.cpp
+++ b/llvm/utils/TableGen/DecoderEmitter.cpp
@@ -2495,7 +2495,9 @@ void DecoderEmitter::parseInstructionEncodings() {
       ++NumEncodingsOmitted;
       continue;
     }
+    unsigned EncodingID = Encodings.size();
     Encodings.emplace_back(EncodingDef, &Target.getInstruction(InstDef));
+    EncodingIDsByHwMode[DefaultMode].push_back(EncodingID);
   }
 
   // Do some statistics.