Update coredistools (#96286)

* Update coredistools

Update to the 1.2.0 version of the coredistools package.

Also, implement a basic late disassembler for RyuJIT based on
coredistools (previously, this depended on closed-source msvcdis).

Various changes:
1. Remove unused AddressMap in SPMI. Removed a bunch of cases in
the NearDiffer that were using it.
2. Change arm64 emitter unit tests to not require `verbose`
3. Always build in the late disassembler under DEBUG. It's only
used when `JitLateDisasm` is set.
4. Implement new arm32 callback mechanism for movw/movt handling
in the NearDiffer.

* Allow fallback to coredistools 1.1.0

* Fixes

* Formatting

* Fixes

1. Fix Risc-V build: only build LATE_DISASM for platforms where
coredistool support exists.
2. Defer load and initialization of coredistools.dll library for
LATE_DISASM until we actually need it.
3. For arm64, in late disasm, if there is a failure, try to recover
by skipping one 4-byte instruction.

* Be smarter about initializing and uninitializing coredistools

In particular, only load the coredistools library once, not
once per function to disassemble.

* Update to 1.4.0

src/coreclr/inc/coredistools.h

@@ -1,11 +1,11 @@
 // Licensed to the .NET Foundation under one or more agreements.
 // The .NET Foundation licenses this file to you under the MIT license.
 
-//===--------- coredistools.h - Dissassembly tools for CoreClr ------------===//
+//===--------- coredistools.h - Disassembly tools for CoreClr ------------===//
 //
-//  Core Disassembly Tools API Version 1.0.1-prerelease
+//  Core Disassembly Tools API Version 1.4.0
 //  Disassembly tools required by CoreCLR for utilities like
-//  GCStress and SuperPMI
+//  GCStress, SuperPMI, and R2RDump.
 //===----------------------------------------------------------------------===//
 
 #if !defined(_COREDISTOOLS_H_)
@@ -41,7 +41,8 @@ enum TargetArch {
     Target_X86,
     Target_X64,
     Target_Thumb,
-    Target_Arm64
+    Target_Arm64,
+    Target_LoongArch64
 };
 
 struct CorDisasm;
@@ -60,14 +61,28 @@ struct PrintControl {
 // The type of a custom function provided by the user to determine
 // if two offsets are considered equivalent wrt diffing code blocks.
 // Offset1 and Offset2 are the two offsets to be compared.
-// BlockOffset is the offest of the instructions (that contain Offset1
+// BlockOffset is the offset of the instructions (that contain Offset1
 // and Offset2) from the beginning of their respective code blocks.
 // InstructionLength is the length of the current instruction being
 // compared for equivalency.
 typedef bool(__cdecl *OffsetComparator)(const void *UserData, size_t BlockOffset,
     size_t InstructionLength, uint64_t Offset1,
     uint64_t Offset2);
 
+// If an OffsetMunger function is defined, it is called before the OffsetComparator.
+// If it returns `true` then:
+// 1. the instructions are considered equivalent
+// 2. the offsets have been decoded and "munged" (changed), and
+//    *Offset1 and *Offset2 are set to the values to use.
+// 3. *SkipInstructions1 instructions in code stream 1 are skipped
+// 4. *SkipInstructions2 instructions in code stream 2 are skipped
+//
+// This is typically used on arm32 to treat "movw/movt" as a single instruction
+// generating a single constant. Similarly, for arm64 mov/movk/movk/movk sequences.
+typedef bool(__cdecl *OffsetMunger)(const void *UserData, size_t BlockOffset,
+    size_t InstructionLength, uint64_t* Offset1, uint64_t* Offset2,
+    uint32_t* SkipInstructions1, uint32_t* SkipInstructions2);
+
 // The Export/Import definitions for CoreDistools library are defined below.
 // A typedef for each interface function's type is defined in order to aid
 // the importer.
@@ -76,6 +91,10 @@ typedef bool(__cdecl *OffsetComparator)(const void *UserData, size_t BlockOffset
 typedef CorDisasm * __cdecl InitDisasm_t(enum TargetArch Target);
 DllIface InitDisasm_t InitDisasm;
 
+// Initialize the disassembler, using buffered print controls
+typedef CorDisasm * __cdecl InitBufferedDisasm_t(enum TargetArch Target);
+DllIface InitBufferedDisasm_t InitBufferedDisasm;
+
 // Initialize the disassembler using custom print controls
 typedef CorDisasm * __cdecl NewDisasm_t(enum TargetArch Target,
     const PrintControl *PControl);
@@ -85,6 +104,28 @@ DllIface NewDisasm_t NewDisasm;
 typedef void __cdecl FinishDisasm_t(const CorDisasm *Disasm);
 DllIface FinishDisasm_t FinishDisasm;
 
+// Initialize a code differ using buffered output.
+typedef CorDisasm * __cdecl InitBufferedDiffer_t(enum TargetArch Target,
+                               const OffsetComparator Comparator);
+DllIface InitBufferedDiffer_t InitBufferedDiffer;
+
+// Initialize the Code Differ
+typedef CorAsmDiff * __cdecl NewDiffer_t(enum TargetArch Target,
+    const PrintControl *PControl,
+    const OffsetComparator Comparator);
+DllIface NewDiffer_t NewDiffer;
+
+// Initialize the Code Differ, with an offset munger.
+typedef CorAsmDiff * __cdecl NewDiffer2_t(enum TargetArch Target,
+    const PrintControl *PControl,
+    const OffsetComparator Comparator,
+    const OffsetMunger Munger);
+DllIface NewDiffer2_t NewDiffer2;
+
+// Delete the Code Differ
+typedef void __cdecl FinishDiff_t(const CorAsmDiff *AsmDiff);
+DllIface FinishDiff_t FinishDiff;
+
 // DisasmInstruction -- Disassemble one instruction
 // Arguments:
 // Disasm -- The Disassembler
@@ -100,22 +141,27 @@ typedef size_t __cdecl DisasmInstruction_t(const CorDisasm *Disasm,
     const uint8_t *Bytes, size_t Maxlength);
 DllIface DisasmInstruction_t DisasmInstruction;
 
-// Initialize the Code Differ
-typedef CorAsmDiff * __cdecl NewDiffer_t(enum TargetArch Target,
-    const PrintControl *PControl,
-    const OffsetComparator Comparator);
-DllIface NewDiffer_t NewDiffer;
-
-// Delete the Code Differ
-typedef void __cdecl FinishDiff_t(const CorAsmDiff *AsmDiff);
-DllIface FinishDiff_t FinishDiff;
+// DumpInstruction -- Disassemble one instruction and output it
+// Arguments:
+// Disasm -- The Disassembler
+// Address -- The address at which the bytes of the instruction
+//            are intended to execute
+// Bytes -- Pointer to the actual bytes which need to be disassembled
+// MaxLength -- Number of bytes available in Bytes buffer
+// Returns:
+//   -- The Size of the disassembled instruction
+//   -- Zero on failure
+typedef size_t __cdecl DumpInstruction_t(const CorDisasm *Disasm,
+	const uint8_t *Address, const uint8_t *Bytes,
+	size_t Maxlength);
+DllIface DumpInstruction_t DumpInstruction;
 
 // NearDiffCodeBlocks -- Compare two code blocks for semantic
 //                       equivalence
 // Arguments:
 // AsmDiff -- The Asm-differ
 // UserData -- Any data the user wishes to pass through into
-//             the OffsetComparator
+//             the OffsetComparator/OffsetMunger
 // Address1 -- Address at which first block will execute
 // Bytes1 -- Pointer to the actual bytes of the first block
 // Size1 -- The size of the first block
@@ -145,4 +191,12 @@ typedef void __cdecl DumpDiffBlocks_t(const CorAsmDiff *AsmDiff,
     const uint8_t *Bytes2, size_t Size2);
 DllIface DumpDiffBlocks_t DumpDiffBlocks;
 
+// Get a pointer to the buffered output buffer.
+typedef const char* __cdecl GetOutputBuffer_t();
+DllIface GetOutputBuffer_t GetOutputBuffer;
+
+// Clear the buffered output buffer.
+typedef void __cdecl ClearOutputBuffer_t();
+DllIface ClearOutputBuffer_t ClearOutputBuffer;
+
 #endif // !defined(_COREDISTOOLS_H_)

src/coreclr/jit/codegenarm64test.cpp

@@ -24,7 +24,6 @@ XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
  */
 void CodeGen::genArm64EmitterUnitTestsGeneral()
 {
-    assert(verbose);
     emitter* theEmitter = GetEmitter();
 
     //
@@ -1824,7 +1823,6 @@ void CodeGen::genArm64EmitterUnitTestsGeneral()
  */
 void CodeGen::genArm64EmitterUnitTestsAdvSimd()
 {
-    assert(verbose);
     emitter* theEmitter = GetEmitter();
 
     ////////////////////////////////////////////////////////////////////////////////
@@ -4557,7 +4555,6 @@ void CodeGen::genArm64EmitterUnitTestsAdvSimd()
  */
 void CodeGen::genArm64EmitterUnitTestsSve()
 {
-    assert(verbose);
     emitter* theEmitter = GetEmitter();
 
     //

src/coreclr/jit/codegenlinear.cpp

@@ -2674,8 +2674,8 @@ void CodeGen::genCodeForSetcc(GenTreeCC* setcc)
 /*****************************************************************************
  * Unit testing of the emitter: If JitDumpEmitUnitTests is set, generate
  * a bunch of instructions, then:
- * Use DOTNET_JitLateDisasm=* to see if the late disassembler thinks the instructions as the same as we do.
- * Or, use DOTNET_JitRawHexCode and DOTNET_JitRawHexCodeFile and dissassemble the output file.
+ * Use DOTNET_JitLateDisasm=* to see if the late disassembler thinks the instructions are the same as we do.
+ * Or, use DOTNET_JitRawHexCode and DOTNET_JitRawHexCodeFile and disassemble the output file.
  *
  * Possible values for JitDumpEmitUnitTests:
  * Amd64: all, sse2
@@ -2688,7 +2688,7 @@ void CodeGen::genEmitterUnitTests()
 {
     const WCHAR* unitTestSection = JitConfig.JitDumpEmitUnitTests();
 
-    if ((!verbose) || (unitTestSection == nullptr))
+    if (unitTestSection == nullptr)
     {
         return;
     }

src/coreclr/jit/compiler.cpp

@@ -1994,6 +1994,12 @@ void Compiler::compInit(ArenaAllocator*       pAlloc,
 
 void Compiler::compDone()
 {
+#ifdef LATE_DISASM
+    if (codeGen != nullptr)
+    {
+        codeGen->getDisAssembler().disDone();
+    }
+#endif // LATE_DISASM
 }
 
 void* Compiler::compGetHelperFtn(CorInfoHelpFunc ftnNum,        /* IN  */