[DebugInfo] Add DW_OP_LLVM_extract_bits (#93990)

This operation extracts a number of bits at a given offset and sign or
zero extends them, which is done by emitting it as a left shift followed
by a right shift.

This is being added for use in clang for C++ structured bindings of
bitfields that have offset or size that aren't a byte multiple. A new
operation is being added, instead of shifts being used directly, as it
makes correctly handling it in optimisations (which will be done in a
later patch) much easier.

Author: john-brawn-arm

llvm/docs/LangRef.rst

@@ -6308,6 +6308,15 @@ The current supported opcode vocabulary is limited:
   (``16`` and ``DW_ATE_signed`` here, respectively) to which the top of the
   expression stack is to be converted. Maps into a ``DW_OP_convert`` operation
   that references a base type constructed from the supplied values.
+- ``DW_OP_LLVM_extract_bits_sext, 16, 8,`` specifies the offset and size
+  (``16`` and ``8`` here, respectively) of bits that are to be extracted and
+  sign-extended from the value at the top of the expression stack. If the top of
+  the expression stack is a memory location then these bits are extracted from
+  the value pointed to by that memory location. Maps into a ``DW_OP_shl``
+  followed by ``DW_OP_shra``.
+- ``DW_OP_LLVM_extract_bits_zext`` behaves similarly to
+  ``DW_OP_LLVM_extract_bits_sext``, but zero-extends instead of sign-extending.
+  Maps into a ``DW_OP_shl`` followed by ``DW_OP_shr``.
 - ``DW_OP_LLVM_tag_offset, tag_offset`` specifies that a memory tag should be
   optionally applied to the pointer. The memory tag is derived from the
   given tag offset in an implementation-defined manner.

llvm/include/llvm/BinaryFormat/Dwarf.h

@@ -138,12 +138,14 @@ enum LocationAtom {
 #include "llvm/BinaryFormat/Dwarf.def"
   DW_OP_lo_user = 0xe0,
   DW_OP_hi_user = 0xff,
-  DW_OP_LLVM_fragment = 0x1000,         ///< Only used in LLVM metadata.
-  DW_OP_LLVM_convert = 0x1001,          ///< Only used in LLVM metadata.
-  DW_OP_LLVM_tag_offset = 0x1002,       ///< Only used in LLVM metadata.
-  DW_OP_LLVM_entry_value = 0x1003,      ///< Only used in LLVM metadata.
-  DW_OP_LLVM_implicit_pointer = 0x1004, ///< Only used in LLVM metadata.
-  DW_OP_LLVM_arg = 0x1005,              ///< Only used in LLVM metadata.
+  DW_OP_LLVM_fragment = 0x1000,          ///< Only used in LLVM metadata.
+  DW_OP_LLVM_convert = 0x1001,           ///< Only used in LLVM metadata.
+  DW_OP_LLVM_tag_offset = 0x1002,        ///< Only used in LLVM metadata.
+  DW_OP_LLVM_entry_value = 0x1003,       ///< Only used in LLVM metadata.
+  DW_OP_LLVM_implicit_pointer = 0x1004,  ///< Only used in LLVM metadata.
+  DW_OP_LLVM_arg = 0x1005,               ///< Only used in LLVM metadata.
+  DW_OP_LLVM_extract_bits_sext = 0x1006, ///< Only used in LLVM metadata.
+  DW_OP_LLVM_extract_bits_zext = 0x1007, ///< Only used in LLVM metadata.
 };
 
 enum LlvmUserLocationAtom {

llvm/lib/BinaryFormat/Dwarf.cpp

@@ -155,6 +155,10 @@ StringRef llvm::dwarf::OperationEncodingString(unsigned Encoding) {
     return "DW_OP_LLVM_implicit_pointer";
   case DW_OP_LLVM_arg:
     return "DW_OP_LLVM_arg";
+  case DW_OP_LLVM_extract_bits_sext:
+    return "DW_OP_LLVM_extract_bits_sext";
+  case DW_OP_LLVM_extract_bits_zext:
+    return "DW_OP_LLVM_extract_bits_zext";
   }
 }
 
@@ -169,6 +173,8 @@ unsigned llvm::dwarf::getOperationEncoding(StringRef OperationEncodingString) {
       .Case("DW_OP_LLVM_entry_value", DW_OP_LLVM_entry_value)
       .Case("DW_OP_LLVM_implicit_pointer", DW_OP_LLVM_implicit_pointer)
       .Case("DW_OP_LLVM_arg", DW_OP_LLVM_arg)
+      .Case("DW_OP_LLVM_extract_bits_sext", DW_OP_LLVM_extract_bits_sext)
+      .Case("DW_OP_LLVM_extract_bits_zext", DW_OP_LLVM_extract_bits_zext)
       .Default(0);
 }
 

llvm/lib/CodeGen/AsmPrinter/DwarfExpression.cpp

@@ -18,6 +18,7 @@
 #include "llvm/CodeGen/Register.h"
 #include "llvm/CodeGen/TargetRegisterInfo.h"
 #include "llvm/IR/DataLayout.h"
+#include "llvm/MC/MCAsmInfo.h"
 #include "llvm/Support/ErrorHandling.h"
 #include <algorithm>
 
@@ -546,6 +547,37 @@ bool DwarfExpression::addExpression(
       LocationKind = Unknown;
       return true;
     }
+    case dwarf::DW_OP_LLVM_extract_bits_sext:
+    case dwarf::DW_OP_LLVM_extract_bits_zext: {
+      unsigned SizeInBits = Op->getArg(1);
+      unsigned BitOffset = Op->getArg(0);
+
+      // If we have a memory location then dereference to get the value
+      if (isMemoryLocation())
+        emitOp(dwarf::DW_OP_deref);
+
+      // Extract the bits by a shift left (to shift out the bits after what we
+      // want to extract) followed by shift right (to shift the bits to position
+      // 0 and also sign/zero extend). These operations are done in the DWARF
+      // "generic type" whose size is the size of a pointer.
+      unsigned PtrSizeInBytes = CU.getAsmPrinter()->MAI->getCodePointerSize();
+      unsigned LeftShift = PtrSizeInBytes * 8 - (SizeInBits + BitOffset);
+      unsigned RightShift = LeftShift + BitOffset;
+      if (LeftShift) {
+        emitOp(dwarf::DW_OP_constu);
+        emitUnsigned(LeftShift);
+        emitOp(dwarf::DW_OP_shl);
+      }
+      emitOp(dwarf::DW_OP_constu);
+      emitUnsigned(RightShift);
+      emitOp(OpNum == dwarf::DW_OP_LLVM_extract_bits_sext ? dwarf::DW_OP_shra
+                                                          : dwarf::DW_OP_shr);
+
+      // The value is now at the top of the stack, so set the location to
+      // implicit so that we get a stack_value at the end.
+      LocationKind = Implicit;
+      break;
+    }
     case dwarf::DW_OP_plus_uconst:
       assert(!isRegisterLocation());
       emitOp(dwarf::DW_OP_plus_uconst);

llvm/lib/IR/DebugInfoMetadata.cpp

@@ -1404,6 +1404,8 @@ unsigned DIExpression::ExprOperand::getSize() const {
   switch (Op) {
   case dwarf::DW_OP_LLVM_convert:
   case dwarf::DW_OP_LLVM_fragment:
+  case dwarf::DW_OP_LLVM_extract_bits_sext:
+  case dwarf::DW_OP_LLVM_extract_bits_zext:
   case dwarf::DW_OP_bregx:
     return 3;
   case dwarf::DW_OP_constu:
@@ -1474,6 +1476,8 @@ bool DIExpression::isValid() const {
     case dwarf::DW_OP_LLVM_convert:
     case dwarf::DW_OP_LLVM_arg:
     case dwarf::DW_OP_LLVM_tag_offset:
+    case dwarf::DW_OP_LLVM_extract_bits_sext:
+    case dwarf::DW_OP_LLVM_extract_bits_zext:
     case dwarf::DW_OP_constu:
     case dwarf::DW_OP_plus_uconst:
     case dwarf::DW_OP_plus:

llvm/test/DebugInfo/X86/DW_OP_LLVM_extract_bits.ll

@@ -0,0 +1,92 @@
+; RUN: llc -mtriple=x86_64-unknown-linux-gnu %s -o %t -filetype=obj
+; RUN: llvm-dwarfdump --debug-info %t | FileCheck %s
+
+%struct.struct_t = type { i8 }
+
+@g = dso_local global %struct.struct_t zeroinitializer, align 1, !dbg !0
+
+; CHECK-LABEL: DW_TAG_subprogram
+; CHECK: DW_AT_name ("test1")
+; CHECK: DW_TAG_variable
+; CHECK: DW_AT_location (DW_OP_fbreg -1, DW_OP_deref, DW_OP_constu 0x3d, DW_OP_shl, DW_OP_constu 0x3d, DW_OP_shr, DW_OP_stack_value)
+; CHECK: DW_AT_name ("x")
+; CHECK: DW_TAG_variable
+; CHECK: DW_AT_location (DW_OP_fbreg -1, DW_OP_deref, DW_OP_constu 0x39, DW_OP_shl, DW_OP_constu 0x3c, DW_OP_shra, DW_OP_stack_value)
+; CHECK: DW_AT_name ("y")
+
+define i32 @test1() !dbg !13 {
+entry:
+  %0 = alloca %struct.struct_t, align 1
+  tail call void @llvm.dbg.declare(metadata ptr %0, metadata !16, metadata !DIExpression(DW_OP_LLVM_extract_bits_zext, 0, 3)), !dbg !17
+  tail call void @llvm.dbg.declare(metadata ptr %0, metadata !18, metadata !DIExpression(DW_OP_LLVM_extract_bits_sext, 3, 4)), !dbg !17
+  ret i32 0, !dbg !17
+}
+
+; CHECK-LABEL: DW_TAG_subprogram
+; CHECK: DW_AT_name ("test2")
+; CHECK: DW_TAG_variable
+; CHECK: DW_AT_location (DW_OP_breg0 {{R[^+]+}}+0, DW_OP_constu 0xff, DW_OP_and, DW_OP_constu 0x3d, DW_OP_shl, DW_OP_constu 0x3d, DW_OP_shr, DW_OP_stack_value)
+; CHECK: DW_AT_name ("x")
+; CHECK: DW_TAG_variable
+; CHECK: DW_AT_location (DW_OP_breg0 {{R[^+]+}}+0, DW_OP_constu 0xff, DW_OP_and, DW_OP_constu 0x39, DW_OP_shl, DW_OP_constu 0x3c, DW_OP_shra, DW_OP_stack_value)
+; CHECK: DW_AT_name ("y")
+
+define i8 @test2() !dbg !20 {
+entry:
+  %0 = load i8, ptr @g, align 1
+  tail call void @llvm.dbg.value(metadata i8 %0, metadata !21, metadata !DIExpression(DW_OP_LLVM_extract_bits_zext, 0, 3)), !dbg !22
+  tail call void @llvm.dbg.value(metadata i8 %0, metadata !23, metadata !DIExpression(DW_OP_LLVM_extract_bits_sext, 3, 4)), !dbg !22
+  ret i8 %0, !dbg !22
+}
+
+; CHECK-LABEL: DW_TAG_subprogram
+; CHECK: DW_AT_name ("test3")
+; CHECK: DW_TAG_variable
+; CHECK: DW_AT_location (DW_OP_breg0 {{R[^+]+}}+0, DW_OP_constu 0x3f, DW_OP_shr, DW_OP_stack_value)
+; CHECK: DW_AT_name ("x")
+; CHECK: DW_TAG_variable
+; CHECK: DW_AT_location (DW_OP_breg0 {{R[^+]+}}+0, DW_OP_constu 0x3f, DW_OP_shra, DW_OP_stack_value)
+; CHECK: DW_AT_name ("y")
+
+define i64 @test3(ptr %p) !dbg !24 {
+entry:
+  %0 = load i64, ptr %p, align 8
+  tail call void @llvm.dbg.value(metadata i64 %0, metadata !25, metadata !DIExpression(DW_OP_LLVM_extract_bits_zext, 63, 1)), !dbg !26
+  tail call void @llvm.dbg.value(metadata i64 %0, metadata !27, metadata !DIExpression(DW_OP_LLVM_extract_bits_sext, 63, 1)), !dbg !26
+  ret i64 %0, !dbg !26
+}
+
+declare void @llvm.dbg.declare(metadata, metadata, metadata)
+declare void @llvm.dbg.value(metadata, metadata, metadata)
+
+!llvm.dbg.cu = !{!2}
+!llvm.module.flags = !{!11, !12}
+
+!0 = !DIGlobalVariableExpression(var: !1, expr: !DIExpression())
+!1 = distinct !DIGlobalVariable(name: "g", scope: !2, file: !3, type: !5, isLocal: false, isDefinition: true)
+!2 = distinct !DICompileUnit(language: DW_LANG_C_plus_plus_14, file: !3, isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug, globals: !4, splitDebugInlining: false, nameTableKind: None)
+!3 = !DIFile(filename: "DW_OP_bit_piece.cpp", directory: "./")
+!4 = !{!0}
+!5 = !DIDerivedType(tag: DW_TAG_typedef, name: "struct_t", file: !3, baseType: !6)
+!6 = distinct !DICompositeType(tag: DW_TAG_structure_type, file: !3, size: 8, flags: DIFlagTypePassByValue, elements: !7, identifier: "_ZTS8struct_t")
+!7 = !{!8, !10}
+!8 = !DIDerivedType(tag: DW_TAG_member, name: "x", scope: !6, file: !3, baseType: !9, size: 3, flags: DIFlagBitField, extraData: i64 0)
+!9 = !DIBasicType(name: "unsigned int", size: 32, encoding: DW_ATE_unsigned)
+!10 = !DIDerivedType(tag: DW_TAG_member, name: "y", scope: !6, file: !3, baseType: !9, size: 4, offset: 3, flags: DIFlagBitField, extraData: i64 0)
+!11 = !{i32 7, !"Dwarf Version", i32 5}
+!12 = !{i32 2, !"Debug Info Version", i32 3}
+!13 = distinct !DISubprogram(name: "test1", linkageName: "test1", scope: !3, file: !3, type: !14, spFlags: DISPFlagDefinition, unit: !2)
+!14 = !DISubroutineType(types: !15)
+!15 = !{!9}
+!16 = !DILocalVariable(name: "x", scope: !13, file: !3, type: !9)
+!17 = !DILocation(line: 0, scope: !13)
+!18 = !DILocalVariable(name: "y", scope: !13, file: !3, type: !19)
+!19 = !DIBasicType(name: "signed int", size: 32, encoding: DW_ATE_signed)
+!20 = distinct !DISubprogram(name: "test2", linkageName: "test2", scope: !3, file: !3, type: !14, spFlags: DISPFlagDefinition, unit: !2)
+!21 = !DILocalVariable(name: "x", scope: !20, file: !3, type: !9)
+!22 = !DILocation(line: 0, scope: !20)
+!23 = !DILocalVariable(name: "y", scope: !20, file: !3, type: !19)
+!24 = distinct !DISubprogram(name: "test3", linkageName: "test3", scope: !3, file: !3, type: !14, spFlags: DISPFlagDefinition, unit: !2)
+!25 = !DILocalVariable(name: "x", scope: !24, file: !3, type: !9)
+!26 = !DILocation(line: 0, scope: !24)
+!27 = !DILocalVariable(name: "y", scope: !24, file: !3, type: !19)