Detailed debug-info (DWARF) support in new backends (initially x64).

This PR propagates "value labels" all the way from CLIF to DWARF
metadata on the emitted machine code. The key idea is as follows:

- Translate value-label metadata on the input into "value_label"
  pseudo-instructions when lowering into VCode. These
  pseudo-instructions take a register as input, denote a value label,
  and semantically are like a "move into value label" -- i.e., they
  update the current value (as seen by debugging tools) of the given
  local. These pseudo-instructions emit no machine code.

- Perform a dataflow analysis *at the machine-code level*, tracking
  value-labels that propagate into registers and into [SP+constant]
  stack storage. This is a forward dataflow fixpoint analysis where each
  storage location can contain a *set* of value labels, and each value
  label can reside in a *set* of storage locations. (Meet function is
  pairwise intersection by storage location.)

  This analysis traces value labels symbolically through loads and
  stores and reg-to-reg moves, so it will naturally handle spills and
  reloads without knowing anything special about them.

- When this analysis converges, we have, at each machine-code offset, a
  mapping from value labels to some number of storage locations; for
  each offset for each label, we choose the best location (prefer
  registers). Note that we can choose any location, as the symbolic
  dataflow analysis is sound and guarantees that the value at the
  value_label instruction propagates to all of the named locations.

- Then we can convert this mapping into a format that the DWARF
  generation code (wasmtime's debug crate) can use.

This PR also adds the new-backend variant to the gdb tests on CI.

.github/workflows/main.yml

@@ -292,6 +292,14 @@ jobs:
       env:
         RUST_BACKTRACE: 1
 
+    # Test debug (DWARF) related functionality on new backend.
+    - run: |
+        sudo apt-get update && sudo apt-get install -y gdb
+        cargo test --features experimental_x64 test_debug_dwarf -- --ignored --test-threads 1
+      if: matrix.target == 'x86_64-unknown-linux-gnu'
+      env:
+        RUST_BACKTRACE: 1
+
     # Build and test lightbeam. Note that
     # Lightbeam tests fail right now, but we don't want to block on that.
     - run: cargo build --package lightbeam

cranelift/codegen/Cargo.toml

@@ -74,7 +74,7 @@ all-arch = [
 ]
 
 # For dependent crates that want to serialize some parts of cranelift
-enable-serde = ["serde"]
+enable-serde = ["serde", "regalloc/enable-serde"]
 
 # Allow snapshotting regalloc test cases. Useful only to report bad register
 # allocation failures, or for regalloc.rs developers.

cranelift/codegen/src/context.rs

@@ -473,6 +473,7 @@ impl Context {
         Ok(build_value_labels_ranges::<ComparableSourceLoc>(
             &self.func,
             &self.regalloc,
+            self.mach_compile_result.as_ref(),
             isa,
         ))
     }

cranelift/codegen/src/ir/mod.rs

@@ -58,6 +58,7 @@ pub use crate::ir::table::TableData;
 pub use crate::ir::trapcode::TrapCode;
 pub use crate::ir::types::Type;
 pub use crate::ir::valueloc::{ArgumentLoc, ValueLoc};
+pub use crate::value_label::LabelValueLoc;
 pub use cranelift_codegen_shared::condcodes;
 
 use crate::binemit;

cranelift/codegen/src/isa/aarch64/inst/emit.rs

@@ -2365,6 +2365,9 @@ impl MachInstEmit for Inst {
                     sink.bind_label(jump_around_label);
                 }
             }
+            &Inst::ValueLabelMarker { .. } => {
+                // Nothing; this is only used to compute debug info.
+            }
         }
 
         let end_off = sink.cur_offset();

cranelift/codegen/src/isa/aarch64/inst/mod.rs

@@ -7,7 +7,7 @@ use crate::binemit::CodeOffset;
 use crate::ir::types::{
     B1, B128, B16, B32, B64, B8, F32, F64, FFLAGS, I128, I16, I32, I64, I8, I8X16, IFLAGS, R32, R64,
 };
-use crate::ir::{ExternalName, MemFlags, Opcode, SourceLoc, TrapCode, Type};
+use crate::ir::{ExternalName, MemFlags, Opcode, SourceLoc, TrapCode, Type, ValueLabel};
 use crate::isa::CallConv;
 use crate::machinst::*;
 use crate::{settings, CodegenError, CodegenResult};
@@ -1208,6 +1208,12 @@ pub enum Inst {
         /// The needed space before the next deadline.
         needed_space: CodeOffset,
     },
+
+    /// A definition of a value label.
+    ValueLabelMarker {
+        reg: Reg,
+        label: ValueLabel,
+    },
 }
 
 fn count_zero_half_words(mut value: u64, num_half_words: u8) -> usize {
@@ -2015,6 +2021,9 @@ fn aarch64_get_regs(inst: &Inst, collector: &mut RegUsageCollector) {
             memarg_regs(mem, collector);
         }
         &Inst::VirtualSPOffsetAdj { .. } => {}
+        &Inst::ValueLabelMarker { reg, .. } => {
+            collector.add_use(reg);
+        }
         &Inst::EmitIsland { .. } => {}
     }
 }
@@ -2765,6 +2774,9 @@ fn aarch64_map_regs<RUM: RegUsageMapper>(inst: &mut Inst, mapper: &RUM) {
         }
         &mut Inst::VirtualSPOffsetAdj { .. } => {}
         &mut Inst::EmitIsland { .. } => {}
+        &mut Inst::ValueLabelMarker { ref mut reg, .. } => {
+            map_use(mapper, reg);
+        }
     }
 }
 
@@ -2960,6 +2972,17 @@ impl MachInst for Inst {
     fn ref_type_regclass(_: &settings::Flags) -> RegClass {
         RegClass::I64
     }
+
+    fn gen_value_label_marker(label: ValueLabel, reg: Reg) -> Self {
+        Inst::ValueLabelMarker { label, reg }
+    }
+
+    fn defines_value_label(&self) -> Option<(ValueLabel, Reg)> {
+        match self {
+            Inst::ValueLabelMarker { label, reg } => Some((*label, *reg)),
+            _ => None,
+        }
+    }
 }
 
 //=============================================================================
@@ -4068,6 +4091,10 @@ impl Inst {
                 format!("virtual_sp_offset_adjust {}", offset)
             }
             &Inst::EmitIsland { needed_space } => format!("emit_island {}", needed_space),
+
+            &Inst::ValueLabelMarker { label, reg } => {
+                format!("value_label {:?}, {}", label, reg.show_rru(mb_rru))
+            }
         }
     }
 }

cranelift/codegen/src/isa/aarch64/mod.rs

@@ -79,6 +79,7 @@ impl MachBackend for AArch64Backend {
             frame_size,
             disasm,
             unwind_info,
+            value_labels_ranges: None,
         })
     }
 

cranelift/codegen/src/isa/arm32/mod.rs

@@ -74,6 +74,7 @@ impl MachBackend for Arm32Backend {
             frame_size,
             disasm,
             unwind_info: None,
+            value_labels_ranges: None,
         })
     }
 

cranelift/codegen/src/isa/mod.rs

@@ -325,6 +325,12 @@ pub trait TargetIsa: fmt::Display + Send + Sync {
         Err(RegisterMappingError::UnsupportedArchitecture)
     }
 
+    #[cfg(feature = "unwind")]
+    /// Map a regalloc::Reg to its corresponding DWARF register.
+    fn map_regalloc_reg_to_dwarf(&self, _: ::regalloc::Reg) -> Result<u16, RegisterMappingError> {
+        Err(RegisterMappingError::UnsupportedArchitecture)
+    }
+
     /// Returns an iterator over legal encodings for the instruction.
     fn legal_encodings<'a>(
         &'a self,

cranelift/codegen/src/isa/x64/inst/emit.rs

@@ -2962,6 +2962,10 @@ pub(crate) fn emit(
         Inst::EpiloguePlaceholder => {
             // Generate no code.
         }
+
+        Inst::ValueLabelMarker { .. } => {
+            // Nothing; this is only used to compute debug info.
+        }
     }
 
     state.clear_post_insn();

cranelift/codegen/src/isa/x64/inst/mod.rs

@@ -1,7 +1,7 @@
 //! This module defines x86_64-specific machine instruction types.
 
 use crate::binemit::{CodeOffset, StackMap};
-use crate::ir::{types, ExternalName, Opcode, SourceLoc, TrapCode, Type};
+use crate::ir::{types, ExternalName, Opcode, SourceLoc, TrapCode, Type, ValueLabel};
 use crate::isa::x64::settings as x64_settings;
 use crate::machinst::*;
 use crate::{settings, settings::Flags, CodegenError, CodegenResult};
@@ -474,6 +474,9 @@ pub enum Inst {
     /// reports its own `def`s/`use`s/`mod`s; this adds complexity (the instruction list is no
     /// longer flat) and requires knowledge about semantics and initial-value independence anyway.
     XmmUninitializedValue { dst: Writable<Reg> },
+
+    /// A definition of a value label.
+    ValueLabelMarker { reg: Reg, label: ValueLabel },
 }
 
 pub(crate) fn low32_will_sign_extend_to_64(x: u64) -> bool {
@@ -532,7 +535,8 @@ impl Inst {
             | Inst::XmmCmpRmR { .. }
             | Inst::XmmLoadConst { .. }
             | Inst::XmmMinMaxSeq { .. }
-            | Inst::XmmUninitializedValue { .. } => None,
+            | Inst::XmmUninitializedValue { .. }
+            | Inst::ValueLabelMarker { .. } => None,
 
             // These use dynamic SSE opcodes.
             Inst::GprToXmm { op, .. }
@@ -1762,6 +1766,10 @@ impl PrettyPrint for Inst {
             Inst::Hlt => "hlt".into(),
 
             Inst::Ud2 { trap_code } => format!("ud2 {}", trap_code),
+
+            Inst::ValueLabelMarker { label, reg } => {
+                format!("value_label {:?}, {}", label, reg.show_rru(mb_rru))
+            }
         }
     }
 }
@@ -2021,6 +2029,9 @@ fn x64_get_regs(inst: &Inst, collector: &mut RegUsageCollector) {
         | Inst::Fence { .. } => {
             // No registers are used.
         }
+        Inst::ValueLabelMarker { reg, .. } => {
+            collector.add_use(*reg);
+        }
     }
 }
 
@@ -2385,6 +2396,8 @@ fn x64_map_regs<RUM: RegUsageMapper>(inst: &mut Inst, mapper: &RUM) {
             dst.map_uses(mapper);
         }
 
+        Inst::ValueLabelMarker { ref mut reg, .. } => map_use(mapper, reg),
+
         Inst::Ret
         | Inst::EpiloguePlaceholder
         | Inst::JmpKnown { .. }
@@ -2473,6 +2486,25 @@ impl MachInst for Inst {
         }
     }
 
+    fn stack_op_info(&self) -> Option<MachInstStackOpInfo> {
+        match self {
+            Self::VirtualSPOffsetAdj { offset } => Some(MachInstStackOpInfo::NomSPAdj(*offset)),
+            Self::MovRM {
+                size: 8,
+                src,
+                dst: SyntheticAmode::NominalSPOffset { simm32 },
+            } => Some(MachInstStackOpInfo::StoreNomSPOff(*src, *simm32 as i64)),
+            Self::Mov64MR {
+                src: SyntheticAmode::NominalSPOffset { simm32 },
+                dst,
+            } => Some(MachInstStackOpInfo::LoadNomSPOff(
+                dst.to_reg(),
+                *simm32 as i64,
+            )),
+            _ => None,
+        }
+    }
+
     fn gen_move(dst_reg: Writable<Reg>, src_reg: Reg, ty: Type) -> Inst {
         let rc_dst = dst_reg.to_reg().get_class();
         let rc_src = src_reg.get_class();
@@ -2636,6 +2668,17 @@ impl MachInst for Inst {
         RegClass::I64
     }
 
+    fn gen_value_label_marker(label: ValueLabel, reg: Reg) -> Self {
+        Inst::ValueLabelMarker { label, reg }
+    }
+
+    fn defines_value_label(&self) -> Option<(ValueLabel, Reg)> {
+        match self {
+            Inst::ValueLabelMarker { label, reg } => Some((*label, *reg)),
+            _ => None,
+        }
+    }
+
     type LabelUse = LabelUse;
 }
 

cranelift/codegen/src/isa/x64/mod.rs

@@ -4,13 +4,14 @@ use self::inst::EmitInfo;
 
 use super::TargetIsa;
 use crate::ir::{condcodes::IntCC, Function};
+use crate::isa::unwind::systemv::RegisterMappingError;
 use crate::isa::x64::{inst::regs::create_reg_universe_systemv, settings as x64_settings};
 use crate::isa::Builder as IsaBuilder;
 use crate::machinst::{compile, MachBackend, MachCompileResult, TargetIsaAdapter, VCode};
 use crate::result::CodegenResult;
 use crate::settings::{self as shared_settings, Flags};
 use alloc::boxed::Box;
-use regalloc::{PrettyPrint, RealRegUniverse};
+use regalloc::{PrettyPrint, RealRegUniverse, Reg};
 use target_lexicon::Triple;
 
 mod abi;
@@ -60,6 +61,7 @@ impl MachBackend for X64Backend {
         let buffer = buffer.finish();
         let frame_size = vcode.frame_size();
         let unwind_info = vcode.unwind_info()?;
+        let value_labels_ranges = vcode.value_labels_ranges()?;
 
         let disasm = if want_disasm {
             Some(vcode.show_rru(Some(&create_reg_universe_systemv(flags))))
@@ -72,6 +74,7 @@ impl MachBackend for X64Backend {
             frame_size,
             disasm,
             unwind_info,
+            value_labels_ranges,
         })
     }
 
@@ -127,6 +130,11 @@ impl MachBackend for X64Backend {
     fn create_systemv_cie(&self) -> Option<gimli::write::CommonInformationEntry> {
         Some(inst::unwind::systemv::create_cie())
     }
+
+    #[cfg(feature = "unwind")]
+    fn map_reg_to_dwarf(&self, reg: Reg) -> Result<u16, RegisterMappingError> {
+        inst::unwind::systemv::map_reg(reg).map(|reg| reg.0)
+    }
 }
 
 /// Create a new `isa::Builder`.

cranelift/codegen/src/machinst/adapter.rs

@@ -10,6 +10,9 @@ use crate::settings::Flags;
 #[cfg(feature = "testing_hooks")]
 use crate::regalloc::RegDiversions;
 
+#[cfg(feature = "unwind")]
+use crate::isa::unwind::systemv::RegisterMappingError;
+
 use core::any::Any;
 use std::borrow::Cow;
 use std::fmt;
@@ -134,6 +137,11 @@ impl TargetIsa for TargetIsaAdapter {
         self.backend.create_systemv_cie()
     }
 
+    #[cfg(feature = "unwind")]
+    fn map_regalloc_reg_to_dwarf(&self, r: Reg) -> Result<u16, RegisterMappingError> {
+        self.backend.map_reg_to_dwarf(r)
+    }
+
     fn as_any(&self) -> &dyn Any {
         self as &dyn Any
     }