start with 64 bit

rvdc/src/lib.rs

@@ -5,6 +5,27 @@
 use core::fmt::{self, Debug, Display};
 use core::ops::RangeInclusive;
 
+/// The register size of the ISA, RV32 or RV64.
+#[derive(Debug, PartialEq, Eq, PartialOrd, Ord, Hash)]
+pub enum Xlen {
+    /// 32 bit
+    Rv32,
+    /// 64 bit
+    Rv64,
+}
+
+impl Xlen {
+    /// Whether this is [`Xlen::Rv32`].
+    pub fn is_32(self) -> bool {
+        matches!(self, Self::Rv32)
+    }
+
+    /// Whether this is [`Xlen::Rv64`].
+    pub fn is_64(self) -> bool {
+        matches!(self, Self::Rv64)
+    }
+}
+
 /// A decoded RISC-V integer register.
 #[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]
 pub struct Reg(pub u8);
@@ -109,7 +130,7 @@ impl Display for Reg {
 ///
 /// This type is XLEN-agnostic, use the XLEN-specific accessors to get the correct value.
 #[derive(Copy, Clone, PartialEq, Eq, Hash)]
-pub struct Imm(u32);
+pub struct Imm(u64);
 
 impl Imm {
     /// The immediate `0`.
@@ -118,23 +139,33 @@ impl Imm {
 
     /// Create a new immediate from the (if necessary) sign-extended value.
     pub const fn new_i32(value: i32) -> Self {
-        Self(value as u32)
+        Self(value as i64 as u64)
     }
 
     /// Create a new immediate from the (if necessary) zero-extended value.
     pub const fn new_u32(value: u32) -> Self {
-        Self(value)
+        Self(value as u64)
     }
 
     /// Get the `u32` (RV32) value of the immediate.
     pub const fn as_u32(self) -> u32 {
-        self.0
+        self.0 as u32
     }
 
     /// Get the `i32` (RV32) value of the immediate.
     pub const fn as_i32(self) -> i32 {
         self.0 as i32
     }
+
+    /// Get the `u64` (RV64) value of the immediate.
+    pub const fn as_u64(self) -> u64 {
+        self.0 as u64
+    }
+
+    /// Get the `i64` (RV64) value of the immediate.
+    pub const fn as_i64(self) -> i64 {
+        self.0 as i64
+    }
 }
 
 impl From<i32> for Imm {
@@ -216,6 +247,8 @@ pub enum Inst {
 
     /// Add Immediate
     Addi { imm: Imm, dest: Reg, src1: Reg },
+    /// Add Immediate 32-bit (**RV64 only**)
+    AddiW { imm: Imm, dest: Reg, src1: Reg },
     /// Set Less Than Immediate (signed)
     Slti { imm: Imm, dest: Reg, src1: Reg },
     /// Set Less Than Immediate Unsigned
@@ -228,10 +261,16 @@ pub enum Inst {
     Andi { imm: Imm, dest: Reg, src1: Reg },
     /// Shift Left Logical Immediate
     Slli { imm: Imm, dest: Reg, src1: Reg },
+    /// Shift Left Logical Immediate 32-bit (**RV64 only**)
+    SlliW { imm: Imm, dest: Reg, src1: Reg },
     /// Shift Right Logical Immediate (unsigned)
     Srli { imm: Imm, dest: Reg, src1: Reg },
+    /// Shift Right Logical Immediate (unsigned) 32-bit (**RV64 only**)
+    SrliW { imm: Imm, dest: Reg, src1: Reg },
     /// Shift Right Arithmetic Immediate (signed)
     Srai { imm: Imm, dest: Reg, src1: Reg },
+    /// Shift Right Arithmetic Immediate (signed) 32-bit (**RV64 only**)
+    SraiW { imm: Imm, dest: Reg, src1: Reg },
 
     /// Add
     Add { dest: Reg, src1: Reg, src2: Reg },
@@ -502,6 +541,13 @@ impl Display for Inst {
                     write!(f, "addi {dest}, {src1}, {}", imm.as_i32())
                 }
             }
+            Inst::AddiW { imm, dest, src1 } => {
+                if imm.as_u32() == 0 {
+                    write!(f, "sext.w {dest}, {src1}")
+                } else {
+                    write!(f, "addi.w {dest}, {src1}, {}", imm.as_i32())
+                }
+            }
             Inst::Slti {
                 imm,
                 dest,
@@ -532,16 +578,31 @@ impl Display for Inst {
                 dest,
                 src1: rs1,
             } => write!(f, "slli {dest}, {rs1}, {}", imm.as_i32()),
+            Inst::SlliW {
+                imm,
+                dest,
+                src1: rs1,
+            } => write!(f, "slliw {dest}, {rs1}, {}", imm.as_i32()),
             Inst::Srli {
                 imm,
                 dest,
                 src1: rs1,
             } => write!(f, "srli {dest}, {rs1}, {}", imm.as_i32()),
+            Inst::SrliW {
+                imm,
+                dest,
+                src1: rs1,
+            } => write!(f, "srliw {dest}, {rs1}, {}", imm.as_i32()),
             Inst::Srai {
                 imm,
                 dest,
                 src1: rs1,
             } => write!(f, "srai {dest}, {rs1}, {}", imm.as_i32()),
+            Inst::SraiW {
+                imm,
+                dest,
+                src1: rs1,
+            } => write!(f, "sraiw {dest}, {rs1}, {}", imm.as_i32()),
             Inst::Add { dest, src1, src2 } => {
                 write!(f, "add {dest}, {src1}, {src2}")
             }
@@ -847,12 +908,15 @@ impl Inst {
     /// If the caller wants to avoid reading more bytes than necessary, [`Self::first_byte_is_compressed`]
     /// can be used to check, read the required bytes, and then call [`Self::decode_compressed`] or
     /// [`Self::decode_normal`] directly.
-    pub fn decode(code: u32) -> Result<(Inst, IsCompressed), DecodeError> {
+    pub fn decode(code: u32, xlen: Xlen) -> Result<(Inst, IsCompressed), DecodeError> {
         let is_compressed = (code & 0b11) != 0b11;
         if is_compressed {
-            Ok((Self::decode_compressed(code as u16)?, IsCompressed::Yes))
+            Ok((
+                Self::decode_compressed(code as u16, xlen)?,
+                IsCompressed::Yes,
+            ))
         } else {
-            Ok((Self::decode_normal(code)?, IsCompressed::No))
+            Ok((Self::decode_normal(code, xlen)?, IsCompressed::No))
         }
     }
 
@@ -867,7 +931,7 @@ impl Inst {
     /// let inst = rvdc::Inst::decode_compressed(x).unwrap();
     /// assert_eq!(inst, expected);
     /// ```
-    pub fn decode_compressed(code: u16) -> Result<Inst, DecodeError> {
+    pub fn decode_compressed(code: u16, xlen: Xlen) -> Result<Inst, DecodeError> {
         let code = InstCodeC(code);
         if code.0 == 0 {
             return Err(decode_error(code, "null instruction"));
@@ -1146,7 +1210,7 @@ impl Inst {
     }
 
     /// Decode a normal (not compressed) instruction.
-    pub fn decode_normal(code: u32) -> Result<Inst, DecodeError> {
+    pub fn decode_normal(code: u32, xlen: Xlen) -> Result<Inst, DecodeError> {
         let code = InstCode(code);
         let inst = match code.opcode() {
             // LUI
@@ -1312,6 +1376,47 @@ impl Inst {
                 },
                 _ => return Err(decode_error(code, "OP-IMM funct3")),
             },
+            // OP-IMM-32
+            0b0011011 => {
+                if !xlen.is_64() {
+                    return Err(decode_error(code, "OP-IMM-32 only on RV64"));
+                }
+
+                match code.funct3() {
+                    0b000 => Inst::AddiW {
+                        imm: code.imm_i(),
+                        dest: code.rd(),
+                        src1: code.rs1(),
+                    },
+                    // SLLIW
+                    0b001 => {
+                        if code.funct7() != 0 {
+                            return Err(decode_error(code, "SLLIW funct7"));
+                        }
+
+                        Inst::SlliW {
+                            imm: Imm::new_u32(code.rs2_imm()),
+                            dest: code.rd(),
+                            src1: code.rs1(),
+                        }
+                    }
+
+                    0b101 => match code.funct7() {
+                        0b0000000 => Inst::SrliW {
+                            imm: Imm::new_u32(code.rs2_imm()),
+                            dest: code.rd(),
+                            src1: code.rs1(),
+                        },
+                        0b0100000 => Inst::SraiW {
+                            imm: Imm::new_u32(code.rs2_imm()),
+                            dest: code.rd(),
+                            src1: code.rs1(),
+                        },
+                        _ => return Err(decode_error(code, "OP-IMM-32 funct7")),
+                    },
+                    _ => return Err(decode_error(code, "OP-IMM-32 funct3")),
+                }
+            }
             // OP
             0b0110011 => {
                 let (dest, src1, src2) = (code.rd(), code.rs1(), code.rs2());
@@ -1465,6 +1570,7 @@ mod tests {
     use crate::Imm;
     use crate::Inst;
     use crate::Reg;
+    use crate::Xlen;
 
     #[test]
     #[cfg_attr(not(slow_tests), ignore = "cfg(slow_tests) not enabled")]
@@ -1476,9 +1582,20 @@ mod tests {
                 std::print!("\r{}{}", "#".repeat(done), "-".repeat(100 - done));
                 std::io::stdout().flush().unwrap();
             }
-            let _ = Inst::decode(i);
+            let _ = Inst::decode(i, Xlen::Rv32);
         }
-        let _ = Inst::decode(u32::MAX);
+        let _ = Inst::decode(u32::MAX, Xlen::Rv32);
+
+        for i in 0..u32::MAX {
+            if (i % (2 << 25)) == 0 {
+                let percent = i as f32 / (u32::MAX as f32);
+                let done = (100.0 * percent) as usize;
+                std::print!("\r{}{}", "#".repeat(done), "-".repeat(100 - done));
+                std::io::stdout().flush().unwrap();
+            }
+            let _ = Inst::decode(i, Xlen::Rv64);
+        }
+        let _ = Inst::decode(u32::MAX, Xlen::Rv64);
     }
 
     #[test]
@@ -1591,7 +1708,7 @@ mod tests {
             let start_time = std::time::Instant::now();
 
             let insts = (start..=start.saturating_add(CHUNK_SIZE))
-                .filter_map(|code| Some((code, Inst::decode_normal(code).ok()?)))
+                .filter_map(|code| Some((code, Inst::decode_normal(code, Xlen::Rv32).ok()?)))
                 .filter(|(_, inst)| is_inst_supposed_to_roundtrip(inst))
                 .collect::<Vec<_>>();
 
@@ -1631,7 +1748,7 @@ mod tests {
     #[ignore = "this doesn't quite work yet because there is often a non-canonical encoding"]
     fn compressed_clang_roundtrip() {
         let insts = (0..=u16::MAX)
-            .filter_map(|code| Some((code, Inst::decode_compressed(code).ok()?)))
+            .filter_map(|code| Some((code, Inst::decode_compressed(code, Xlen::Rv32).ok()?)))
             .filter(|(_, inst)| is_compressed_inst_supposed_to_roundtrip(inst))
             .collect::<Vec<_>>();
 

src/elf.rs

@@ -32,18 +32,30 @@ pub struct Header {
 }
 
 #[derive(Debug)]
-pub struct Phdr32 {
+pub struct Phdr {
     pub p_type: u32,
-    pub p_offset: u32,
-    pub p_vaddr: u32,
-    pub p_paddr: u32,
-    pub p_filesz: u32,
-    pub p_memsz: u32,
+    pub p_offset: Offset,
+    pub p_vaddr: Addr,
+    pub p_paddr: Addr,
+    pub p_filesz: u64,
+    pub p_memsz: u64,
     pub p_flags: u32,
-    pub p_align: u32,
+    pub p_align: u64,
 }
 
 impl Elf<'_> {
+    fn class(&self) -> Result<ElfClass> {
+        let (_, class) = self.content.split_u32()?;
+        let (class, _) = class.split_bytes(1)?;
+        Ok(match class[0] {
+            // ELFCLASS32
+            1 => ElfClass::Elf32,
+            // ELFCLASS64
+            2 => ElfClass::Elf64,
+            _ => bail!("not a ELF32 or ELF64 file (EI_CLASS={})", class[0]),
+        })
+    }
+
     pub fn header(&self) -> Result<Header> {
         let (ident, rest) = self.content.split_bytes(16)?;
         if ident[..4] != *b"\x7fELF" {
@@ -140,8 +152,9 @@ impl Elf<'_> {
         })
     }
 
-    pub fn segments_32(&self) -> Result<Vec<Phdr32>> {
+    pub fn segments(&self) -> Result<Vec<Phdr>> {
         let header = self.header()?;
+        let class = self.class()?;
 
         let (_, phdrs) = self.content.split_bytes(header.e_phoff.0 as usize)?;
         let (mut phdrs, _) = phdrs.split_bytes((header.e_phentsize * header.e_phnum) as usize)?;
@@ -152,6 +165,8 @@ impl Elf<'_> {
             let phdr;
             (phdr, phdrs) = phdrs.split_bytes(header.e_phentsize as usize)?;
 
+            let phdr = match class {
+                ElfClass::Elf32 => {
                     let (p_type, phdr) = phdr.split_u32()?;
                     let (p_offset, phdr) = phdr.split_u32()?;
                     let (p_vaddr, phdr) = phdr.split_u32()?;
@@ -161,16 +176,40 @@ impl Elf<'_> {
                     let (p_flags, phdr) = phdr.split_u32()?;
                     let (p_align, _) = phdr.split_u32()?;
 
-            parsed_phdrs.push(Phdr32 {
+                    Phdr {
                         p_type,
-                p_offset,
-                p_vaddr,
-                p_paddr,
+                        p_offset: Offset(p_offset as u64),
+                        p_vaddr: Addr(p_vaddr as u64),
+                        p_paddr: Addr(p_paddr as u64),
+                        p_filesz: p_filesz as u64,
+                        p_memsz: p_memsz as u64,
+                        p_flags,
+                        p_align: p_align as u64,
+                    }
+                }
+                ElfClass::Elf64 => {
+                    let (p_type, phdr) = phdr.split_u32()?;
+                    let (p_flags, phdr) = phdr.split_u32()?;
+                    let (p_offset, phdr) = phdr.split_u64()?;
+                    let (p_vaddr, phdr) = phdr.split_u64()?;
+                    let (p_paddr, phdr) = phdr.split_u64()?;
+                    let (p_filesz, phdr) = phdr.split_u64()?;
+                    let (p_memsz, phdr) = phdr.split_u64()?;
+                    let (p_align, _) = phdr.split_u64()?;
+
+                    Phdr {
+                        p_type,
+                        p_offset: Offset(p_offset),
+                        p_vaddr: Addr(p_vaddr),
+                        p_paddr: Addr(p_paddr),
                         p_filesz,
                         p_memsz,
                         p_flags,
                         p_align,
-            });
+                    }
+                }
+            };
+            parsed_phdrs.push(phdr);
         }
 
         Ok(parsed_phdrs)

src/emu.rs

@@ -20,44 +20,50 @@ impl Memory {
             Ok(())
         }
     }
-    pub fn slice<XLEN: XLen>(&self, addr: XLEN, len: u32) -> Result<&[u8], Status> {
+    pub fn slice<XLEN: XLen>(&self, addr: XLEN, len: XLEN) -> Result<&[u8], Status> {
         self.mem
             .get((addr.as_usize())..)
             .ok_or(Status::InvalidMemoryAccess(addr.as_usize()))?
-            .get(..(len as usize))
+            .get(..(len.as_usize()))
             .ok_or(Status::InvalidMemoryAccess(addr.as_usize()))
     }
-    pub fn slice_mut<XLEN: XLen>(&mut self, addr: XLEN, len: u32) -> Result<&mut [u8], Status> {
+    pub fn slice_mut<XLEN: XLen>(&mut self, addr: XLEN, len: XLEN) -> Result<&mut [u8], Status> {
         self.mem
             .get_mut((addr.as_usize())..)
             .ok_or(Status::InvalidMemoryAccess(addr.as_usize()))?
-            .get_mut(..(len as usize))
+            .get_mut(..(len.as_usize()))
             .ok_or(Status::InvalidMemoryAccess(addr.as_usize()))
     }
 
     pub fn load_u8<XLEN: XLen>(&self, addr: XLEN) -> Result<u8, Status> {
-        Ok(u8::from_le_bytes(self.slice(addr, 1)?.try_into().unwrap()))
+        Ok(u8::from_le_bytes(
+            self.slice(addr, XLEN::from_32_z(1))?.try_into().unwrap(),
+        ))
     }
     pub fn load_u16<XLEN: XLen>(&self, addr: XLEN) -> Result<u16, Status> {
-        Ok(u16::from_le_bytes(self.slice(addr, 2)?.try_into().unwrap()))
+        Ok(u16::from_le_bytes(
+            self.slice(addr, XLEN::from_32_z(2))?.try_into().unwrap(),
+        ))
     }
     pub fn load_u32<XLEN: XLen>(&self, addr: XLEN) -> Result<u32, Status> {
-        Ok(u32::from_le_bytes(self.slice(addr, 4)?.try_into().unwrap()))
+        Ok(u32::from_le_bytes(
+            self.slice(addr, XLEN::from_32_z(4))?.try_into().unwrap(),
+        ))
     }
     pub fn store_u8<XLEN: XLen>(&mut self, addr: XLEN, value: u8) -> Result<(), Status> {
-        self.slice_mut(addr, 1)?
+        self.slice_mut(addr, XLEN::from_32_z(1))?
             .copy_from_slice(&value.to_le_bytes());
         Ok(())
     }
     pub fn store_u16<XLEN: XLen>(&mut self, addr: XLEN, value: u16) -> Result<(), Status> {
         self.check_align(addr, 2)?;
-        self.slice_mut(addr, 2)?
+        self.slice_mut(addr, XLEN::from_32_z(2))?
             .copy_from_slice(&value.to_le_bytes());
         Ok(())
     }
     pub fn store_u32<XLEN: XLen>(&mut self, addr: XLEN, value: u32) -> Result<(), Status> {
         self.check_align(addr, 4)?;
-        self.slice_mut(addr, 4)?
+        self.slice_mut(addr, XLEN::from_32_z(4))?
             .copy_from_slice(&value.to_le_bytes());
         Ok(())
     }
@@ -139,6 +145,28 @@ fn hash_color(value: usize) -> impl Display {
 }
 
 impl<XLEN: XLen> Emulator<XLEN> {
+    pub fn new(
+        mem: Memory,
+        start: XLEN,
+        break_addr: XLEN,
+        debug: bool,
+        ecall_handler: Box<dyn FnMut(&mut Memory, &mut [XLEN; 32]) -> Result<(), Status>>,
+    ) -> Self {
+        Self {
+            mem,
+            xreg: [XLEN::ZERO; 32],
+            xreg0_value: XLEN::ZERO,
+            pc: start,
+            reservation_set: None,
+
+            is_breaking: false,
+
+            break_pc: break_addr,
+            debug,
+            ecall_handler,
+        }
+    }
+
     pub fn start_linux(&mut self) -> Status {
         self.setup_linux_stack().unwrap();
 
@@ -210,7 +238,7 @@ impl<XLEN: XLen> Emulator<XLEN> {
             print!("0x{:x} ", self.pc.as_usize());
         }
 
-        let (inst, was_compressed) = Inst::decode(code)?;
+        let (inst, was_compressed) = Inst::decode(code, XLEN::XLEN)?;
 
         if self.debug {
             println!(
@@ -649,11 +677,21 @@ impl<XLEN: XLen> Emulator<XLEN> {
 }
 
 pub trait XLen: Copy + PartialEq + Eq {
+    type Signed;
+    type NextUnsigned;
+    type NextSigned;
+
+    const XLEN: rvdc::Xlen;
+
     const ZERO: Self;
     const SIGNED_MIN: Self;
     const MAX: Self;
 
-    fn from_bool(v: bool) -> Self;
+    fn switch<R>(self, on_32: impl FnOnce(u32) -> R, on_64: impl FnOnce(u64) -> R) -> R;
+
+    fn from_bool(v: bool) -> Self {
+        Self::from_32_z(v as u32)
+    }
     fn from_8_z(v: u8) -> Self {
         Self::from_32_z(v as u32)
     }
@@ -704,29 +742,17 @@ pub trait XLen: Copy + PartialEq + Eq {
     fn unsigned_rem(self, other: Self) -> Self;
 }
 
-impl XLen for u32 {
+macro_rules! xlen_impl {
+    () => {
         const ZERO: Self = 0;
-    const SIGNED_MIN: Self = i32::MIN as u32;
-    const MAX: Self = u32::MAX;
+        const SIGNED_MIN: Self = Self::Signed::MIN as Self;
+        const MAX: Self = Self::MAX;
 
-    fn from_bool(v: bool) -> Self {
-        v as u32
-    }
-    fn from_32_s(v: u32) -> Self {
-        v
-    }
-    fn from_32_z(v: u32) -> Self {
-        v
-    }
-    fn from_imm(v: Imm) -> Self {
-        v.as_u32()
-    }
         fn as_usize(self) -> usize {
             self as usize
         }
-
         fn truncate32(self) -> u32 {
-        self
+            self as u32
         }
 
         fn add(self, other: Self) -> Self {
@@ -745,13 +771,13 @@ impl XLen for u32 {
             self ^ other
         }
         fn signed_lt(self, other: Self) -> bool {
-        (self as i32) < (other as i32)
+            (self as Self::Signed) < (other as Self::Signed)
         }
         fn unsigned_lt(self, other: Self) -> bool {
             self < other
         }
         fn signed_ge(self, other: Self) -> bool {
-        (self as i32) >= (other as i32)
+            (self as Self::Signed) >= (other as Self::Signed)
         }
         fn unsigned_ge(self, other: Self) -> bool {
             self >= other
@@ -763,42 +789,93 @@ impl XLen for u32 {
             self.wrapping_shr(other)
         }
         fn signed_shr(self, other: u32) -> Self {
-        ((self as i32).wrapping_shr(other)) as u32
+            ((self as Self::Signed).wrapping_shr(other)) as Self
         }
         fn signed_min(self, other: Self) -> Self {
-        (self as i32).min(other as i32) as u32
+            (self as Self::Signed).min(other as Self::Signed) as Self
         }
         fn unsigned_min(self, other: Self) -> Self {
             self.min(other)
         }
         fn signed_max(self, other: Self) -> Self {
-        (self as i32).max(other as i32) as u32
+            (self as Self::Signed).max(other as Self::Signed) as Self
         }
         fn unsigned_max(self, other: Self) -> Self {
             self.max(other)
         }
         fn mul_lower(self, other: Self) -> Self {
-        (self as i32).wrapping_mul(other as i32) as u32
+            (self as Self::Signed).wrapping_mul(other as Self::Signed) as Self
         }
         fn signed_mul_upper(self, other: Self) -> Self {
-        let mul_result = (self as i32 as i64).wrapping_mul(other as i32 as i64);
-        let shifted = (mul_result as u64) >> 32;
-        shifted as u32
+            let mul_result = (self as Self::Signed as Self::NextSigned)
+                .wrapping_mul(other as Self::Signed as Self::NextSigned);
+            let shifted = (mul_result as Self::NextUnsigned) >> Self::BITS;
+            shifted as Self
         }
         fn unsigned_mul_upper(self, other: Self) -> Self {
-        let shifted = ((self as u64).wrapping_mul(other as u64)) >> 32;
-        shifted as u32
+            let shifted = ((self as Self::NextUnsigned).wrapping_mul(other as Self::NextUnsigned))
+                >> Self::BITS;
+            shifted as Self
         }
         fn signed_div(self, other: Self) -> Self {
-        ((self as i32) / (other as i32)) as u32
+            ((self as Self::Signed) / (other as Self::Signed)) as Self
         }
         fn unsigned_div(self, other: Self) -> Self {
             self / other
         }
         fn signed_rem(self, other: Self) -> Self {
-        ((self as i32) % (other as i32)) as u32
+            ((self as Self::Signed) % (other as Self::Signed)) as Self
         }
         fn unsigned_rem(self, other: Self) -> Self {
             self % other
         }
+    };
+}
+
+impl XLen for u32 {
+    type Signed = i32;
+    type NextUnsigned = u64;
+    type NextSigned = i64;
+
+    const XLEN: rvdc::Xlen = rvdc::Xlen::Rv32;
+
+    fn switch<R>(self, on_32: impl FnOnce(u32) -> R, _: impl FnOnce(u64) -> R) -> R {
+        on_32(self)
+    }
+
+    fn from_32_s(v: u32) -> Self {
+        v
+    }
+    fn from_32_z(v: u32) -> Self {
+        v
+    }
+    fn from_imm(v: Imm) -> Self {
+        v.as_u32()
+    }
+
+    xlen_impl!();
+}
+
+impl XLen for u64 {
+    type Signed = i64;
+    type NextUnsigned = u128;
+    type NextSigned = i128;
+
+    const XLEN: rvdc::Xlen = rvdc::Xlen::Rv64;
+
+    fn switch<R>(self, _: impl FnOnce(u32) -> R, on_64: impl FnOnce(u64) -> R) -> R {
+        on_64(self)
+    }
+
+    fn from_32_s(v: u32) -> Self {
+        v as i32 as i64 as u64
+    }
+    fn from_32_z(v: u32) -> Self {
+        v as u64
+    }
+    fn from_imm(v: Imm) -> Self {
+        v.as_u64()
+    }
+
+    xlen_impl!();
 }

src/lib.rs

@@ -9,13 +9,14 @@ const MEMORY_SIZE: usize = 2 << 21;
 pub fn execute_linux_elf(
     elf: &[u8],
     debug: bool,
-    break_addr: u32,
-    ecall_handler: Box<dyn FnMut(&mut emu::Memory, &mut [u32; 32]) -> Result<(), emu::Status>>,
+    break_addr: u64,
+    ecall_handler32: Box<dyn FnMut(&mut emu::Memory, &mut [u32; 32]) -> Result<(), emu::Status>>,
+    ecall_handler64: Box<dyn FnMut(&mut emu::Memory, &mut [u64; 32]) -> Result<(), emu::Status>>,
 ) -> eyre::Result<emu::Status> {
     let elf = elf::Elf { content: elf };
     let header = elf.header()?;
 
-    let segments = elf.segments_32()?;
+    let segments = elf.segments()?;
 
     let mut mem = emu::Memory {
         mem: vec![0; MEMORY_SIZE],
@@ -30,13 +31,13 @@ pub fn execute_linux_elf(
                 if phdr.p_filesz > 0 {
                     let contents = &elf
                         .content
-                        .get((phdr.p_offset as usize)..)
+                        .get((phdr.p_offset.0 as usize)..)
                         .ok_or_eyre("invalid offset")?
                         .get(..(phdr.p_filesz as usize))
                         .ok_or_eyre("invalid offset")?;
 
                     mem.mem
-                        .get_mut((phdr.p_vaddr as usize)..)
+                        .get_mut((phdr.p_vaddr.0 as usize)..)
                         .ok_or_eyre("invalid offset")?
                         .get_mut(..(phdr.p_filesz as usize))
                         .ok_or_eyre("invalid offset")?
@@ -61,19 +62,16 @@ pub fn execute_linux_elf(
 
     let start = header.e_entry;
 
-    let mut emu = emu::Emulator {
-        mem,
-        xreg: [0; 32],
-        xreg0_value: 0,
-        pc: start.0 as u32,
-        reservation_set: None,
-
-        is_breaking: false,
-
-        break_pc: break_addr,
-        debug,
-        ecall_handler,
-    };
-
+    match header.class {
+        elf::ElfClass::Elf32 => {
+            let mut emu =
+                emu::Emulator::<u32>::new(mem, start.0 as u32, break_addr as u32, debug, ecall_handler32);
             Ok(emu.start_linux())
+        }
+        elf::ElfClass::Elf64 => {
+            let mut emu =
+                emu::Emulator::<u64>::new(mem, start.0, break_addr, debug, ecall_handler64);
+            Ok(emu.start_linux())
+        }
+    }
 }

src/main.rs

@@ -1,7 +1,7 @@
-use std::io::Write;
+use std::{cell::RefCell, io::Write, sync::Arc};
 
 use eyre::eyre;
-use rustv32i::emu::{self, Memory};
+use rustv32i::emu::{self, Memory, XLen};
 use rvdc::Reg;
 
 fn main() -> eyre::Result<()> {
@@ -12,22 +12,26 @@ fn main() -> eyre::Result<()> {
         .nth(1)
         .map(|addr| {
             if let Some(addr) = addr.strip_prefix("0x") {
-                u32::from_str_radix(addr, 16)
+                u64::from_str_radix(addr, 16)
             } else {
-                u32::from_str_radix(&addr, 10)
+                u64::from_str_radix(&addr, 10)
             }
         })
         .unwrap_or(Ok(0))?;
 
     let debug = std::env::args().any(|arg| arg == "--debug");
 
-    let mut syscall_state = SyscallState { set_child_tid: 0 };
+    let syscall_state = Arc::new(RefCell::new(SyscallState { set_child_tid: 0 }));
+    let syscall_state64 = syscall_state.clone();
 
     let status = rustv32i::execute_linux_elf(
         &content,
         debug,
         break_addr,
-        Box::new(move |mem, xreg| ecall_handler(mem, xreg, &mut syscall_state)),
+        Box::new(move |mem, xreg| ecall_handler::<u32>(mem, xreg, &mut syscall_state.borrow_mut())),
+        Box::new(move |mem, xreg| {
+            ecall_handler::<u64>(mem, xreg, &mut syscall_state64.borrow_mut())
+        }),
     )?;
 
     std::io::stdout().flush()?;
@@ -44,12 +48,12 @@ fn main() -> eyre::Result<()> {
 }
 
 struct SyscallState {
-    set_child_tid: u32,
+    set_child_tid: u64,
 }
 
-fn ecall_handler(
+fn ecall_handler<XLEN: Into<u64> + From<u32> + XLen>(
     mem: &mut Memory,
-    xreg: &mut [u32; 32],
+    xreg: &mut [XLEN; 32],
     syscall_state: &mut SyscallState,
 ) -> Result<(), emu::Status> {
     let nr = xreg[Reg::A7.0 as usize];
@@ -60,22 +64,22 @@ fn ecall_handler(
 
     // https://jborza.com/post/2021-05-11-riscv-linux-syscalls/
     // https://github.com/torvalds/linux/blob/master/include/uapi/asm-generic/unistd.h
-    match nr {
+    match nr.into() {
         // ioctl
         29 => {
             let fd = arg0;
             let request = arg1;
 
-            match request {
+            match request.into() {
                 // TIOCGWINSZ
                 0x5413 => {
-                    let wsz_ptr = xreg[Reg::A2.0 as usize];
+                    let wsz_ptr = xreg[Reg::A2.0 as usize].into();
 
                     let mut wsz: libc::winsize = unsafe { std::mem::zeroed() };
 
-                    let r = unsafe { libc::ioctl(fd as i32, libc::TIOCGWINSZ, &mut wsz) };
+                    let r = unsafe { libc::ioctl(fd.into() as i32, libc::TIOCGWINSZ, &mut wsz) };
 
-                    xreg[Reg::A0.0 as usize] = r as u32;
+                    xreg[Reg::A0.0 as usize] = (r as u32).into();
                     if r >= 0 {
                         mem.store_u16(wsz_ptr, wsz.ws_row)?;
                         mem.store_u16(wsz_ptr + 2, wsz.ws_col)?;
@@ -83,16 +87,16 @@ fn ecall_handler(
                         mem.store_u16(wsz_ptr + 6, wsz.ws_ypixel)?;
                     }
                 }
-                _ => todo!("unknown ioctl: {request}"),
+                _ => todo!("unknown ioctl: {}", request.into()),
             }
         }
         // read
         63 => {
-            let fd = arg0;
+            let fd = arg0.into();
             let ptr = xreg[Reg::A1.0 as usize];
             let len = xreg[Reg::A2.0 as usize];
 
-            let buf = mem.slice_mut(ptr, len)?;
+            let buf = mem.slice_mut::<XLEN>(ptr, len)?;
 
             let len = unsafe { libc::read(fd as i32, buf.as_mut_ptr().cast(), buf.len()) };
             let ret = if len < 0 {
@@ -101,11 +105,11 @@ fn ecall_handler(
                 len as u32
             };
 
-            xreg[Reg::A0.0 as usize] = ret;
+            xreg[Reg::A0.0 as usize] = ret.into();
         }
         // write
         64 => {
-            let fd = arg0;
+            let fd = arg0.into();
             let ptr = xreg[Reg::A1.0 as usize];
             let len = xreg[Reg::A2.0 as usize];
 
@@ -118,13 +122,13 @@ fn ecall_handler(
                 len as u32
             };
 
-            xreg[Reg::A0.0 as usize] = ret;
+            xreg[Reg::A0.0 as usize] = ret.into();
         }
         // https://man7.org/linux/man-pages/man3/writev.3p.html
         66 => {
-            let fd = arg0;
-            let iovec = arg1;
-            let iovcnt = arg2;
+            let fd = arg0.into();
+            let iovec = arg1.into();
+            let iovcnt = arg2.into();
 
             let mut written = 0;
 
@@ -142,27 +146,27 @@ fn ecall_handler(
                 };
 
                 if (ret as i32) < 0 {
-                    xreg[Reg::A0.0 as usize] = ret;
+                    xreg[Reg::A0.0 as usize] = ret.into();
                     return Ok(());
                 } else {
                     written += ret;
                 }
             }
 
-            xreg[Reg::A0.0 as usize] = written;
+            xreg[Reg::A0.0 as usize] = written.into();
         }
         // exit | exit_group
         93 | 94 => {
             return Err(emu::Status::Exit {
-                code: xreg[Reg::A0.0 as usize] as i32,
+                code: xreg[Reg::A0.0 as usize].into() as i32,
             });
         }
         // <https://man7.org/linux/man-pages/man2/set_tid_address.2.html>
         96 => {
             let tidptr = arg0;
-            syscall_state.set_child_tid = tidptr;
+            syscall_state.set_child_tid = tidptr.into();
 
-            xreg[Reg::A0.0 as usize] = 1; // thread ID
+            xreg[Reg::A0.0 as usize] = 1.into(); // thread ID
         }
         // ppoll - called for some stdin/stdout/stderr check.
         414 => {
@@ -170,10 +174,10 @@ fn ecall_handler(
             // and opens /dev/null for them if they are.
             // They're always valid here, so just get out.
 
-            xreg[Reg::A0.0 as usize] = 0;
+            xreg[Reg::A0.0 as usize] = 0.into();
         }
         _ => {
-            todo!("unkonwn syscall: {nr}");
+            todo!("unkonwn syscall: {}", nr.into());
         }
     }
 

tests/check.rs

@@ -27,21 +27,30 @@ fn check() -> eyre::Result<()> {
             continue;
         }
 
-        test_case(tmpdir.path(), &file, name, "rv32ima")?;
-        test_case(tmpdir.path(), &file, name, "rv32imac")?;
+        test_case(tmpdir.path(), &file, name, 32, "rv32ima")?;
+        test_case(tmpdir.path(), &file, name, 32, "rv32imac")?;
+
+        // test_case(tmpdir.path(), &file, name, 64, "rv64ima")?;
+        // test_case(tmpdir.path(), &file, name, 64, "rv64imac")?;
     }
 
     Ok(())
 }
 
-fn test_case(tmpdir: &Path, file: &DirEntry, name: &str, march: &str) -> eyre::Result<()> {
+fn test_case(
+    tmpdir: &Path,
+    file: &DirEntry,
+    name: &str,
+    size: u8,
+    march: &str,
+) -> eyre::Result<()> {
     let name = format!("{name} ({march})");
     write!(std::io::stdout(), "test {name} ...")?;
     std::io::stdout().flush()?;
 
     eprintln!("---- START TEST {name} -----");
 
-    let output = build(&tmpdir, &file.path(), march).wrap_err(format!("building {name}"))?;
+    let output = build(&tmpdir, &file.path(), size, march).wrap_err(format!("building {name}"))?;
     let content =
         std::fs::read(&output).wrap_err(format!("reading output from {}", output.display()))?;
 
@@ -60,6 +69,17 @@ fn test_case(tmpdir: &Path, file: &DirEntry, name: &str, march: &str) -> eyre::R
                 Err(rustv32i::emu::Status::Trap("wrong syscall"))
             }
         }),
+        Box::new(|_, xreg| {
+            if xreg[Reg::A7.0 as usize] == u32::MAX as u64 {
+                if xreg[Reg::A0.0 as usize] == 1 {
+                    Err(rustv32i::emu::Status::Exit { code: 0 })
+                } else {
+                    Err(rustv32i::emu::Status::Trap("fail"))
+                }
+            } else {
+                Err(rustv32i::emu::Status::Trap("wrong syscall"))
+            }
+        }),
     )
     .wrap_err(format!("{name} failed"))?;
 
@@ -71,11 +91,12 @@ fn test_case(tmpdir: &Path, file: &DirEntry, name: &str, march: &str) -> eyre::R
     }
 }
 
-fn build(tmpdir: &Path, src: &Path, march: &str) -> eyre::Result<PathBuf> {
+fn build(tmpdir: &Path, src: &Path, size: u8, march: &str) -> eyre::Result<PathBuf> {
     let out_path = tmpdir.join(Path::new(src.file_name().unwrap()).with_extension(""));
 
     let mut cmd = std::process::Command::new("clang");
-    cmd.args(["-target", "riscv32-unknown-none-elf", "-nostdlib"]);
+    cmd.args(["-target", &format!("riscv{size}-unknown-none-elf")]);
+    cmd.arg("-nostdlib");
     cmd.arg(format!("-march={march}"));
     cmd.arg(src);
     cmd.arg("-o");