Skip to content

Pseudo Assembly

This is the IR generated by the compiler for further compilation into byte code

1. PA Syntax

  1. Addressed Instruction \({\tt li} \rightarrow {\tt l : i}\)
  2. Addressed Instructions \({\tt lis} \rightarrow {\tt li}\ |\ {\tt li\ lis}\)
  3. Address \({\tt l} \rightarrow 1\ |\ 2\ |\ ...\)

  4. Instruction:

    \[{\tt i} \rightarrow \quad {\tt d} \leftarrow {\tt s}\ |\ {\tt d} \leftarrow {\tt s\ op\ s}\ | \ {\tt ret}\ | \ {\tt ifn\ s\ goto\ l}\ |\ {\tt goto\ l}\ |\ {\tt push\ s}\ |\ {\tt pop\ d}\ |\ {\tt label}\ |\ {\tt jmp\ label}\ |\ {\tt d \leftarrow icast\ s}\ |\ {\tt d \leftarrow fcast\ s}\ |\ {\tt d \leftarrow scast\ s}\ |\ {\tt d \leftarrow malloc\ n}\ |\ {\tt free\ s}\]

  5. Operator:

    \[{\tt op} \rightarrow {\tt +\ |\ -\ |\ *\ |\ /\ |\ \%\ |\ <\ |\ >\ |\ <=\ |\ >=\ |\ ==\ |\ !=\ |\ \&\&\ |\ \texttt{||}}\]
    • Note: there are multiple variants of each operator for each type of input they support, denoted by a subscript: \({\tt op_i, op_f, op_b, op_s, op_v}\) for \({\tt int}\), \({\tt float}\), \({\tt bool}\), \({\tt string}\), and \({\tt vector}\) respectively. See ยง 2.5. Operations under Type Checking for details.

  6. Operand \({\tt d, s} \rightarrow {\tt c\ |\ r\ |\ v\ |\ mem[s]}\)

  7. Constant \({\tt c} \rightarrow {\tt LIT}\)
  8. Register \({\tt r} \rightarrow {\tt r_i}\)
  9. Identifier \({\tt v} \rightarrow {\tt x\ |\ y\ |\ ...}\)

Notes:

  • \({\tt string}\)s are implemented as a pointer to a memory address with a \({\tt u32}\) length prefix followed by that many \({\tt u8}\) values
  • \({\tt vector}\)s are implemented as a pointer to a memory address with three \({\tt f32}\) values for \(x\), \(y\), and \(z\) respectively

2. PA Semantics & aliases

  • PA assumes the availability of an infinite amount of identifiers (and memory) which are used like registers. Register allocation is deferred until bytecode generation.
  • The availability of 32 enumerated 32-bit registers is assumed. The following aliases are used:
  • \({\tt r_{ret} = r_0}\): return value register
  • \({\tt r_{sp} = r_{27}}\): stack pointer
  • \({\tt r_{bp} = r_{28}}\): base pointer
  • \({\tt r_{lp} = r_{29}}\): link pointer
  • \({\tt r_{xp} = r_{30}}\): exception pointer
  • The value of \({\tt r_{31}}\) is read-only and always \(0\)
  • \({\tt alloc\ n}\) is a mnemonic alias for \({\tt r_{sp} \leftarrow r_{sp} + n}\)
  • \({\tt dealloc\ n}\) is a mnemonic alias for \({\tt r_{sp} \leftarrow r_{sp} - n}\)
  • A stack is maintained to support procedures.
  • A heap is maintained for dynamic memory allocation.
  • The \({\tt goto}\) instruction simply jumps to an address, whereas the \({\tt jmp}\) instruction jumps to a label and writes the current address to \({\tt r_{lp}}\)
  • \({\tt ret}\) jumps to the address in \({\tt r_{lp}} + 1\)