Skip to content

Latest commit

 

History

History
263 lines (185 loc) · 7.79 KB

spec.md

File metadata and controls

263 lines (185 loc) · 7.79 KB
tags: Alucard

Alucard Specification

This document lays out the specification and documentation for the various items found in the Alucard(Alu) programming language.

This document also serves as a companion piece to:

  • A detailed Syntax Guide.
    • This piece gives an interactive tutorial to the langugage, meaning that every piece of code in the tutorial can be run in the Alucard REPL.
    • The syntax guide introduces new concepts and continuously builds on new ones as the tutorial goes on.
    • The loadable file of the tutorial can be found here in the docs folder.
  • A Reference Manual
    • This document gives a more practical overview of the functions given by Alucard.

Interface to Alucard

In order to give users of Alucard flexiblity, Alucard can be used either in a traditional batch compiler or as a fully interactive system.

Scripting Mode (PARTIALLY IMPLEMENTED)

Scripting mode is the default behavior when launching the alucard program. It gives the user an interface to input any valid alu expression.

Sadly out of the box the user interface (hereby known as the REPL) does not support very good editor or auto-completion facilities. To alleviate this we recommend:

  1. Launching Alucard in your editor of choice (vscode, vim, emacs, etc.).
  2. Running the cli in a program like rlwrap.

The Alucard user interface is in reality a thin sheen over the Common Lisp REPL interface. Therefore you are able to reuse the tooling that comes with a Common Lisp system.

Since the Common Lisp system is an interactive system, everytime one uses editor integration to interact with the compiler, we are using the scripting mode of the system. From this scripting mode, the user can write the same program they would have in batch compilation mode. Further, since the tooling is integrated with one's editor they can experiment with:

  1. Running tests
  2. Having variations of the same circuit to optimize the performance or to more rapidly investigate new ideas
  3. Having graphical outputs of the circuit live as they edit
  4. Have multiple different strategies to compare
  5. etc.

Batch Compilation Mode (PARTIALLY IMPLEMENTED)

The system can easily be used in batch mode, meaning that you can give the compiler a file with the -i flag and tell it to produce a vampir file with the -o flag.

 % ./build/alu.image -i alu/example.lisp -o alu/exampale.vampir

If the -o flag is not given, then it will go into scripting mode.

Data Types

Alucard provides a variety of types, which can be broken up into three broad groups:

  1. Numerical types
  2. Arrays
  3. User defined data types

Numeric Types

Numerical types are all variations on the finite field elements.

Fields Elements

The Field data type is intended to represent a positive element of a finite field. The value can be in [0, p - 1] where p is a large prime number.

Overflow is at p and thus 0 - 1 ≡ p

Integers

The Integer type is a constrained field element to be within some bound.

For example if we have a circuit where we have x of type (int 32), then x can have values between 0 and 2^32 - 1 ([0, 2^32 - 1]).

To make working with standard integer types easier the following are defined as short hand:

  • int64
  • int32
  • int16
  • int8
  • bool ; To be thought of as (int 2)

With the more general form being invoked like (int k) where k is some natural number:

  • (int 9) ; for integers [0, 2^32 - 1]
  • (int 63)
Booleans

Boolean types can be thought of as a field element that is constrained to either be 0 for false, and 1 for true.

Arrays (PARTIALLY IMPLEMENTED)

Arrays are a fixed length data type in Alucard. Arrays are defined in terms of both the field type, and the size of the Array. Because the type is defined over the size, arrays of different sizes are considered different types and are incompatible.

Arrays have a lookup operation, and are indexed from 0.

Although Alucard functions take arrays of a specific size, generic functions over any size can be defined over them on the Common Lisp side. This is often seen in functions like map that can take any size array.

An example definition is shown below for map:

;; how do we determine the type of new-arr?
;; seems we need the type of the output of the lambda
(defun map (lamb array)
  (def ((length  (arr-length array))
        (new-arr (make-array :length length)))
    (dotimes (i length)
      (= (lookup new-array i)
         (funcall lamb (lookup array i))))
    new-arr))

User Defined Types

Users can define more complex types with the deftype construct.

Structs

Structs are very much like structs in the C programming language. They are defined with a specified type and name. The name compiles to a record lookup function, meaning that we can lookup the field by simply stating (name struct-instance).

Array Types (PARTIALLY IMPLEMENTED)

Standard Library Functions

Utilizing Common Lisp for Higher Level Code

Documentation and Specification of Specific Terms

Defcircuit

Defcircuit defines a custom gate that can be applied to other Alucard values.

(defcircuit name (<parameter>* <return>) body)

Evaluating a defcircuit causes the name to be bound as a function that is specified by the body with the parameters being in the lexical scope. The body consists of any number of expressions (zero or more). These expressions are evaluated in order and the result of the last expression is returned unless the return type specifies otherwise. The return type is currently required and specifies the type given to name. Iff the return type specifies void as the return type, then no values are returned from the function.

The complete syntax for parameterers is:

parameter ::= (<privacy> <symbol> <type>)
privacy   ::= public | private

The first value specifies whether that parameter is a public or private input field. If another circuit is specified to be compiled and that circuit calls this circuit, then the privacy of the parameters are deferred to the privacy controls of that defcircuit.

The second value specifies the name of the parameter.

The third value specifies what type the parameter is considered to be. Curently no inference is given and the type is required.

The complete syntax for returns is:

return ::= (return <type>)

The name return specifies the value as the return type.

(defcircuit poly-check ((public x int)
                        (output bool))
  (= (+ (exp x 3)
        (* 3 (exp x 2))
        (* 2 x)
        4)
     0))

In this example we define out poly-check which takes a public input x of type int and returns the type of bool.

The rest of the body is a list of expressions, in this example we simply state the relation 0 = x^3 + 3 * x^2 + 2 * x + 4.

Defgate (NOT IMPLEMENTED)

Deftype

The deftype construct is the way to make a new custom user data type. The data types

(deftype point ()
  (x int)
  (y int))

(deftype nested ()
  (plane point)
  (time  point))

In the above example we make two rec

Def

Standard Library Functions