2023-04-08 13:38:21 -04:00
<!DOCTYPE html>
< html >
< head >
2023-04-08 14:05:48 -04:00
< title > Kraken Quals Presentation< / title >
2023-04-11 23:17:55 -04:00
< link id = "theme" rel = "stylesheet" type = "text/css" href = "recursive.css" / >
<!-- <link id="theme" rel="stylesheet" type="text/css" href="slick.css"/> -->
2023-04-08 13:38:21 -04:00
< link href = "favicon.png" rel = "icon" type = "image/png" / >
< meta http-equiv = "Content-Type" content = "text/html; charset=UTF-8" / >
< style type = "text/css" >
2023-04-08 14:16:59 -04:00
body { max-width: unset; }
2023-04-11 23:17:55 -04:00
.title > h1 {
font-size: 4em;
line-height:1;
}
h1 {
line-height:1;
2023-04-24 10:48:24 -04:00
--rec-wght: 700;
//--rec-slnt: -0;
--rec-casl: 0.40;
2023-04-11 23:17:55 -04:00
}
2023-04-18 01:47:10 -04:00
.pull-left > * {
float: left;
width: 48%;
}
.pull-right > * {
float: right;
width: 48%;
}
2023-04-23 13:02:01 -04:00
.fullWidthImg > * { width: 95%; }
2023-04-08 14:16:59 -04:00
.rerun_container { position: relative; }
2023-04-23 13:02:01 -04:00
.mathSize9 { font-size: 0.9em; }
.mathSize8 { font-size: 0.8em; }
.mathSize6 { font-size: 0.6em; }
2023-04-08 13:38:21 -04:00
< / style >
< / head >
< body onload = "loadEverything();" >
< textarea id = "source" >
class: center, middle, title
# Kraken
2023-04-11 23:17:55 -04:00
_Fexprs are a better foundation for functional Lisps_
2023-04-08 13:38:21 -04:00
---
# Agenda
2023-04-25 21:58:25 -04:00
1. Lisp Background
2023-04-18 13:02:13 -04:00
2. Macro and Fexpr comparison
3. Fexpr problems
4. Past Work: Practical compilation of fexprs using partial evaluation
2023-04-25 21:58:25 -04:00
1. Kraken Language
2. Partial Evalaution
3. Optimizations
5. Current & Future Work: Scheme & the more generic re-do, Delimited Continuations, Automatic Differientation, Type Systems
2023-04-11 23:17:55 -04:00
---
class: center, middle, title
2023-04-25 21:58:25 -04:00
# Macros and Fexprs
2023-04-11 23:17:55 -04:00
2023-04-18 13:02:13 -04:00
_Lisp Background_
2023-04-11 23:17:55 -04:00
---
# Background: Lisp
Essentially every non-atomic expression is a parentheses delimited list.
This is true for:
< pre > < code class = "remark_code" > (+ 1 2) ; function calls, evaluates to 3
2023-04-25 21:58:25 -04:00
(or a b) ; macro invocations, expands to
; (let ((t a)) (if t t b))
(if false (+ 1 2) (- 1 2)) ; special forms like if
2023-04-11 23:17:55 -04:00
< / code > < / pre >
---
# Background: Lisp
One of the key hallmarks of Lisp is macros
< pre > < code class = "remark_code" > (or a b)
< / code > < / pre >
becomes
< pre > < code class = "remark_code" > (let ((temp a))
(if temp
temp
b))
< / code > < / pre >
---
2023-04-25 21:58:25 -04:00
# Background: Lisp Macros
2023-04-11 23:17:55 -04:00
2023-04-25 21:58:25 -04:00
Procedure like:
2023-04-11 23:17:55 -04:00
< pre > < code class = "remark_code" > (define-maco (or . body)
(cond
((nil? (cdr body)) (car body))
(else (list 'let (list (list 'temp (car body)))
(list 'if 'temp 'temp (car (cdr body)))))))
< / code > < / pre >
2023-04-25 21:58:25 -04:00
Pattern matching:
2023-04-11 23:17:55 -04:00
< pre > < code class = "remark_code" > (letrec-syntax
((or (syntax-rules ()
((or a) a)
((or a b)
(let ((temp a))
(if temp
temp
b)))))))
< / code > < / pre >
2023-04-08 13:38:21 -04:00
---
2023-04-25 21:58:25 -04:00
# Background: Lisp Macros
As we've mentioned, in Scheme _or_ is a macro expanding
< pre > < code class = "remark_code" > (or a b)
< / code > < / pre >
to
< pre > < code class = "remark_code" > (let ((temp a)) (if temp temp b))
< / code > < / pre >
So passing it to a higher-order function doesn't work, you have to wrap it in a function:
< pre > < code class = "remark_code" > > (fold or #f (list #t #f))
Exception: invalid syntax and
< / code > < / pre >
< pre > < code class = "remark_code" > > (fold (lambda (a b) (or a b)) #f (list #t #f))
#t
< / code > < / pre >
---
2023-04-15 15:41:11 -04:00
# Background: Fexprs
2023-04-08 13:38:21 -04:00
2023-04-18 13:02:13 -04:00
Something of a combo between the two - direct style, but naturally hygienic by default.
2023-04-11 23:17:55 -04:00
< pre > < code class = "remark_code" > (vau de (a b) (let ((temp (eval a de)))
(if temp
temp
(eval b de))))
< / code > < / pre >
2023-04-25 21:58:25 -04:00
But in Kraken, _or_ is a combiner (an operative!), so it's first-class
So it's perfectly legal to pass to a higher-order combiner:
< pre > < code class = "remark_code" > > (foldl or false (array true false))
true
< / code > < / pre >
---
# Background: Fexprs in Lisp
- 1960s - In earliest Lisps
- 1980 - Argument that they shoudln't be included due to inability for static analysis to fix optimization problems (Kent Pitman, "Special Forms in Lisp")
- 1998 - Adding fexprs to lambda calculus produces a trivial theory (Mitchell Wand, "The Theory of Fexprs is Trivial")
- 2010 - Fexprs are not trivial with internal syntax, make sense with lexical scoping (John Shutt, "Fexprs as the basis of Lisp function application or $ vau: the ultimate abstraction", also earlier papers)
---
(from Wikipedia)
.fullWidthImg[]
---
(from Wikipedia)
.fullWidthImg[]
2023-04-11 23:17:55 -04:00
---
2023-04-15 15:41:11 -04:00
# Background: Fexprs - detail
Ok, Fexprs are calls to combiners - combiners are either applicatives or operatives.
2023-04-18 13:02:13 -04:00
Combiners are introduced with _vau_ and take an extra parameter (here called _dynamic_env_, earlier called _de_) which is the dynamic environment.
2023-04-15 15:41:11 -04:00
< pre > < code class = "remark_code" > (vau dynamicEnv (normalParam1 normalParam2) (body of combiner))
< / code > < / pre >
--
Lisps, as well as Kraken, have an _eval_ function.
This function takes in code as a data structure, and in R5RS Scheme an "environment specifier", and in Kraken, a full environment (like what is passed as _dynamicEnv_).
< pre > < code class = "remark_code" > (eval some_code an_environment)
2023-04-25 21:58:25 -04:00
---
# Pros and Cons
**Pros:**
1. Vau/Combiners unify and make first class functions, macros, and built-in forms in a single simple system
2. They are also much simpler conceptually than macro systems while being hygienic by default
**Cons:**
1. The code of the operative combiner (analogus to a macro invocation) is re-executed at runtime, every time it is encountered
2. Additionally, because it is unclear what code will be evaluated as a parameter to a function call and what code must be passed unevaluated to the combiner, little optimization can be done.
2023-04-15 15:41:11 -04:00
---
# Background: Fexprs - detail
- **Normal Lisp** (Scheme, Common Lisp, etc)
- Functions - runtime, evaluate parameters once, return value
- Macros - expansion time, do not evaluate parameters, return code to be inlined
- Special Forms - look like function or macro calls, but do something special (if, lambda, etc)
--
- **Kraken** (and Kernel)
- Combiners
- Applicatives (like normal functions, combiners that evaluate all their parameters once in their dynamic environment)
- Operatives (combiners that do something unusual with their parameters, do not evaluate them right away)
---
2023-04-11 23:17:55 -04:00
# Solution: Partial Eval
2023-04-24 10:48:24 -04:00
1. Evaluate parts of program that only depend on statically-known data ahead of time and insert resulting values into generated code
2. The parts of the resulting partially-evaluated program that only contains static references to a subset of built in combiners and functions (combiners that evaluate their parameters exactly once) can be compiled just like it was a normal Scheme program
2023-04-16 00:02:29 -04:00
2023-04-25 21:58:25 -04:00
**This is novel! No one has sucessfully pulled this off, to our knowledge.**
Why?
- Can't use a binding time analysis pass with offline partial evaluation, which eliminates quite a bit of mainline partial evaluation research
- Online partial evaluation research generally does not have to deal with the same level of partially/fully dynamic first-class explicit environments
2023-04-16 00:02:29 -04:00
---
2023-04-25 21:58:25 -04:00
# Research Contributions
2023-04-16 00:02:29 -04:00
2023-04-25 21:58:25 -04:00
- Programming Language, Kraken, that is entirely based on fexprs
- Novel Partial Evaluation algorithm
2023-04-16 00:02:29 -04:00
- Heavily specialized to optimize away operative combiners like macros written in a specific way
2023-04-25 21:58:25 -04:00
- Compiler
2023-04-18 13:02:13 -04:00
- Static calls fully optimized like a normal function call in other languages
2023-04-16 00:02:29 -04:00
- Dynamic calls have a single branch of overhead - if normal applicative combiner function like call, post-branch optimized as usual
- Optimizes away Y Combinator recursion to static recursive jumps (inc tail call opt)
2023-04-25 21:58:25 -04:00
- Prototype faster than Python and other interpreted Lisps
- Paper: *Practical compilation of Fexprs using partial evaluation*
- Currently under review for ICFP '23
---
# Kraken
- Entirly based on fexprs as only means of computation
- No macros, no special forms, no seperate functions
- Not just tacked-on to an existing language, entirely different foundation
- Purely Functional
- callbacks/monad-like interaction with outside world
- In the Lisp tradition
2023-04-16 00:02:29 -04:00
---
2023-04-18 01:47:10 -04:00
# Base Language: Syntax
$$
\newcommand{\alt} {\mid}
\newcommand{\kraken} {\textit{Kraken}}
\newcommand{\kprim} [2] {\langle #1~\textbf{#2} \rangle}
\newcommand{\kenv} [3] {\langle \langle #1~|#2,~#3 \rangle \rangle}
\newcommand{\kcomb} [5] {\langle \textbf{comb} ~ #1 ~ #2 ~ #3 ~ #4 ~ #5\rangle}
\newcommand{\keval} [2] {[\text{eval} ~ #1 ~ #2]}
\newcommand{\kcombine} [3] {[\text{combine} ~ #1 ~ #2 ~ #3 ]}
$$
.mathSize8[
$$
\begin{array}{rcll}
n & \in & \mathbb{N} & \text{(Integers)} \\\
s & \in & Symbols & \\\
o & \in & \kprim{1}{eval}, \kprim{0}{vau},\kprim{1}{wrap}, \kprim{1}{unwrap}, & \\\
& & \kprim{0}{if0}, \kprim{0}{vif0}, \kprim{1}{int-to-symbol},& \\\
& & \kprim{1}{symbol?}, \kprim{1}{int?}, \kprim{1}{combiner?},\kprim{1}{env?},& \\\
& & \kprim{1}{array?}, \kprim{1}{len}, \kprim{1}{idx}, \kprim{1}{concat},& \\\
& & \kprim{1}{+}, \kprim{1}{< =} & \text{(Primitives)}\\\
E & :=& \kenv{(s \leftarrow T)\dots}{}{E} \alt \kenv{(s \leftarrow T)\dots}{s' \leftarrow E}{E} & \text{(Environments)}\\\
A & :=& (T \dots)& \text{(Arrays)}\\\
C & :=& \kcomb{n}{s'}{E}{(s\dots)}{T} & \text{(Combiners)}\\\
S & :=& n \alt o \alt E \alt C & \text{(Self evaluating terms)}\\\
V & :=& S \alt s \alt A & \text{(Values)}\\\
T & :=& V \alt AT & \text{(Terms)}\\\
AT & :=& \keval{T}{E} \alt \kcombine{T}{(T\dots)}{E} & \text{(Active terms)}\\\
\end{array}
$$
]
2023-04-16 00:02:29 -04:00
---
2023-04-18 01:47:10 -04:00
# Base Language: Contexts
$$
\newcommand{\Ctxt} {\mathcal{E}}
\newcommand{\InCtxt} [1] {\Ctxt[#1]}
$$
.mathSize8[
$$
\begin{array}{rcl}
\Ctxt & :=& \square \alt \kcombine{\Ctxt}{(T\dots)}{E} \alt \kcombine{T}{(\Ctxt,T\dots)}{E}\\\
& & \alt \kcombine{T}{(T\dots,\Ctxt,T\dots)}{E} \alt \kcombine{T}{(T\dots,\Ctxt)}{E}\\\
\end{array}
$$
]
2023-04-16 00:02:29 -04:00
2023-04-18 01:47:10 -04:00
---
# Base: Small-Step Semantics
.mathSize8[
$$
\begin{array}{rcl}
\InCtxt{E} & \rightarrow& \InCtxt{E'} ~ (\text{if } E \rightarrow E')\\\
\keval{S}{E} & \rightarrow& S\\\
\keval{s}{E} & \rightarrow& lookup(s,E)\\\
\keval{(T_1~T_2\dots)}{E} & \rightarrow& \kcombine{\keval{T_1}{E}}{(T_2\dots)}{E}\\\
\\\
\kcombine{\kcomb{(S~n)}{s'}{E'}{(s\dots)}{Tb}}{(V\dots)}{E} & \rightarrow& \kcombine{\kcomb{n}{s'}{E'}{s}{Tb}}{\\\& & \keval{V}{E}\dots}{E}\\\
\kcombine{\kcomb{0}{s'}{E'}{(s\dots)}{Tb}}{(V\dots)}{E} & \rightarrow& \keval{Tb}{\kenv{(s \leftarrow V)\dots}{s' \leftarrow E}{E'}}\\\
\\\
\kcombine{\kprim{(S~n)}{o}}{(V\dots)}{E} & \rightarrow& \kcombine{\kprim{n}{o}}{(\keval{V}{E}\dots)}{E}\\\
\end{array}
$$
]
2023-04-16 00:02:29 -04:00
2023-04-11 23:17:55 -04:00
---
2023-04-18 01:47:10 -04:00
# Base: Selected Primitives
.mathSize8[
$$
\begin{array}{rcl}
\kcombine{\kprim{0}{eval}}{(V~E')}{E} & \rightarrow& \keval{V}{E'}\\\
\kcombine{\kprim{0}{vau}}{(s'~(s\dots)~V)}{E} & \rightarrow& \kcomb{0}{s'}{E}{(s\dots)}{V}\\\
\kcombine{\kprim{0}{wrap}}{\kcomb{0}{s'}{E'}{(s\dots)}{V}}{E} & \rightarrow& \kcomb{1}{s'}{E'}{(s\dots)}{V}\\\
\kcombine{\kprim{1}{unwrap}}{\kcomb{1}{s'}{E'}{(s\dots)}{V}}{E} & \rightarrow& \kcomb{0}{s'}{E'}{(s\dots)}{V}\\\
2023-04-23 13:02:01 -04:00
\kcombine{\kprim{0}{if0}}{(V_c~V_t~V_e)}{E} & \rightarrow& \kcombine{\kprim{0}{vif0}}{\\\& & (\keval{V_c}{E}~V_t~V_e)}{E}\\\
2023-04-18 01:47:10 -04:00
\kcombine{\kprim{0}{vif0}}{(0~V_t~V_e)}{E} & \rightarrow& \keval{V_t}{E}\\\
\kcombine{\kprim{0}{vif0}}{(n~V_t~V_e)}{E} & \rightarrow& \keval{V_e}{E} ~\text{(n != 0)}\\\
\kcombine{\kprim{0}{int-to-symbol}}{(n)}{E} & \rightarrow& 'sn ~\text{(symbol made out of the number n)}\\\
\kcombine{\kprim{0}{array}}{(V\dots)}{E} & \rightarrow& (V\dots)\\\
\end{array}
$$
]
2023-04-11 23:17:55 -04:00
---
2023-04-18 01:47:10 -04:00
# Base Language Summary
- This base calculus defined above is not only capable of normal lambda-calculus computations with primitives and derived user applicatives, but also supports a superset of macro-like behaviors via its support for operatives.
- All of the advantages listed in the introduction apply to this calculus, as do the performance drawbacks, at least if implemented naively. Our partial evaluation and compilation framework will demonstrate how to compile this base language into reasonably performant binaries (WebAssembly bytecode, for our prototype).
2023-04-11 23:17:55 -04:00
---
2023-04-18 01:47:10 -04:00
class: center, middle, title
# Slow
2023-04-11 23:17:55 -04:00
---
# Partial Eval: How it works
2023-04-18 01:47:10 -04:00
2023-04-25 21:58:25 -04:00
Marco expansion kind of *is* partial evaluation!
2023-04-18 01:47:10 -04:00
- Online, no binding time analysis
- Partially Evaluate combiners with partially-static environments
- Prevent infinate recursion by blocking on
- Recursive calls underneath a partially evaluated body
- Recursive path to *if*
- Track call frames that need to be real to progress on every AST node
- Can zero-in on areas that will make progress
- Also tracks nodes previously stopped by recursion-stopper in case no longer under the frame that stopped the recursion
- Evaluate derived calls with parameter values, inline result even if not value if it doesn't depend on call frame
---
2023-04-25 21:58:25 -04:00
# Partial Eval Rule Walkthrough
2023-04-18 01:47:10 -04:00
---
class: center, middle, title
# Optimizations
---
2023-04-25 21:58:25 -04:00
# Optimization Adgenda
- "The Trick" - handling dynamic calls
- Lazy Environment Instantiation
- Type-Inference-Based Primitive Inlining
- Immediately-Called Closure Inlining
- Y-Combinator Elimination
---
# "The Trick"
2023-04-18 01:47:10 -04:00
< pre > < code class = "remark_code" > (lambda (f) (f (+ 1 2)))
< / code > < / pre >
To something like
< pre > < code class = "remark_code" > function(f):
if wrap_level(f) == 1:
f(3)
else:
f([`+ 1 2])
< / code > < / pre >
- Insert runtime check for dynamic call sites
- When compiling in the wraplevel=1 side of conditional, further partial evaluate the parameter value
- Only a single branch of overhead for dynamic function calls
2023-04-11 23:17:55 -04:00
2023-04-25 21:58:25 -04:00
*Really a critical part of the dance, not an optional optimization*
2023-04-11 23:17:55 -04:00
---
2023-04-18 01:47:10 -04:00
# Lazy Environment Instantiation
< pre > < code class = "remark_code" > (lambda (f) (f))
< / code > < / pre >
2023-04-24 10:48:24 -04:00
compiled to equivalent of
2023-04-18 01:47:10 -04:00
< pre > < code class = "remark_code" > function(f):
if uses_env(f):
2023-04-24 10:48:24 -04:00
if not env_cache:
env_cache = make_env()
2023-04-18 01:47:10 -04:00
f(env_cache)
else:
f()
< / code > < / pre >
---
# Type-Inference-Based Primitive Inlining
For instance, consider the following code:
2023-04-25 21:58:25 -04:00
< pre > < code class = "remark_code" > (if (and (array? a) (= 3 (len a))) (idx a 2)
nil)
2023-04-18 01:47:10 -04:00
< / code > < / pre >
- Call to *idx* fully inlined without type or bounds checking
- No type information is needed to inline type predicates, as they only need to look at the tag bits.
- Equality checks can be inlined as a simple word/ptr compare if any of its parameters are of a type that can be word/ptr compared (ints, bools, and symbols).
2023-04-25 21:58:25 -04:00
< pre > < code class = "remark_code" > if is_array(a) and len(a) == 3:
*(a+2)
else: nil
< / code > < / pre >
2023-04-18 01:47:10 -04:00
---
# Immediately-Called Closure Inlining
Inlining calls to closure values that are allocated and then immediately used:
2023-04-25 21:58:25 -04:00
This is partial-evaled
< pre > < code class = "remark_code" > (let (b (+ a 2))
(+ b 3))
2023-04-18 01:47:10 -04:00
< / code > < / pre >
to this
2023-04-25 21:58:25 -04:00
< pre > < code class = "remark_code" > ((wrap (vau (b) (+ b 3))) (+ a 2))
2023-04-18 01:47:10 -04:00
< / code > < / pre >
and then inlined (plus lazy environment allocation)
2023-04-25 21:58:25 -04:00
< pre > < code class = "remark_code" > b = a + 2;
b + 3
< / code > < / pre >
2023-04-18 01:47:10 -04:00
---
# Y-Combinator Elimination
2023-04-25 21:58:25 -04:00
< pre > < code class = "remark_code" > (Y (lambda (self) (lambda (n)
(if (= 0 n) 1
(* n (self (- n 1)))))))
< / code > < / pre >
2023-04-18 01:47:10 -04:00
- When compiling a combiner, pre-emptive memoization
- Partial-evaluation to normalize
- Eager lang - extra lambda - eta-conversion in the compiler
2023-04-25 21:58:25 -04:00
< pre > < code class = "remark_code" > def fact(n):
if n == 0:
1
else:
n * fact(n - 1)
< / code > < / pre >
2023-04-18 01:47:10 -04:00
---
2023-04-24 10:48:24 -04:00
# Outcomes
1. All macro-like combiner calls are partially evaluated away
2023-04-25 21:58:25 -04:00
- No downside to using fexprs for everything
2023-04-24 10:48:24 -04:00
2. No interpreted evaluation calls remain
3. Optimizations allow reasonable performance
2023-04-11 23:17:55 -04:00
---
2023-04-18 01:47:10 -04:00
# Benchmarks
- Fib - Calculating the nth Fibonacci number
- RB-Tree - Inserting n items into a red-black tree, then traversing the tree to sum its values
- Deriv - Computing a symbolic derivative of a large expression
- Cfold - Constant-folding a large expression
- NQueens - Placing n number of queens on the board such that no two queens are diagonal, vertical, or horizontal from each other
2023-04-25 21:58:25 -04:00
There isn't a standard suite to use here, this was a set we liked from a previosu paper that is relevent for both functional and imperitive languages.
2023-04-18 01:47:10 -04:00
---
# Results:
Number of eval calls with no partial evaluation for Fexprs
.mathSize8[
$$
\begin{array}{||c | c c c c c ||}
\hline
& Evals & Eval w1 Calls & Eval w0 Calls & Comp Dyn & Comp Dyn\\\
& & & & w1 Calls & w0 Calls\\\
\hline\hline
Cfold 5 & 10897376 & 2784275 & 879066 & 1 & 0 \\\
\hline
Deriv 2 & 11708558 & 2990090 & 946500 & 1 & 0 \\\
\hline
NQueens 7 & 13530241 & 3429161 & 1108393 & 1 & 0 \\\
\hline
Fib 30 & 119107888 & 30450112 & 10770217 & 1 & 0 \\\
\hline
RB-Tree 10 & 5032297 & 1291489 & 398104 & 1 & 0 \\\
\hline
\end{array}
$$
]
Number of eval calls in Partially Evaluated Fexprs
.mathSize8[
$$
\begin{array}{||c | c c c c c ||}
\hline
& Evals & Eval w1 Calls & Eval w0 Calls & Comp Dyn & Comp Dyn\\\
& & & & w1 Calls & w0 Calls\\\
\hline\hline
Cfold 5 & 0 & 0 & 0 & 0 & 0 \\\
\hline
Deriv 2 & 0 & 0 & 0 & 2 & 0 \\\
\hline
NQueens 7 & 0 & 0 & 0 & 0 & 0 \\\
\hline
Fib 30 & 0 & 0 & 0 & 0 & 0 \\\
\hline
RB-Tree 10 & 0 & 0 & 0 & 10 & 0 \\\
\hline
\end{array}
$$
]
---
# Results:
Number of calls to the runtime's eval function for RB-Tree. The table shows the non-partial evaluation numbers -> partial evaluation numbers.
.mathSize8[
$$
\begin{array}{||c | c c c c c ||}
\hline
& Evals & Eval w1 Calls & Eval w0 Calls & Comp Dyn & Comp Dyn\\\
& & & & w1 Calls & w0 Calls\\\
\hline\hline
RB-Tree 7 & 2952848 -> 0 & 757932 -> 0 & 233513 -> 0 & 1 -> 7 & 0 -> 0\\\
\hline
RB-Tree 8 & 3532131 -> 0 & 906548 -> 0 & 279379 -> 0 & 1 -> 8 & 0 -> 0\\\
\hline
RB-Tree 9 & 4278001 -> 0 & 1097965 -> 0 & 3383831 -> 0 & 1 -> 9 & 0 -> 0\\\
\hline
\end{array}
$$
]
---
# Results:
2023-04-24 10:48:24 -04:00
.fullWidthImg[]
2023-04-18 01:47:10 -04:00
---
# Results:
2023-04-24 10:48:24 -04:00
.fullWidthImg[]
2023-04-18 01:47:10 -04:00
---
# Results:
2023-04-24 10:48:24 -04:00
.fullWidthImg[]
2023-04-18 01:47:10 -04:00
---
# Results:
2023-04-24 10:48:24 -04:00
.fullWidthImg[]
2023-04-18 01:47:10 -04:00
---
2023-04-24 10:48:24 -04:00
# Results: (log scale)
.fullWidthImg[]
---
# Results:
.fullWidthImg[]
---
# Results:
.fullWidthImg[]
2023-04-18 01:47:10 -04:00
---
2023-04-25 21:58:25 -04:00
# Research Conclusion
- Purely functional Lisp based entirely on Fexprs - Kraken
- Novel Partial Evaluation algorithm specifically designed to remove all macro-like fexprs
- No interpreted evaluation calls remain
- Compiler implementing partial evaluation and additional optimizations achiving feasible performance (70,000x+ speedup, etc)
*Little to no downside to basing Lisp-like functional language on fexprs*
---
2023-04-24 10:48:24 -04:00
# Current & Future
2023-04-18 01:47:10 -04:00
- Partial Evaluation / Kraken evolution:
2023-04-24 10:48:24 -04:00
- More normal language: purely functional Scheme
2023-04-18 01:47:10 -04:00
- More standard Scheme
- Environments as association-lists
2023-04-24 10:48:24 -04:00
- Fully manipulateable as normal list/pairs
- Partial evaluation that supports naturally-written operative combiners, like the running *or* example
- Performance: Better Reference Counting, Tail-Recursion Modulo Cons
2023-04-18 01:47:10 -04:00
- Implement Delimited Continuations as Fexprs
2023-04-24 10:48:24 -04:00
- Implement Automatic Differentiation as Fexprs
- Allow type systems to be built using Fexprs, like the type-systems-as-macros paper
2023-04-25 21:58:25 -04:00
- Investigate Hardware as Fexprs
---
class: center, middle, title
# Thank you!
2023-04-18 01:47:10 -04:00
---
class: center, middle, title
# Backup Slides
2023-04-24 10:48:24 -04:00
---
2023-04-25 21:58:25 -04:00
# Partial Eval Semantics:
.pull-left[]
.pull-right[]
---
# Partial Eval Semantics:
.pull-left[]
.pull-right[]
---
# Partial Eval Semantics:
.pull-left[]
.pull-right[]
---
# Background: Fexprs - detail
< pre > < code class = "remark_code" > foldr:
(rec-lambda recurse (f z l)
(if (= nil l)
z
(lapply f (list (car l) (recurse f z (cdr l))))))
< / code > < / pre >
(lapply reduces the wrap-level of the function by 1, equivalent to quoting the inputs)
< pre > < code class = "remark_code" > foldr:
(rec-lambda recurse (f z l)
(if (= nil l)
z
(f (car l) (recurse f z (cdr l)))))
< / code > < / pre >
---
2023-04-24 10:48:24 -04:00
# Background: Fexprs - detail
All special forms in Kaken are combiners too, and are thus also first class.
In this case, we can not only pass the raw _if_ around, but we can make an _inverse_if_ which inverts its condition (kinda macro-like) and pass it around.
< pre > < code class = "remark_code" > > (let ((use_if (lambda (new_if) (new_if true 1 2)))
(inverse_if (vau de (c t e) (if (not (eval c de))
(eval t de)
(eval e de))))
)
(list (use_if if) (use_if inverse_if)))
(1 2)
< / code > < / pre >
What were special forms in Lisp are now just built-in combiners in Kraken.
*if* is not any more special than *+*, and in both cases you can define your own versions that would be indistinguishable, and in both cases they are first-class.
---
# Solution: Partial Eval
1. Partially evaluate a purely functional version of this language in a nearly-single pass over the entire program
2. Environment chains consisting of both "real" environments with every contained symbol mapped to a value and "fake" environments that only have placeholder values.
4. The parts of the resulting partially-evaluated program that only contains static references to a subset of built in combiners and functions (combiners that evaluate their parameters exactly once) can be compiled just like it was a normal Scheme program
---
2023-04-25 21:58:25 -04:00
# Intuition
Macros, especially *define-macro* macros, are essentially functions that run at expansion time and compute new code from old code.
This is essentially partial evaluation / inlining, depending on how you look at it.
It thus makes sense to ask if we can identify and partial evaluate / inline operative combiners to remove and optimize them like macros. Indeed, if we can determine what calls are to applicative combiners we can optimize their parameters, and if we can determine what calls are to macro-like operative combiners, we can try to do the equivalent of macro expansion.
For Kraken, this is exactly what we do, using a specialized form of Partial Evaluation to do so.
---
2023-04-24 10:48:24 -04:00
# Selected Explanations
.mathSize9[
- \\(\kprim{0}{eval}\\): evaluates its argument in the given environment.
- \\(\kprim{0}{vau}\\): creates a new combiner and is analogous to lambda in other languages, but with a "wrap level" of 0, meaning the created combiner does not evaluate its arguments.
- \\(\kprim{0}{wrap}\\): increments the wrap level of its argument. Specifically, we are "wrapping" a "wrap level" n combiner (possibly "wrap level" 0, created by *vau* to create a "wrap level" n+1 combiner. A wrap level 1 combiner is analogous to regular functions in other languages.
- \\(\kprim{0}{unwrap}\\): decrements the "wrap level" of the passed combiner, the inverse of *wrap*.
- \\(\kprim{0}{if}\\): evaluates only its condition and converts to the \\(\kprim{0}{vif}\\) primitive for the next step. It cannot evaluate both branches due to the risk of non-termination.
- \\(\kprim{0}{vif}\\): evaluates and returns one of the two branches based on if the condition is non-zero.
- \\(\kprim{0}{int-to-symbol}\\): creates a symbol out of an integer.
- \\(\kprim{0}{array}\\): returns an array made out of its parameter list.
]
---
# Less Interesting Prims
.mathSize9[
- \\(\kcombine{\kprim{0}{type-test?}}{(A)}{E}\\): *array?*, *comb?*, *int?*, and *symbol?*, each return 0 if the single argument is of that type, otherwise they return 1.
- \\(\kcombine{\kprim{0}{len}}{(A)}{E}\\): returns the length of the single array argument.
- \\(\kcombine{\kprim{0}{idx}}{(A~n)}{E}\\): returns the nth item array A.
- \\(\kcombine{\kprim{0}{concat}}{(A~B)}{E}\\): combines both array arguments into a single concatenated array.
- \\(\kcombine{\kprim{0}{+}}{(A~A)}{E}\\): adds its arguments
- \\(\kcombine{\kprim{0}{< =}}{(A~A)}{E}\\): returns 0 if its arguments are in increasing order, and 1 otherwise.
]
---
# Results:
.pull-left[]
.pull-right[]
2023-04-11 23:17:55 -04:00
---
2023-04-08 13:38:21 -04:00
# Introduction
2023-04-11 23:17:55 -04:00
Here's some test code:
2023-04-08 13:38:21 -04:00
.run_container[
< div class = "editor" id = "hello_editor" > ; Of course
(println "Hello World")
; Just print 3
2023-04-08 14:05:48 -04:00
(println "Math workssss:" (+ 1 2 4))
2023-04-08 13:38:21 -04:00
< / div >
]
--
2023-04-08 14:05:48 -04:00
.rerun_container[
2023-04-08 13:38:21 -04:00
< pre > < code class = "remark_code" id = "hello_editor_output" > output here...< / code > < / pre >
2023-04-08 14:05:48 -04:00
< button class = "run_button" onclick = "executeKraken(hello_editor_jar[1].toString(), 'hello_editor_output')" > Rerun< / button > < br >
]
2023-04-08 13:38:21 -04:00
< / textarea >
< link rel = "stylesheet" href = "./default.min.css" >
< script src = "./highlight.min.js" > < / script >
< script type = "module" >
import {CodeJar} from './codejar.js'
window.loadEverything = function() {
2023-04-18 01:47:10 -04:00
var renderMath = function() {
//renderMathInElement(document.body);
//renderMathInElement(document.body, {delimiters: [
// {left: "$$", right: "$$", display: true},
// {left: "$", right: "$", display: false},
// {left: "\\[", right: "\\]", display: true},
// {left: "\\(", right: "\\)", display: false},
//]});
}
//var slideshow = remark.create({}, renderMath);
2023-04-08 13:38:21 -04:00
var slideshow = remark.create();
document.querySelectorAll('.editor').forEach((editor_div) => {
2023-04-08 14:05:48 -04:00
if (window[editor_div.id + "_jar"] == undefined) {
window[editor_div.id + "_jar"] = []
}
window[editor_div.id + "_jar"].push(CodeJar(editor_div, hljs.highlightElement))
2023-04-08 13:38:21 -04:00
});
slideshow.on('showSlide', function (slide) {
2023-04-08 14:05:48 -04:00
//console.log('Navigated to', slide)
for (const c of slide.content) {
if (c.class == "run_container") {
//console.log("found editor", c)
const re = /class="editor" id="([^"]+)"/;
let id = c.content[0].match(re)[1]
let editors = window[id + "_jar"]
if (slide.properties.continued == "true") {
editors[1].updateCode(editors[0].toString())
//console.log("Got editors", editors, "running kraken")
executeKraken(editors[1].toString(), id + "_output")
} else {
editors[0].updateCode(editors[1].toString())
}
}
2023-04-08 13:38:21 -04:00
}
})
2023-04-18 01:47:10 -04:00
MathJax.Hub.Config({
tex2jax: {
skipTags: ['script', 'noscript', 'style', 'textarea', 'pre'],
}
});
MathJax.Hub.Configured();
2023-04-08 13:38:21 -04:00
}
< / script >
< script >
var output_name = ""
var Module = {
noInitialRun: true,
onRuntimeInitialized: () => {
},
print: txt => {
document.getElementById(output_name).innerHTML += txt + "\n";
},
printErr: txt => {
document.getElementById(output_name).innerHTML += "STDERR:[" + txt + "]\n";
}
};
function executeKraken(code, new_output_name) {
output_name = new_output_name
2023-04-08 14:05:48 -04:00
document.getElementById(new_output_name).innerHTML = "";
2023-04-08 13:38:21 -04:00
Module.callMain(["-C", code]);
}
< / script >
< script type = "text/javascript" src = "k_prime.js" > < / script >
2023-04-08 14:05:48 -04:00
< script src = "remark-latest.min.js" > < / script >
2023-04-18 01:47:10 -04:00
<!-- <script src="MathJax.js"></script> -->
< script src = "https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.5/MathJax.js?config=TeX-AMS_HTML&delayStartupUntil=configured" type = "text/javascript" > < / script >
2023-04-08 13:38:21 -04:00
< / body >
< / html >
