Lambda the Ultimate

Compiling a Subset of APL Into a Typed Intermediate Language

by Martin Elsman, Martin Dybdal

Traditionally, APL is an interpreted language ... In this paper, we present a compiler that compiles a subset of APL into a typed intermediate representation, which should serve as a practical and well-defined intermediate format for targeting parallel-architectures through a large number of existing tools and frameworks. The intermediate language is conceptually close to the language Repa. It supports shape-polymorphic functions and types that classify shapes. The compiler takes a simplified approach to certain aspects of APL. Following other APL compilation approaches, the compiler is based on lexical (i.e., static) identifier scoping and has no support for dynamic compilation (APL execute).

Terseness of APL is legendary, for good or bad. I keep finding more and more papers by Haskell community (and especially GHC contributors) working on efficient (parallel) arrays in Haskell.

There is a growing set of fascinating interviews with PL folks at People of Programming Languages.

Exploiting Vector Instructions with Generalized Stream Fusion

By Geoffrey Mainland, Roman Leshchinskiy, and Simon Peyton Jones.

Our ideas are implemented in modified versions of the GHC compiler and vector library. Benchmarks show that high-level Haskell code written using our compiler and libraries can produce code that is faster than both compiler- and hand-vectorized C.

This paper continues the promising line of research started in 1990 by Wadler (at least, that was how I learned of deforestation). Of course, there was a lot of development since then, but this specific paper introduces an interesting idea of multiple representations - potentially changing the game.

Partial sharing graphs offer a reduction model for the lambda calculus that is optimal in a sense put forward by Jean Jacques Levy. This model has seen interest wax and wane: initially it was thought to offer the most efficient possible technology for implementing the lambda calculus, but then an important result showed that bookkeeping overheads of any such model could be very high (Asperti & Mairson 1998). This result had a chilling effect on the initial wave of excitement over the technology.

Now Stefano Guerrini, one of the early investigators of partial sharing graphs, has an article with Marco Solieri (Guerrini & Solieri 2017) arguing that the gains from optimality can be very high and that partial sharing graphs can be relatively close to the best possible efficiency, within a quadratic factor on a conservative analysis (this is relatively close in terms of elementary recursion). Will the argument and result lead to renewed interest in partial sharing graphs from implementors of functional programming? We'll see...

(Asperti & Mairson 1998) Parallel beta reduction is not elementary recursive.

(Guerrini & Solieri 2017) Is the Optimal Implementation inefficient? Elementarily not.

No! (As Simon Thompson explains.)
You cannot not love the "exploration of the length function" at the bottom. Made me smile in the middle of running errands.

Found on the ARM community's embedded blog. It seems that Forth may be making a comeback.
"8th" - a gentle introduction to a modern Forth

8th is a secure, cross-platform programming language based on Forth which lets you concentrate on your application’s logic instead of worrying about differences between platforms. It lets you write your code once, and simultaneously produce applications running on multiple platforms. Its built-in encryption helps protect your application from hackers. Its interactive nature makes debugging and testing your code much easier.

As of this writing it supports 32-bit and 64-bit variants of:

• Windows, macOS and Linux for desktop or server systems
• Android (32-bit Arm only) and iOS for mobile systems
• Raspberry Pi (Raspbian etc) for embedded Linux Arm-based systems.

...
8th differs from more traditional Forths in a number of ways. First of all, it is strongly typed and has a plethora of useful types (dynamic strings, arrays, maps, queues and more).

Other differences from traditional Forth appear to include automatic memory management and some kind of signed and encrypted application deployment.

[Edit: per gasche's comment, please note that 8th appears to be closed source. From their FAQ:

"Is 8th a GPL-Licensed product? No, it is a commercial product. None of the libraries it uses are under the GPL or LGPL. Due to the desire for security, 8th includes its required libraries in the binary, and the GPL family of licenses is therefore not appropriate."

Let the arguments about the effectiveness of security-by-obscurity begin. Source is apparently available if you buy an Enterprise license and sign an NDA.]

Just saw this on Hacker News -- Project Loom: Fibers and Continuations for the Java Virtual Machine with the following overview:

Project Loom's mission is to make it easier to write, debug, profile and maintain concurrent applications meeting today's requirements. Threads, provided by Java from its first day, are a natural and convenient concurrency construct (putting aside the separate question of communication among threads) which is being supplanted by less convenient abstractions because their current implementation as OS kernel threads is insufficient for meeting modern demands, and wasteful in computing resources that are particularly valuable in the cloud. Project Loom will introduce fibers as lightweight, efficient threads managed by the Java Virtual Machine, that let developers use the same simple abstraction but with better performance and lower footprint. We want to make concurrency simple(r) again! A fiber is made of two components — a continuation and a scheduler. As Java already has an excellent scheduler in the form of ForkJoinPool, fibers will be implemented by adding continuations to the JVM.

I'm a fan of fibers, and this has quite a bit of interesting material in it for like-minded folks.

Copattern matching and first-class observations in OCaml, with a macro, by Paul Laforgue and Yann Regis-Gianas:

Infinite data structures are elegantly defined by means of copattern matching, a dual construction to pattern matching that expresses the outcomes of the observations of an infinite structure. We extend the OCaml programming language with copatterns, exploiting the duality between pattern matching and copattern matching. Provided that a functional programming language has GADTs, every copattern matching can be transformed into a pattern matching via a purely local syntactic transformation, a macro. The development of this extension leads us to a generalization of previous calculus of copatterns: the introduction of first-class observation queries. We study this extension both from a formal and practical point of view.

Codata isn't well supported enough in languages IMO, and supporting it via a simple macro translation will hopefully make it easier to adopt, at least for any language with GADTs.

Non-determinism: a sublanguage rather than a monad

A puzzlingly named, exceedingly technical device introduced to structure the denotational semantics has by now achieved cult status. It has been married to effects -- more than once. It is compulsively looked for in all manner of things, including burritos. At least two ICFP papers brought it up without a rhyme or reason (or understanding), as the authors later admitted. I am talking about monads.

In truth, effects are not married to monads and approachable directly. The profound insight behind monads is the structuring, the separation of `pure' (context-independent) and effectful computations. The structuring can be done without explicating mathematical monads, and especially without resorting to vernacular monads such as State, etc. This article gives an example: a simple, effectful, domain-specific sublanguage embedded into an expressive `macro' metalanguage. Abstraction facilities of the metalanguage such higher-order functions and modules help keep the DSL to the bare minimum, often to the first order, easier to reason about and implement.

The key insight predates monads and goes all the way back to the origins of ML, as a scripting language for the Edinburgh LCF theorem prover. What has not been clear is how simple an effectful DSL may be while remaining useful. How convenient it is, especially compared to the monadic encodings. How viable it is to forsake the generality of first-class functions and monads and what benefits it may bring. We report on an experiment set out to explore these questions.

We pick a rather complex effect -- non-determinism -- and use it in OCaml, which at first blush seems unsuitable since it is call-by-value and has no monadic sugar. And yet, we can write non-deterministic programs just as naturally and elegantly as in Haskell or Curry.

The running tutorial example is computing all permutations of a given list of integers. The reader may want to try doing that in their favorite language or plain OCaml. Albeit a simple exercise, the code is often rather messy and not obviously correct. In the functional-logic language Curry, it is strikingly elegant: mere foldr insert []. It is the re-statement of the specification: a permutation is moving the elements of the source list one-by-one into some position in the initially empty list. The code immediately tells that the number of possible permutations of n elements is n!. From its very conception in the 1959 Rabin and Scott's paper, non-determinism was called for to write clear specifications -- and then to make them executable. That is what will shall do.

A fun pearl by William E. Byrd, Michael Ballantyne, Gregory Rosenblatt, and Matthew Might from ICFP: seven programming challenges solved (easily!) using a relational interpreter. One challenge, for example, is to find quines. Another is to find programs that produce different results with lexical vs. dynamic scope.

The interpreter is implemented in miniKanren (of course), inside Racket (of course).