GNU Guile Scheme Gets a Register VM and CPS-Based IL 42
In late November, Andy Wingo pushed a new register VM to Guile's (the GNU implementation of the Scheme language) master branch. It brought a number of performance improvements, but led to a bit of a conceptual mismatch between the compiler's direct-style intermediate language and the virtual machine. Earlier this week Andy Wingo announced a new continuation-passing style intermediate language for Guile. From the article:
"To recap, we switched from a stack machine to a register machine because, among other reasons, register machines can consume and produce named intermediate results in fewer instructions than stack machines, and that makes things faster. To take full advantage of this new capability, it is appropriate to switch at the same time from the direct-style intermediate language (IL) that we had to an IL that names all intermediate values. ... In Guile I chose a continuation-passing style language. ... Guile's CPS language is composed of terms, expressions, and continuations. It was heavily inspired by Andrew Kennedy's 'Compiling with Continuations, Continued' paper. ... The optimizations I have currently implemented for CPS are fairly basic. Contification was tricky. One thing I did recently was to make all non-tail $call nodes require $kreceive continuations; if, as in the common case, extra values were unused, that was reflected in an unused rest argument. This required a number of optimizations to clean up and remove the extra rest arguments for other kinds of source expressions: dead-code elimination, the typical beta/eta reduction, and some code generation changes."
The article describes the CPS language provided by Guile and explains the reasons behind choosing CPS over SSA or A-Normal Form. The Guile manual contains draft documentation. The new VM and Intermediate Language will be released with Guile 2.2, which should be out later this year.
Re: (Score:2, Troll)
Sonic BOOM!
Re:frost pist with a wager (Score:5, Insightful)
Remember when a nerd was someone who cared about tail call optimizations and SSA, and not which corporation made their cellphone?
Off my lawn.
Re: (Score:2)
Re: (Score:1)
So? I fit that, and still consider this "stuff that matters". I was wondering why they thought their register VM was applicable to a scheme variant, and I had started to cross Guile off my list of relevant Scheme implementations. I'll be waiting until I see what actually gets documented before I weigh in further though.
Then again, I've been interested in Apple products since the 70's, so get off my lawn :D
Re: (Score:2)
I was wondering why they thought their register VM was applicable to a scheme variant
Why would it not be relevant?
Re: (Score:2)
Because they already had a stack machine variant. And stack machines tend to have a lot of activity on the top of the stack, and in some lisp implementations that can result in a lot of garbage collection. Even without garbage being created it's a lot of extra churning, with more stack manipulation instructions being generated by the compiler. Note that a lot of Forth machines and implementations implement a top-of-stack register for efficiency.
Re: (Score:2)
one of the most widely deployed vms on the planet is register based - dalvik. consequently, such an architecture may be back in vogue.
Re: (Score:3, Interesting)
Register VMs began to come back into vogue at least since the Dis VM in Plan 9. The Bell Labs developers wrote a paper about Dis which effectively stated that, "yo, yo, register VMs are faster out-of-the-box, easier to optimize, and easier to translate to native code".
By the time Dalvik came along (10 years after the Dis paper) it was already conventional wisdom that register-based VMs were the obvious choice for performance. But most projects still use stack-based VMs because they're easier to implement an
This stuff is starting to matter. (Score:3, Interesting)
Re: (Score:2)
This is awesome. Where else other than slashdot can we discuss continuation passing as a compilation method with peers who understand what we're talking about?
Re: (Score:2, Informative)
Usenet?
lambda-the-ultimate.org?
Slashdot has been overrun by tech consumers.
Re: (Score:1)
Re: (Score:2)
I don't understand CPS or SSA and register-based vs. stack-based well enough to know the answer.
Re: (Score:2)
SSA can use registers. In fact it usually does use registers.
Re: (Score:2)
Scheme-Great way to learn function programming (Score:2)
Re: (Score:2)
Anything that has parentheses-matching as a requirement is simply self-flagellation.
That's what the backronym says: Lots of Idiotic Single Parentheses.
Hierarchic thinking has infested Computer Science for decades. The entire Object Oriented Programming paradigm was founded on hierarchy, with Inheritance defined as from one parent only, and many not too sure if multiple inheritance is a good idea or needed. That the notion of inheritance even has to be qualified with that word "multiple", because it implicitly means single otherwise, is a barrier to thought. Trees are useful data stru
Re: (Score:2)
True, you can do DAGs in trees, which encompasses most of the hard stuff in theory. But with Lisp you can have the cycles too, so it really does implement a universal data structure.
Re:Scheme-Great way to learn function programming (Score:5, Insightful)
Hierarchic thinking has infested Computer Science for decades. The entire Object Oriented Programming paradigm was founded on hierarchy, with Inheritance defined as from one parent only, and many not too sure if multiple inheritance is a good idea or needed. That the notion of inheritance even has to be qualified with that word "multiple", because it implicitly means single otherwise, is a barrier to thought. Trees are useful data structures, but they aren't the ultimate, universal data structure that can succinctly describe all other organizations of data.
There's a lot of foolish programmers who think that because they've got inheritance available to them, they've got to use it. It's the If All You Have Is A Hammer principle. Delegation/composition is much more useful in practice (and can express arbitrary graphs just fine).
Parentheses matching not required (Score:4, Informative)
GNU guile's built-in reader includes support for SRFI-105, so you can use infix expressions directly. In particular, you can use {...} instead of (...) and put the operator in the EVEN position, e.g., {n https://www.gnu.org/software/guile/manual/html_node/SRFI_002d105.html
If you want to eliminate more of the parens, you can use guile with SRFI-110, which provides support for indentation-sensitive semantics. An implementation is available with an MIT license. See more here: http://readable.sourceforge.net/ [sourceforge.net]
Re: (Score:2, Informative)
www.lilypond.org
Thanks
Slashdot in Greek (Score:2, Funny)
I don't think I have ever encountered a title at Slashdot more meaningless to me than this one. There might be a lesson in that, and I don't think the lesson is for me.
Re:Slashdot in Greek (Score:4, Insightful)
To those interested in the implementation of programming languages, it is immediately apparent that this is a fundamental change in the compiler behind the GNU Guile system which implements the Scheme programming language, inasmuch as it now has a virtual machine based on the register model instead of the implied stack model, along with an intermediate language in its compilation path that is based on continuation-passing style.
I think that the lesson here is for everyone: There are many segments of nerd culture, and it is very unlikely that any randomly-selected Slashdot reader understands and appreciates all of those segments. For example, the earlier headline today "Why Transivity Violations Can Be Rational" has no meaning to many readers, even after the title was corrected to spell transitivity correctly. After reading a little about that topic, I see that it is an area of of obvious interest to many nerds.
That being said, there are plenty of topics Slashdot poorly reports on which are not of interest to any segment of nerd culture, at least not beyond the overlap between nerd culture and the mainstream news where we already read the same information three days earlier except through the words of a literate, competent reporter with real editing before it hit the press.
Re: (Score:2)
It's about defining an intermediate syntax for the compiler to use, so that optimisations can be performed efficiently. As TFA suggests, if you need an overview on the differences & similarities of SSA and CPS, you could read his earlier article [wingolog.org].
I've experimented a little with LLVM's SSA form for compiling and optimising high level procedural languages. And I think he makes a few quite valid criticisms of the way SSA form is defined and used.
Great language that no one uses.. (Score:3)
The Guile team is doing amazing work. Starting with version 2, Scheme is no longer something that always runs slow. My simple test code run several times faster in Guile 2 compared to similar code in python 3. The problem is that no one seems to be interested. It's quite sad. Guile is simply a fantastic, fairly complete programming environment with a very fast VM. In the early days of Gnome, GNU elevated Scheme to the number 1 status among the Gnome languages, but it was dropped since then. I don't know whether this is from the perception of LISP as the old-school language or the rejection of the code aesthetics due to meticulous use of parenthesis.