Archive

Racket

One thing that’s great about many modern package managers like npm is that they make it really easy to install software from source repositories, not just from packages. Fortunately, it’s really easy to build a (very simple) version of this on top of Racket’s raco link command. Thus, the raco git tool. Here’s how to use it:

% raco git --github offby1 rudybot

And that’s all you need to do to install rudybot, everyone’s favorite IRC bot. Now there’s a Racket collection set up called rudybot, and you can do things like:

% racket -t rudybot/loop

To install raco git, do the following:

% git clone http://github.com/samth/raco-git.git
% raco link raco-git
% raco setup raco-git

Recently, Matthew Flatt added programmatic logging of garbage collection events in Racket.  Based on this, I’ve built a tool for summarizing the GC behavior of Racket programs.  Here’s an example of how to use it:

 % racket -l gcstats -u my-program.rkt

This says to first require the gcstats module before running my-program.rkt. When you do this you’ll get a big printout after your program runs; below the fold, we’ll take a look at each line in detail.

    39,703,916 bytes allocated in the heap
    28,890,688 bytes collected by GC
    17,083,432 bytes max heap size
    16,604,120 bytes max slop
    28,229,632 bytes peak total memory use

Generation 0:       5 collections,       32ms,    31.71ms elapsed
Generation 1:       0 collections,        0ms,        0ms elapsed

INIT  time       256 ms
MUT   time       132 ms (    129.98 ms elapsed)
GC    time        32 ms (     31.71 ms elapsed)
TOTAL time       420 ms (    417.69 ms elapsed)

%GC time       19.51%   ( 19.61% elapsed)

Alloc rate     300,787,242 bytes per MUT second

To install, follow the instructions on the GitHub page.

Right now, the tool is preliminary, but useful. There are a few limitations:

  1. There are a few GCs before the tool starts — it can’t report anything about them.
  2.  If you have multiple places, it will only report information from the initial place. Fixing this will require  more information from Racket.
  3. The current architecture keeps too much info around during the run of the program.  I hope to fix that soon.

Hopefully, this gives people some better information about how the Racket GC behaves.  The output formatting and information gathered is inspired by similar output from the GHC runtime system. Read More

The Computer Language Shootout is a popular if not-so-informative way to compare the speed of various language implementations.  Racket does pretty well on their benchmarks, thanks to a lot of effort from various people, especially Eli. They run benchmarks on both 1 core and 4 core machines, so languages with support for parallelism can take advantage in many cases.  However, up until this past week, there were no parallel versions of the Racket programs, and therefore Racket didn’t even show up on the 4-core benchmarks. I set out to fix this, in order to advertise Racket’s up-and-coming parallelism constructs.

There are now two new Racket versions of the benchmarks, one each using futures and places. The mandelbrot benchmark uses futures, getting a speedup of approximately 3.2x on 4 cores, and the binary-trees benchmark uses places, with a speedup of almost exactly 2x.

I learned a few things writing these programs:

  1. Racket’s parallelism constructs, though new, are quite performant, at least on microbenchmarks.  With only two parallel programs, Racket is right now competitive with Erlang on 4 cores.
  2. Futures are really easy to use; places take a little more getting used to. Both are quite simple once you get the hang of it, especially if you’ve written concurrent Racket programs before using Racket’s threads.
  3. It can be very surprising which languages are easiest to translate to Racket.  F# and OCaml were the easiest, with Scala similar.  Programs written in Common Lisp, though fast, were much harder to convert to Racket.
  4. My quick rule of thumb for whether to choose places or futures: if you program does much allocation in parallel, or it needs to synchronize, then use places.  Otherwise, futures are probably easier.  I think this is roughly in line with the original design, and there are more applications where synchronization is unnecessary than you would think.

There are a bunch more programs that could have parallel implementations; feel free to hack on them, or to improve mine.