JRuby+Truffle - a High-Performance Truffle Backend for JRuby
The Truffle runtime of JRuby is an experimental implementation of an interpreter for JRuby using the Truffle AST interpreting framework and the Graal compiler. It’s an alternative to the IR interpreter and bytecode compiler. The goal is to be significantly faster, simpler and to have more functionality than other implementations of Ruby.
JRuby+Truffle is a project of Oracle Labs and academic collaborators at the Institut für Systemsoftware at Johannes Kepler University Linz.
Authors
-
Chris Seaton
-
Benoit Daloze
-
Kevin Menard
-
Petr Chalupa
-
Thomas Würthinger
-
Matthias Grimmer
-
Josef Haider
-
Fabio Niephaus
-
Matthias Springer
-
Brandon Fish
-
Lucas Allan Amorim
-
Aditya Bhardwaj
The best way to get in touch with us is to join us in #jruby
on Freenode, but you can also Tweet to @chrisgseaton, @nirvdrum or
@eregontp, or email chris.seaton@oracle.com.
Using Truffle
To run JRuby in Truffle mode, pass the -X+T
option.
JRuby+Truffle is designed to be run with a JVM that has the Graal compiler. The easiest way to get this is via the GraalVM, available from the Oracle Technology Network.
github.com/jruby/jruby/wiki/Downloading-GraalVM
But you can also build it yourself, which you will need to do if you are on
the truffle-head
branch.
github.com/jruby/jruby/wiki/Building-Graal
You then need to set the JAVACMD
environment variable as
described on those pages.
$ JAVACMD=... bin/jruby -X+T ...
What to expect
JRuby+Truffle is a research project and is not yet a finished product. Arbitrary programs are very unlikely to run due to missing functionality, and if they do run they are unlikely to run fast yet due to requiring new functionality to be tuned. We are at least a year away from being able to run significant programs without needing new methods to be implemented.
Windows is currently not supported.
How we benchmark
We use the bench9000 benchmarking tool. This includes classic synthetic benchmarks such as mandelbrot, n-body and fannkuch, and also kernels from two real-word Ruby gems, chunky_png and psd.rb.
Research
-
C. Seaton. Specialising Dynamic Techniques for Implementing the Ruby Programming Language. PhD thesis, University of Manchester, 2015.
-
M. Grimmer, C. Seaton, R. Schatz, T. Würthinger, H. Mössenböck. High-Performance Cross-Language Interoperability in a Multi-Language Runtime. In Proceedings of 11th Dynamic Languages Symposium (DLS).
-
F. Niephaus, M. Springer, T. Felgentreff, T. Pape, R. Hirschfeld. Call-target-specific Method Arguments. In Proceedings of the 10th Implementation, Compilation, Optimization of Object-Oriented Languages, Programs and Systems Workshop (ICOOOLPS), 2015.
-
B. Daloze, C. Seaton, D. Bonetta, H. Mössenböck. Techniques and Applications for Guest-Language Safepoints. In Proceedings of the 10th Implementation, Compilation, Optimization of Object-Oriented Languages, Programs and Systems Workshop (ICOOOLPS), 2015.
-
M. Grimmer, C. Seaton, T. Würthinger, H. Mössenböck. Dynamically Composing Languages in a Modular Way: Supporting C Extensions for Dynamic Languages. In Proceedings of the 14th International Conference on Modularity, 2015.
-
A. Wöß, C. Wirth, D. Bonetta, C. Seaton, C. Humer, and H. Mössenböck. An object storage model for the Truffle language implementation framework. In Proceedings of the International Conference on Principles and Practices of Programming on the Java Platform (PPPJ), 2014.
-
C. Seaton, M. L. Van De Vanter, and M. Haupt. Debugging at full speed. In Proceedings of the 8th Workshop on Dynamic Languages and Applications (DYLA), 2014.
Also see the Ruby Bibliography, and publications specifically on Truffle and Graal.
Truffle-Specific Functionality
Options
There are runtime configuration options that can be set on the command line
with -Xtrufle.option=value
. To see a list of these run
-Xtruffle...
.
Truffle Module
The Truffle
, Truffle::Debug
and
Truffle::Interop
modules include Truffle-specific
functionality. They're documented for the current development version
at lafo.ssw.uni-linz.ac.at/graalvm/jruby/doc/.
Debugger
See the documentation of the Truffle::Debug
module at lafo.ssw.uni-linz.ac.at/graalvm/jruby/doc/.
Truffle::Debug.break
will enter a shell and allow to
introspect the program.
If you don't want to modify the program to include a call to
Truffle::Debug.break
you can break the main thread externally.
Run with the instrumentation server enabled,
-Xtruffle.instrumentation_server_port=8080
. Then you can send
a message to the runtime to break at the current location:
curl http://localhost:8080/break
Stack Server
To dump the call stacks of a running Ruby program in Truffle, run with the
instrumentation server enabled and the passalot
option,
-Xtruffle.instrumentation_server_port=8080
-Xtruffle.passalot=1
. Then you can dump the current call stack of
all threads:
curl http://localhost:8080/stacks
Workflow Tool
Truffle is built as part of JRuby, but if you are working on the Truffle
code base you may appreciate the jt
tool. To use it alias it
in your shell function jt { ruby tool/jt.rb $@; }
, then run
jt --help
to see the commands available. You need a standard
ruby
from your system to run the tool.
For example:
-
jt build
builds JRuby and Truffle -
jt run args...
runs JRuby in Truffle mode -
jt run --graal args...
runs JRuby in Truffle mode, using Graal -
jt test fast
runs a subset of Truffle tests