Bytecode for transfer

Currently transfer is the bottleneck in Apertium, processing here takes 95% CPU. This is because the transfer file is being interpreted (tree walking of the XML in the transfer t1x file) instead of being compiled into machine code.

The Java transfer bytecode compiler converts arbitrarily complex transfer files into Java source code, which is then compiled into platform-indepent bytecode.

During transfer the Java Virtual Machine will convert the most used part (the 'hot spots') into machine code.

This enables

• Faster transfer (currently factor 5) of a corpus
• Debuggable transfer (using a Java development tool, for example Netbeans, you can step thru the transfer code to see exactly what is happening)
• Validating transfer files

A concrete example: Esperanto-English

Take a look at apertium-eo-en.eo-en.t1x and compare with the Java version [http://apertium.svn.sourceforge.net/viewvc/apertium/trunk/lttoolbox-java/src/org/apertium/transfer/generated/apertium_eo_en_eo_en_t1x.java?view=markup apertium_eo_en_eo_en_t1x.java]. This is compiled into bytecode and executed with the Java JIT (Just-in-time) compiler.

Parsing /home/j/esperanto/apertium-svn/apertium/trunk/lttoolbox-java/testdata/transfer/apertium-eo-en.eo-en.t1x
// WARNING: Attribute a_np_acr is not defined. Valid attributes are: [a_nom, a_prp, a_adv, a_adj, a_vrb, a_vrb2, a_det, a_ord, a_prn, a_tns, a_nepersonaj_tempoj, a_gen, a_prs, a_nbr, a_cas, lem, lemq, lemh, whole, tags, chname, chcontent, content]
// Replacing with error_UNKNOWN_ATTR - for <transfer default="chunk">/<section-def-macros>/<def-macro n="firstWord" npar="1">/<choose>/<when>/<test>/<equal>/<clip part="a_np_acr" pos="1" side="sl">
Compiling: javac -cp dist/lttoolbox.jar transfertest/res/lttoolbox-java/testdata/transfer/apertium_eo_en_eo_en_t1x.java

Here is a speed comparison:

Interpreted transfer took 91.59 secs
bytecode compiled transfer took 15.88 secs
Speedup factor: 5.76

Further work

• The Java code have not been optimized for speed, so perhaps the real potential speedup is 6-8, or even a higher factor, if using a mixed mode (mixing C and Java code instead of doing pure-Java).
• Memory usage is also higher than really needed. I.a.
• The underlying library, lttoolbox-java, is using 50% of the CPU, and there are some well known performance issues which are fixable
• The bytecode should be pulled thru an optimizer, like Soot
• There is a zillion of Open Source Java bytecode interpreters to choose from, most prominent Sun's own and http://kaffe.org. Only Sun's have been tested. At least GCJ should be tried out.
• A step for post-compiling to native code should be tried out.
• With http://xmlvm.org/ there could be a way for iPhones as well
• Considering that we have a full port lttoolbox, Apertium could be made to run purely on Java, enabling a wide range of platforms, i.a. Windows, phones (J2ME or Android), web pages, server systems. Only the tagger is missing for a full system.