Re: Testing JIT compilers

Tim Harris <tlh20@cam.ac.uk> writes:
> I was thinking about testing compilers a bit last week while working
> on a Java-bytecode to x86-native-code compiler. I'd be interested in
> feedback on the prototype system described below -- e.g. flaws in it,
> or that it has been done before :-).
>


The CACAO JIT compiler has a trace option where method invocations, their
arguments and the return value is traced. The heap and the string table
is mmaped to a fixed address. So the arguments of methods have always
the same values, even if otpimizations are added. If the new version of
the JIT compiler is wrong, I do a diff between the traces to find the
faulting method. If you know your changes, it is usually easy to find
the problem. This scheme of testing worked even between CACAO on the
Alpha and CACAO on the MIPS. Some programs behave differently, e.g.
programs which use date, time or the environment.


--
andi@complang.tuwien.ac.at Andreas Krall
http://www.complang.tuwien.ac.at/andi/ Inst. f. Computersprachen, TU Wien
tel: (+431) 58801/18511 Argentinierstr. 8/4/1851
fax: (+431) 58801/18598 A-1040 Wien AUSTRIA EUROPE

Return to the comp.compilers page.
Search the comp.compilers archives again.