On 2013-03-25 19:02, Zoltán Herczeg wrote: >It is also possible that the JIT does not help you. No solution is
> perfect, everything has advantages and disadvantages.
> I tests JIT against individual patterns. Results are awesome! For some of
them JIT is about 6 times faster than interpreter. It's really excellent
thing!
My real application uses a finite number of patterns. All of them are
precompiled. Application uses them in a jumble order against short, long
and huge subject strings. Adding a JIT results with slowing due to very
long JIT-study time. Average study+execution time with JIT is much greater
then average execution time by interpreter.
Of course I can continue getting along without JIT. IMHO large study time
is real bottleneck and don't allow to use JIT in most complex
applications. But if a precompiling of JIT-data becames possible then it
will be a breakthrough in PCRE's life.
