Hi,
I read in your other mail that you are retired, but your words suggest that noone other working on PCRE, which is a surpise for me, as PCRE is an excellent piece of software. I hope you can find someone who would continue PCRE development.
Answering your qestions, yes, it would require a complete redesign of the matching part (the parser probably don't need to be touched), but it would probably worth it, since current JS engines can do pattern matching 5 or more times faster in JIT mode than in interpreted mode. For most patterns, only a few machine instructions are generated, and it can be much faster than executing those big switch()-es employed by any interpreters. SLJIT was written in C as well, but it can be compiled by a C++ compiler, which provides a great flexibility.
Regards,
Zoltan
Philip Hazel <ph10@???> írta:
>On Sun, 29 Aug 2010, Herczeg Zoltán wrote:>
>
> Have you ever considered adding JIT support for PCRE?>
>
No. I suspect it would require a complete redesign and rewrite of the>
library (which is written in C, as I guess you already know). That won't>
ever be done by me as I am now retired. I will continue to maintain PCRE >
for a while, but I do not see myself doing any major development of any >
kind, let along a radical re-write such as that.>
>
Would JIT support *really* give better performance? PCRE doesn't take >
very long to compile a regex, and you have the advantage of discovering >
any syntax errors before you start to try to match anything.>
>
Philip>
>
-- >
Philip Hazel
