[Webkit-unassigned] [Bug 42264] Prepare YARR JIT for matching regexps with iterative parentheses

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=42264





--- Comment #6 from Peter Varga <pvarga at inf.u-szeged.hu>  2010-07-16 00:31:07 PST ---
(In reply to comment #5)


Hi Gavin,

With my previous patch I just want to show this logic doesn't cause performance regression. 
IMHO the two kinds of allocation in different cases still cause less complexity and maintance overhead
to the engine than do fallback to PCRE. Of course this patch doesn't solve the stack allocation problem
which you mentioned at point (1). To solve that problem I need further investigation.

I think it is important to write an engine which avoids to use of PCRE by reason of the different results of
matching as well.
At the moment I see two ways to solve the problem:
(a) To extend the current YARR JIT implementation with capable of matching the fallback cases. This solution
changes the current logic. To refer at your point (1) I'm thinking on a solution which keeps the original stack
preallocating logic and use heap only for store each iteration of parentheses.
(b) Write a NEW regex JIT engine which doesn't do fallback and is faster than PCRE. If it will slower than
the YARR JIT in the currently handled cases then the YARR JIT can do fallback to this (instead of PCRE).

What would you prefer to solve this problem? Maybe you can suggest another way? Perhaps my train of thought is wrong?

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.