That leads to some depressing conclusions.

Indeed, but Rasmus has spelled them out(12 Dec. 2005)

Bart de Boer went ahead anyway and asked whether there are any future plans for native multithreading support in PHP?
Rasmus denied it, saying that asynchronous mechanisms offer a more efficient approach. Bart clarified this – by ‘asynchronous mechanisms’, did Rasmus mean calling other scripts, such as web services, from within the main script? He’d regard that as a good alternative, apart from the extra overhead, scripts and code…

It’s an answer just as cryptic as you’d expect from the Oracle at Delphi (not the IT-stuff, but Greek mythology). Why would and how could webservices make up for PHP extensions? GD-library via webservices? Did some more research and at some point my head started spinning with acronyms and comparing concepts I knew not too much about at all. Then decided to think “What an interesting discussion” and have a cup of tea.

I’d bet other design decisions have a far greater impact on raw speed than the running with thread safety.

tried to find a list of thread-safe (or ZTS: “Zend Thread Safe”) extensions. Couldn’t find anything,

Actually never seen anything like that either – I’ve seen PCRE and GD mentioned before (somewhere – perhaps PHP internals list – can’t remember) as being non-reentrant but perhaps I shouldn’t believe everything I read online.

You got me wondering more though and perhaps this highlights the problem with even compiling a list in the first place;

> Is there a short list of some common extensions that do work reliably
> with Apache 2? (eg MySQL, GD etc…)

Nope. Make one. It’s not all that easy to create such a list. Just take
a single library and go through it and tell me if it is threadsafe. Can
you do that? Do you know what to look for? If it is threadsafe on one
platform, do you think it is threadsafe on another? Does it depend on
which version of libc is on a platform? Does it depend on how it was
compiled?

There’s also this remark (Andi Gutmans);

You will most likely find that the “faster” Apache way with
thread-safe PHP is slower than the slower Apache way with
non-thread-safe PHP. And even FastCGI will be faster :)

That seems to tally with what Sara Golemon says What the heck is TSRMLS_CC anyway? (comments);

> Is the overhead of lookups in tsrm_ls significant,
> when in ZTS “mode”? Put another way, why not have everything thread safe
> and discourage the use of globals completely?
>
At the CPU instruction level, it’s the difference between:
Fetch a local value

and

Fetch a local value,
dereference it,
Fetch a local value,
Seek forward from that first dereference by that second local value,
Dereference that,
seek into that pointer by X bytes and dereference that value

Any guesses which is faster? Now multiply that difference by EVERY INSTANCE of xG() in the engine.

That leads to some depressing conclusions.

one word

lighttpd

one word: mod_rewrite ;)

The Worker MPM (which is what you meant, of course) is thread-based, and improves performance for most applications.

Thanks for picked nit ;) That’s what I meant.

lighttpd

But that’s FastCGI