Re: [PATCH] No runtime fetching of built-in global constants

Hi Dmitry, Well, it's been awhile since Alpha 1 :-), so I wanted to finally resend this before Alpha 2! I agree that the additional optimization probably wouldn't happen often, as there won't be that much namespace usage right away, I assume. But I think it makes sense to handle :: prefix constants, since they're known to have global scope, and I can see the future [online] optimization tip: "When using namespaces, use :: for global constants to get compile-time substitution." ;-) The code is the same as before, just updated the patch. It doesn't really seem less clear, to me, than the current code. In zend_do_fetch_constant(), the "check_namespace" variable may be better named something like "from_namespace" according to my changes, to be more clear, but I left that out to make the patch as simple as possible. http://realplain.com/php/ct_const_fixes.diff http://realplain.com/php/ct_const_fixes_5_3.diff Thanks, Matt ----- Original Message ----- From: "Dmitry Stogov" Sent: Thursday, July 31, 2008 > Hi Matt, > > Now I know. :) > Anyway, I don't think this optimization will work often. > > Send me the patch after Alpha1 release. > > Thanks. Dmitry. > > > Matt Wilmas wrote: > > Hi Dmitry, > > > > Do you know that with your changes, no substitution will happen in a > > namespace even when using :: prefix? :-/ (That's what I would do when I > > know it's global, for optimization.) Or is that what you meant by "not so > > optimal?" > > > > > > - Matt > > > > > > ----- Original Message ----- > > From: "Dmitry Stogov" > > Sent: Thursday, July 31, 2008 > > > >> Thanks Matt. I committed near the same patch. > >> It's not so optimal, but little bit more clear. > >> > >> Thanks. Dmitry. > >> > >> Matt Wilmas wrote: > >>> Hi Dmitry, > >>> > >>> For the behavior change that I mentioned in the other thread, with this > >>> code: > >>> > >>> function foo() { > >>> static $a = -PHP_INT_MAX; > >>> } > >>> > >>> Which could work sometimes, and sometimes not (if in a namespace or > >>> ZEND_COMPILE_NO_CONSTANT_SUBSTITUTION is set). I changed things so that > >>> there is no substitution of constants (except with CT_SUBST flag, like > >>> always) for compile time constants (ZEND_CT mode). They don't create a > >>> FETCH_CONSTANT opcode anyway. :-) > >>> > >>> Another thing I realized wasn't getting optimized (runtime constants), > > which > >>> can be, is with: > >>> > >>> namespace foo; > >>> $a = ::PHP_INT_MAX; // :: for global scope > >>> > >>> So the patch allows substitution there as well. > >>> > >>> http://realplain.com/php/ct_const_fixes.diff > >>> http://realplain.com/php/ct_const_fixes_5_3.diff > >