Optimize SplFixedArray when magic methods aren't overridden

[TysonAndre]

This decreases the memory usage of SplFixedArrays by 32 bytes per object
on 64-bit systems (use 1 null pointer instead of 5 null pointers)
If allocating a lot of arrays of size 1, memory usage was 19.44MiB before this
change, and 16.24MiB after the change.

Existing tests continue to pass.

Subclassing SplFixedArray is already inefficient and rarely done - it checks for the existence of 5 methods every time a subclass is instantiated.
(and has to switch back from C to the php vm to call those methods)

<?php
/*
Before(last run)
Change to memory in SplFixedArray: 79.124MiB elapsed: 0.114
Change to memory in array:        204.781MiB elapsed: 0.076

After optimizations (last run)
Change to memory in SplFixedArray: 63.124MiB elapsed: 0.098
Change to memory in array:        204.781MiB elapsed: 0.075

 */
ini_set('memory_limit', '2G');
function bench() {
    call_user_func(function () {
        gc_collect_cycles();
        $start = microtime(true);
        $count = 500000;
        $old_memory = memory_get_usage();
        $values = new SplFixedArray($count);
        for ($i = 0; $i < $count; $i++) {
            $values[$i] = new SplFixedArray(1);
            $values[$i]->offsetSet(0, $i);
        }
        $new_memory = memory_get_usage();
        unset($values);
        $end = microtime(true);
        printf("Change to memory in SplFixedArray: %.3fMiB elapsed: %.3f\n", ($new_memory - $old_memory)/1000000, $end-$start);
    });
    call_user_func(function () {
        gc_collect_cycles();
        $start = microtime(true);
        $count = 500000;
        $old_memory = memory_get_usage();
        $values = [];
        for ($i = 0; $i < $count; $i++) {
            $values[$i] = [$i];
        }
        $new_memory = memory_get_usage();
        unset($values);
        $end = microtime(true);
        printf("Change to memory in array:        %.3fMiB elapsed: %.3f\n", ($new_memory - $old_memory)/1000000, $end-$start);
    });
}
for ($i = 0; $i < 10; $i++) {
    bench();
}

[TysonAndre]

@morrisonlevi @nikic @kocsismate - any objections?

[morrisonlevi]

You reduce the size of each SplFixedArray by 32 bytes, yet in the script comments it says after applying this optimization you went from 79.124MiB down to 63.124MiB, yet it only instantiates 10 SplFixedArrays. What am I missing?

[TysonAndre]

There's $count = 500000 arrays being created inside the test. 0.5MiB * 32 is 16MiB, which is what's being reported by the calculation

I ran the benchmark as a whole 10 times in the loop at the bottom

[morrisonlevi]

Ah, I see. I'm not sure how common it is to instantiate more than one at a time.

Pro: obviously it is more space efficient when SplFixedArray isn't inherited from (common).
Con: if you do inherit, it does more allocations. We can't pack it into the same allocation dynamically because SplFixedArray can have dynamic properties and can be inherited from, so unless we can pack it into an official zval (weird) then I don't see an easy way around it, even though we only really need at most one allocation per inherited class.

I will support whatever Nikita or other SPL contributors think is best, as I don't feel strongly about it.

[nikic]

Ideally these methods would be cached in the class_entry, but that's a larger change.

[nikic]

Suggested change

	spl_fixedarray_methods* methods;
	spl_fixedarray_methods *methods;

[TysonAndre]

Done in linked commit

[TysonAndre]

Ideally these methods would be cached in the class_entry, but that's a larger change.

True - I wasn't sure if there was already a way to automatically run a C callback to modify any subclass of a class such as SplFixedArray but it didn't seem straightforward.

Using two sets of handlers and different sized allocations to avoid checking ->methods in the base class seemed possible, but harder to maintain

[TysonAndre]

I'd misunderstood what you meant earlier.

	/* allocated only if class implements Iterator or IteratorAggregate interface */
	zend_class_iterator_funcs *iterator_funcs_ptr;

If a new arraylike_funcs_ptr were added to php for ArrayAccess and handled when a class implemented ArrayAccess, instance->ce->arrayaccess_funcs_ptr->offsetget->scope != my_base_class_entry could be used to check if the internal implementation was overridden. Is that what you mean?

zend_ce_iterator->interface_gets_implemented = zend_implement_iterator; exists for the Iterator interface, and this would be something similar for ArrayAccess? (instance->ce->arrayaccess_funcs_ptr->offsetget->scope != my_base_class_entry could be used to check if the internal implementation was overridden)