Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:37216
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: error (pb1.pair.com: domain hristov.com from 85.92.73.163 cause and error)
Message-ID: <4811E99B.6020004@hristov.com>
Date: Fri, 25 Apr 2008 16:24:27 +0200
User-Agent: Thunderbird 2.0.0.12 (X11/20080226)
MIME-Version: 1.0
To: Matt Wilmas <php_lists@realplain.com>
CC: internals@lists.php.net
References: <014b01c8a6db$57d53ba0$0201a8c0@pc1>
In-Reply-To: <014b01c8a6db$57d53ba0$0201a8c0@pc1>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Subject: Re: [PHP-DEV] [PATCH] More array filling optimizations
From: php@hristov.com (Andrey Hristov)

  Hi,
Matt Wilmas wrote:
> Hi all,
> 
> Expanding on the idea of passing a size other than 0 to zend_hash_init(),
> when possible, which was done awhile ago in a few areas (to save
> resize/rehash operations), I finally added an "array_init_size()" that can
> be used instead of array_init(), likewise, when a size is known.  As an
> example to start, I updated these functions:
> 
> array_*: change_key_case, chunk, combine, fill, fill_keys, flip, keys
> (without $search_value), map, rand, reverse, slice, splice, unique (since it
> copies all entries first), values; and also compact; func_get_args;
> get_defined_vars; str_split; and a couple internal uses.

heh, this optimisitations was one of the pluses mysqlnd had over 
mysqli/libmysql. I am using the following macro:

#if PHP_MAJOR_VERSION < 6
#define mysqlnd_array_init(arg, field_count) \
{ \
	ALLOC_HASHTABLE_REL(Z_ARRVAL_P(arg));\
	zend_hash_init(Z_ARRVAL_P(arg), (field_count), NULL, ZVAL_PTR_DTOR, 0); \
	Z_TYPE_P(arg) = IS_ARRAY;\
}
#else
#define mysqlnd_array_init(arg, field_count) \
{ \
	ALLOC_HASHTABLE_REL(Z_ARRVAL_P(arg));\
	zend_u_hash_init(Z_ARRVAL_P(arg), (field_count), NULL, ZVAL_PTR_DTOR, 
0, 0);\
	Z_TYPE_P(arg) = IS_ARRAY;\
}
#endif

> Those were the simplest, most obvious cases I found. :-)  I didn't test
> every function, but for example, the array functions that don't do much work
> I found to be ~10% faster (with more than 8 elements, otherwise no resizing
> would be needed).  If these changes are applied, more can be done where
> possible, by others, or I can and send further patches, or update them
> myself if I have CVS access...
> 
> http://realplain.com/php/array_init_size.diff
> http://realplain.com/php/array_init_size_5_3.diff
> 
> I was playing with initializing scripts' array( ... ) constructs with the
> correct size too, but wasn't sure about the changes or what the opinion
> would be.  Runtime creation is improved with >8 elements, though it adds a
> bit of compile overhead.  Anyway, separate patches for that:
> 
> http://realplain.com/php/array_init_size_vm.diff
> http://realplain.com/php/array_init_size_vm_5_3.diff
> 
> 
> Thanks,
> Matt
> 
>