PHP should be configured and built with --enable-huge-code-pages, OS should be configured to provide huge pages.
It's possible to enable/disable this future in php.ini through opcache.enable_huge_code_pages=0/1.
The feature was tested on Linux and provided 2% improvement on real-life apps, because of 2-3 times reduction in number of iTLB misses.
Squashed commit of the following:
commit d96eab8d79
Author: Francois Laupretre <francois@tekwire.net>
Date: Fri Jun 26 01:23:31 2015 +0200
Use the new 'ZSTR' macros in the rest of the code.
Does not change anything to the generated code (thanks to compat macros) but cleaner.
commit b352643910
Author: Francois Laupretre <francois@tekwire.net>
Date: Thu Jun 25 13:45:06 2015 +0200
Improve zend_string API
Add missing methods
Removed HashTable->arHash (reduced memory consumption). Now hash slots may be accessed using HT_HASH() macro.
Hash slotas are allocated together with Buckets (before them) and lay in reverse order from HashTable->arData base address (see comments in Zend/zend_types.h)
Indexes in hash table and conflict resolution chains (Z_NEXT) may be stored as indeces or offsets in bytes, depending on system (32 or 64-bit).
HashTable data filelds are reordered to keep the most useful for zend_hash_find() data in the same CPU cache line.
At least in apache, killing a locker with SIGKILL from inside opcache prevents
the child process from emitting a corresponding log event to the access or
error logs, therefore requests are termed with no record. This makes is
incredibly difficult to debug problems espically on systems with high counts
of cache restarts. Since INFO level is far too noisy for production use,
switch the killing to a higher level so we can see it easier.
Now each HashTable is also zend_array, so it's refcounted and may be a subject for Copy on Write
zend_array_dup() was changed to allocate and return HashTable, instead of taking preallocated HashTable as argument.
* origin/master: (36 commits)
NEWS
adapt test for error message introduce in fix for #68463
Fix bug #68463 listen.allowed_clients can silently result in no allowed access
run a few request in this test
fix test
fix test
added info about some other macro changes
added note about toolset and phpize
added note about the response files
split msgfmt_format_intlcalendar_variant*.phpt for ICU 54.1
split formatter_format*.phpt for ICP 54.1
split dateformat_timezone_arg_variations*.phpt for ICU 54.1
split dateformat_get_set_timezone_variant*.phpt for ICU 54.1
fix icu version to test
split ext/intl/tests/dateformat_get_set_calendar_variant*.phpt for 54.1
split dateformat_formatObject_datetime_variant*.phpt for ICU 54.1
split dateformat_formatObject_calendar_variant*.phpt for ICU 54.1
split dateformat_create_cal_arg_variant2.phpt for ICU 53.1 and 54.1
clone dateformat_calendars_variant3.phpt for ICU 54.1
split collator_get_sort_key_variant3.phpt for icu 54.1
...
* origin/master: (398 commits)
NEWS
add test for bug #68381
Fixed bug #68381 Set FPM log level earlier during init
proper dllexport
move to size_t where zend_string is used internally
fix some datatype mismatches
return after the warning, to fix uninitialized salt usage
fix datatype mismatches
add missing type specifier
fix datatype mismatches
fix unsigned check
"extern" shouldn't be used for definitions
joined identical conditional blocks
simplify fpm tests
SEND_VAR_NO_REF optimization
Add test for bug #68442
Add various tests for FPM - covering recent bugs (68420, 68421, 68423, 68428) - for UDS - for ping and status URI - for multi pool and multi mode
Include small MIT FastCGI client library from https://github.com/adoy/PHP-FastCGI-Client
Get rid of zend_free_op structure (use zval* instead). Get rid of useless TSRMLS arguments.
Add new FPM test for IPv4/IPv6
...
Conflicts:
win32/build/config.w32
