Optimized HashTable size calculation

eliminate one branch when computing the initial table size

micro optimized initial hash size calculation

most of the calls are with a size <= 8, so don't even bother any
bitshifts. For the bigger sizes, use an intrinsic (should be extended
w

micro optimized initial hash size calculation

most of the calls are with a size <= 8, so don't even bother any
bitshifts. For the bigger sizes, use an intrinsic (should be extended
with gcc ones though).

show more ...

Micro optimization for zend_hash_next_index_insert_new()

bring back all the TSRMLS_FETCH() stuff

for better comparability with the mainstream

cleanup TSRMLS_FETCH

Avoid double IS_INTERNED() check

Split big _zend_handle_numeric_str() into small always inlined function that makes initial check and regular big function

optimized colision list traverse

Remove HASH_NEXT_INSERT flag

We already pass ht->nNextFreeElement and the rest is handled by
ZEND_HASH_ADD.

Remove two more always-true conditions

Optimized FE_RESET and FE_FETCH

Remove always-true comparisons

Update types in zend_hash

* String lengths are size_t
* Indexes are uint32_t
* Flags are uint32_t
* Booleans are zend_bool

Split and optimize loop

master renames phase 3

master renames phase 2

master renames phase 1

further fixes to ext/standard and zend

first shot on merging the core fro the int64 branch

Fixed conversion of empty packed array into hash

Use reference counting instead of zval duplication

Introduced immutable arrays. They don't need to be copyed and may be used directly from SHM.

Use specialized functions instead of macros

Removed useless initializations