Sort blocks in loop identification

The previous implementation was doing O(blocks*levels) iterations,
which for a linear-ish domtree is O(blocks^2). Avoid this by sorting
the blocks

Sort blocks in loop identification

The previous implementation was doing O(blocks*levels) iterations,
which for a linear-ish domtree is O(blocks^2). Avoid this by sorting
the blocks by decreasing level upfront.

show more ...

Update copyright headers to 2017

Update copyright headers to 2017

Fixed possible false assertion failure

Revert "Fix memory leaks with FE_RESET upon empty iterator"

This reverts commit 00e516e57549f29387ef858de7847e7e4ca0c1c0.

Fix memory leaks with FE_RESET upon empty iterator

Solved by manually freeing and jumping to the opcode *after* FE_FREE

Allow optional splitting of basic blocks at RECV/RECV_INIT opcodes.

Fixed CFG construction for stackless execution (split basic blocks after calls)

Fixed compilation warnings

Don't optimize special dynamic calls to non-dynamic

As it drops the warning. This is more problematic with constant
propagation, as tests would fail.

Extract a zend_optimizer_cl

Don't optimize special dynamic calls to non-dynamic

As it drops the warning. This is more problematic with constant
propagation, as tests would fail.

Extract a zend_optimizer_classify_function() function, as its now
needed by zend_cfg and update_opN.

show more ...

Revert "assert won't cause indrect var access if it's disabled"

in the other hand, it should not be there if it's disabled...

This reverts commit 27f40676e6f506f353ad8d679dd9b21aa87

Revert "assert won't cause indrect var access if it's disabled"

in the other hand, it should not be there if it's disabled...

This reverts commit 27f40676e6f506f353ad8d679dd9b21aa87b828e.

show more ...

assert won't cause indrect var access if it's disabled

Ensure no entry predecessors for SSA construction

Replace BB end with BB len

And support empty blocks everywhere.

Improve previous fix

Do not mark loop var free blocks as reachable after all -- as we
can't construct SSA for unreachable blocks, this would cause
issues down the line.

Inst

Improve previous fix

Do not mark loop var free blocks as reachable after all -- as we
can't construct SSA for unreachable blocks, this would cause
issues down the line.

Instead add an extra UNREACHABLE_FREE flag and retain only the
FREE instruction during NOP removal. (If we retain all
instructions in the BB we might leave a jump instruction that goes
into the nowhere.)

show more ...

Fix reachability detection if live ranges aren't split

Even if we don't split, we still want to keep the same logic for
reachibility detection.

Mark assert() as INDIRECT_VAR_ACCESS

I don't like this... we may want to detect "obviously not string"
operands by checking for the result of a comparison instruction.

Eliminated checks for (func->op_array.fn_flags & ZEND_ACC_GENERATOR) in fast path of DO_FCALL* handlers.
This slightly improves calls to regular function and method calls in cost of a bit slower

Eliminated checks for (func->op_array.fn_flags & ZEND_ACC_GENERATOR) in fast path of DO_FCALL* handlers.
This slightly improves calls to regular function and method calls in cost of a bit slower generator initialization.
Separate call frame for generators, allocated on heap, now created by ZEND_GENERATOR_CREATE instruction.

show more ...

Evaluate arguments of new for classes without ctor

ML: http://markmail.org/message/4b3mk7jid64zvz34

Fix build

Fix dominator tree construction

The used dominator intersection algorithm assumes a postorder
numbering of the CFG. The reversal of our natural numbering is
quite similar to postorde

Fix dominator tree construction

The used dominator intersection algorithm assumes a postorder
numbering of the CFG. The reversal of our natural numbering is
quite similar to postorder, but not the same.

In the future we should precompute both preorder/postorder
numberings and orderings, as these are useful in many places.

show more ...

Support CFG construction without live range splitting

We must not split at live range boundaries for SSA constructions,
otherwise an OP_DATA instruction may be separated into new block

Support CFG construction without live range splitting

We must not split at live range boundaries for SSA constructions,
otherwise an OP_DATA instruction may be separated into new block
and not picked up during renaming.

It's also unnecessary for this use case and only blows up the CFG.

show more ...

Don't insert duplicate predecessors

Otherwise we'll get corrupt phis

More accurate handling of isset() and unset()

Mark isset($$var) as INDIRECT_VAR_ACCESS