Don't store the live range of the freed variable for FREE_ON_RETURN
frees, instead look it up at runtime. As this is an extremely
unlikely codepath (in particular, it requires a loop variable with
a throwing destructor), saving the runtime lookup of the live range
is not worth the extra complexity this adds everywhere else.
- Get rid of ZEND_ISEMPTY and ZEND_ISSET_ISEMPTY_MASK. Use just single ZEND_ISSET bit to make distinct between isset() and empty()
- Use ZEND_FETCH_GLOBAL, ZEND_FETCH_LOCAL and ZEND_FETCH_GLOBAL_LOCK as bitmask
- Removed unused ZEND_FETCH_STANDARD
- Extended ZEND_FETCH_ARG_MASK
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.)
For live-variable analysis it does not matter if def includes
variables that are previously use in the same block, the data flow
equations still have the same result. As such there is no need to
compute separate gen & def sets.
I'm keeping the name "def", because use of "gen" in this context is
pretty confusing (gen is usually the use set, not the def set).
* Add FETCH_LIST to inference
* Restrict JMP_SET/COALESCE result type
* Fix typos in range inference
* Add const annotations in zend_ssa.h
* For pi statements dump the associated predecessor block
* If type can be both true and false, dump bool instead
This means we no longer allocate an unused VAR for the retval of
instructions that support unused results.
Nearly all instructions already used the result variable only if
it was used. The only exception to this was the return value
variable for internal function call results. I've adjusted the code
to use a stack zval for the unused return case now. As we have
retval specialization now, we know that it doesn't matter.
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.
This opcodes inserts a local CV into the closure static variable
table. This replaces the previous mechanism of having static
variables marked as LEXICAL, which perform a symtable lookup
during copying.
This means a) functions which contain closures no longer have to
rebuild their symtable (better performance) and b) we can now track
used variables in SSA.
Dmitry Stogov