On some systems, like Alpine, the thread stack size is small by default.
The last step of SSA construction involves variable renaming that is
recursive, and also makes copies of their version of the renamed
variables on the stack. This combination causes a stack overflow during
compilation on Alpine. Triggerable for example with very long match
statements.
A stop-gap solution would be to use heap allocated arrays for the
renamed variable list, but that would only delay the error as increasing
the number of match arms increases the depth of the dominator tree, and
will eventually run into the same issue.
This patch transforms the algorithm into an iterative one.
There are two states stored in a worklist stack: positive numbers
indicate that the block still needs to undergo variable renaming.
Negative numbers indicate that the block and its dominated children are
already renamed. Because 0 is also a valid block number, we bias the
block numbers by adding 1.
To restore to the right variant when backtracking the "recursive" step,
we index into an array pointing to the different variable renaming
variants.
Closes GH-19083.
Previously, this variable was necessary because of auto-vivification on
UNDEF/null/false. It's now only used for RC inference, as auto-vivification has
been removed.
This implicitly solves an inference problem for $obj->bar &= $obj; where we get
a new variable for both literal references to $obj, with the first one getting
the RCn flag, and the second one getting the MAY_BE_REFERENCE flag. Thus, the
first variable will be missing the reference type, causing a false-positive type
inference warning.
If we want to verify RC inference at some point we'll need a better solution.
Closes GH-13233
While JMPZNZ can avoid execution of a separate JMP opcode in some
cases, it also prevents smart branch optimization, so creating
JMPZNZ may actually have a negative effect. It also adds additional
complexity for optimizations.
Drop JMPZNZ in favor of JMPZ+JMP or JMPNZ+JMP.
Closes GH-7857.
This handles references to the current class through its name
rather than self (and for cases where is is not linked yet and
thus not covered by the context lookup). Rather than handling this
only for FETCH_CLASS_CONSTANT optimization, integrate this into
the generic get_class_entry() utility.
This code is repeated a few time. Two occurrences additionally
contained checks for user classes in CG(class_table) with the
same file name, but as far as I know these should always be in
the script class_table, so I'm omitting the check here.
1. Update: http://www.php.net/license/3_01.txt to https, as there is anyway server header "Location:" to https.
2. Update few license 3.0 to 3.01 as 3.0 states "php 5.1.1, 4.1.1, and earlier".
3. In some license comments is "at through the world-wide-web" while most is without "at", so deleted.
4. fixed indentation in some files before |
This only moves the files, adjusts the build system, exports APIs
and does minor fixups to make sure the code builds.
This does not yet try to make the optimizer usable independently
of opcache.
Closes GH-6642.
2021-01-28 10:38:25 +01:00
Renamed from ext/opcache/Optimizer/zend_ssa.c (Browse further)
