#
b3e12325 |
| 31-Jul-2020 |
Dmitry Stogov |
Fixed JIT when result of ASSIGN is in register, but the original variable might need to be destroyed
|
#
8ea9e1ae |
| 30-Jul-2020 |
Dmitry Stogov |
We check result type guard in FETCH_DIM_R/IS only if index is long or string.
|
#
8b439764 |
| 30-Jul-2020 |
Dmitry Stogov |
Fixed memory leaks caused by incorrect AVOID_REFCOUNTING flag
|
#
3690a805 |
| 30-Jul-2020 |
Nikita Popov |
Check non-zero in is_power_of_two() And assert non-zero in floor_log2(). Fixes DASM_S_RANGE_I in ext/simplexml/tests/021.phpt.
|
#
bbbccf08 |
| 29-Jul-2020 |
Dmitry Stogov |
Force type update if the previous value was in register only
|
#
44384dfb |
| 29-Jul-2020 |
Dmitry Stogov |
Fixed temporary register usage
|
#
9e805580 |
| 29-Jul-2020 |
Dmitry Stogov |
Load ptoper IP value
|
#
02fae1fc |
| 29-Jul-2020 |
Dmitry Stogov |
Fixed bug #79888 (Incorrect execution with JIT enabled)
|
#
2a72778e |
| 28-Jul-2020 |
Dmitry Stogov |
Check for EG(exception) after leaving function frame
|
#
251e6674 |
| 28-Jul-2020 |
Dmitry Stogov |
Prevent incorrect optimization
|
#
2dbb4ecf |
| 28-Jul-2020 |
Dmitry Stogov |
Avoid redundand EX(opline) initialization
|
#
04c3e318 |
| 27-Jul-2020 |
Dmitry Stogov |
More accurate checks to eliminate reloading
|
#
f9f61239 |
| 27-Jul-2020 |
Nikita Popov |
Fix verify type jit for complex type masks We should not try to check things like MAY_BE_CALLABLE in the fast path, because those aren't real types. This is only actively wrong for t
Fix verify type jit for complex type masks We should not try to check things like MAY_BE_CALLABLE in the fast path, because those aren't real types. This is only actively wrong for the concrete_type case, for the mask case it just uses an unnecessarily large immediate.
show more ...
|
#
6d97e7f0 |
| 27-Jul-2020 |
Nikita Popov |
JIT internal functions with return type in debug mode The return type check is just a debug assertion, it's okay if it is not performed in JIT mode. We already don't perform all the
JIT internal functions with return type in debug mode The return type check is just a debug assertion, it's okay if it is not performed in JIT mode. We already don't perform all the argument validation in that case. Just disable the one test that checks for this. This removes an annoying discrepancy between debug&release mode.
show more ...
|
#
c56c19ce |
| 27-Jul-2020 |
Dmitry Stogov |
Fixed incorrect reference counting (we shouldn't skip ADDREF for IS_CV result)
|
#
0ef8e0c1 |
| 24-Jul-2020 |
Dmitry Stogov |
Result of branching opcode in side trace is unknown
|
#
b3e51b767 |
| 23-Jul-2020 |
Dmitry Stogov |
Allow keeping result of FETCH_DIM_R in CPU register
|
#
7de23662 |
| 23-Jul-2020 |
Dmitry Stogov |
Fixed 32-bit build
|
#
f6164b84 |
| 23-Jul-2020 |
Dmitry Stogov |
Revert "Fixed DynASM failure" This reverts commit 62ca6b567cdae0f8f7e02e6e2ebe6bb64ee1dd3c.
|
#
62ca6b56 |
| 23-Jul-2020 |
Dmitry Stogov |
Fixed DynASM failure
|
#
b0c3ca0e |
| 23-Jul-2020 |
Dmitry Stogov |
Avoid register reloading
|
#
545928e4 |
| 23-Jul-2020 |
Nikita Popov |
Fix some clang warnings
|
#
b6b6e769 |
| 23-Jul-2020 |
Nikita Popov |
Try to fix 32-bit NTS build
|
#
bc1d7b60 |
| 22-Jul-2020 |
Dmitry Stogov |
Reserve WIN64 shadow space
|
#
aaedbde8 |
| 22-Jul-2020 |
Nikita Popov |
Mark zend_jit_patch memory accesses as unaligned This prevents ubsan from complaining.
|