GH-118095: Add tier 2 support for YIELD_VALUE #118380
Conversation
|
The stats show that this is working as intended. |
| assert(frame != &entry_frame); | ||
| frame->instr_ptr = next_instr; | ||
| #endif | ||
| frame->instr_ptr++; |
There was a problem hiding this comment.
Would be good if the DSL had a symbol for "this instruction's size".
| @@ -763,7 +763,7 @@ translate_bytecode_to_trace( | |||
| Py_FatalError("garbled expansion"); | |||
| } | |||
|
|
|||
| if (uop == _POP_FRAME || uop == _RETURN_GENERATOR) { | |||
| if (uop == _POP_FRAME || uop == _RETURN_GENERATOR || uop == _YIELD_VALUE) { | |||
There was a problem hiding this comment.
@markshannon Shouldn't this same change be made to optimizer_analysis.c (in get_code())?
For your information, from gcc-14, since it enabled |
This PR:
YIELD_VALUEsuitable for tier 2_YIELD_VALUEmicro-op._FOR_ITER_GEN_FRAMEto the tier 2 optimizer to get rid of a compiler warning. This isn't strictly necessary for this PR, but is just good housekeeping.