mirrors/cpython
1
0
Fork 0
mirror of https://github.com/python/cpython.git synced 2025-08-27 04:05:34 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 8

gh-129987: Disable GCC SLP autovectorization for the interpreter loop on x86-64 (#132295)

Browse source 
The SLP autovectorizer can cause poor code generation for opcode dispatch, negating any benefit we get from vectorization elsewhere in the interpreter loop.
This commit is contained in:
mpage 2025-04-09 10:34:12 -07:00 • committed by GitHub
parent 67ded6a4fa
commit 1f5682f3a2
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 12 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

13
Python/ceval.c
View file

@ -948,7 +948,18 @@ _PyObjectArray_Free(PyObject **array, PyObject **scratch)
#include "generated_cases.c.h"
#endif
PyObject* _Py_HOT_FUNCTION
#if (defined(__GNUC__) && !defined(__clang__)) && defined(__x86_64__)
/*
* gh-129987: The SLP autovectorizer can cause poor code generation for opcode
* dispatch, negating any benefit we get from vectorization elsewhere in the
* interpreter loop.
*/
#define DONT_SLP_VECTORIZE __attribute__((optimize ("no-tree-slp-vectorize")))
#else
#define DONT_SLP_VECTORIZE
#endif
PyObject* _Py_HOT_FUNCTION DONT_SLP_VECTORIZE
_PyEval_EvalFrameDefault(PyThreadState *tstate, _PyInterpreterFrame *frame, int throwflag)
{
_Py_EnsureTstateNotNULL(tstate);