mirrors/cpython
1
0
Fork 0
mirror of https://github.com/python/cpython.git synced 2025-08-30 21:48:47 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 9

GH-91719: Make MSVC generate somewhat faster switch code (#91718)

Browse source 
Apparently a switch on an 8-bit quantity where all cases are
present generates a more efficient jump (doing only one indexed
memory load instead of two).

So we make opcode and use_tracing uint8_t, and generate a macro
full of extra `case NNN:` lines for all unused opcodes.

See https://github.com/faster-cpython/ideas/issues/321#issuecomment-1103263673
This commit is contained in:
Guido van Rossum 2022-04-21 11:53:57 -07:00 • committed by GitHub
parent d44815cabc
commit f8dc6186d1
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
4 changed files with 90 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

6
Python/ceval.c
View file

@ -1662,7 +1662,9 @@ _PyEval_EvalFrameDefault(PyThreadState *tstate, _PyInterpreterFrame *frame, int
#ifdef Py_STATS
int lastopcode = 0;
#endif
int opcode; /* Current opcode */
// opcode is an 8-bit value to improve the code generated by MSVC
// for the big switch below (in combination with the EXTRA_CASES macro).
uint8_t opcode; /* Current opcode */
int oparg; /* Current opcode argument, if any */
_Py_atomic_int * const eval_breaker = &tstate->interp->ceval.eval_breaker;
@ -5645,7 +5647,7 @@ handle_eval_breaker:
#if USE_COMPUTED_GOTOS
_unknown_opcode:
#else
default:
EXTRA_CASES // From opcode.h, a 'case' for each unused opcode
#endif
fprintf(stderr, "XXX lineno: %d, opcode: %d\n",
_PyInterpreterFrame_GetLine(frame), opcode);