|
|
@ -162,3 +162,11 @@ $(PY_BUILD)/gc.o: CFLAGS += $(CSUPEROPT) |
|
|
|
|
|
|
|
# optimising vm for speed, adds only a small amount to code size but makes a huge difference to speed (20% faster)
|
|
|
|
$(PY_BUILD)/vm.o: CFLAGS += $(CSUPEROPT) |
|
|
|
# Optimizing vm.o for modern deeply pipelined CPUs with branch predictors
|
|
|
|
# may require disabling tail jump optimization. This will make sure that
|
|
|
|
# each opcode has its own dispatching jump which will improve branch
|
|
|
|
# branch predictor efficiency.
|
|
|
|
# http://article.gmane.org/gmane.comp.lang.lua.general/75426
|
|
|
|
# http://hg.python.org/cpython/file/b127046831e2/Python/ceval.c#l828
|
|
|
|
# http://www.emulators.com/docs/nx25_nostradamus.htm
|
|
|
|
#-fno-crossjumping
|
|
|
|