bpo-45885: Add more stats for COMPARE_OP in specialize.c

[sweeneyde]

https://bugs.python.org/issue45885

[sweeneyde]

Stats on some of pyperformance (before I considered float_int and int_float): https://gist.github.com/sweeneyde/157d2a24ddb7f21f2038ce1997d1b175

[sweeneyde]

From all of pyperformance:

COMPARE_OP_INT_JUMP:   454436174 1.9% 58.9%
COMPARE_OP:             59845807 0.2% 95.2%
COMPARE_OP_ADAPTIVE:    48912937 0.2% 96.1%
COMPARE_OP_STR_JUMP:    20343452 0.1% 98.1%
COMPARE_OP_FLOAT_JUMP:  18384022 0.1% 98.4%

COMPARE_OP:
 unquickened:    59845807 10.0%
    deferred:    47982034 8.0%
       deopt:        5709 0.0%
         hit:   492833412 82.0%
        miss:      330236 0.1%
  success:      110039
  failure:      820864
    kind  0:    50088 6.1%
    kind 12:    56388 6.9%
    kind 13:       45 0.0%
    kind 14:    60430 7.4%
    kind 15:    35732 4.4%
    kind 16:    18354 2.2%
    kind 17:    40646 5.0%
    kind 18:     8579 1.0%
    kind 19:   227608 27.7%  (Note: mostly frozenset <= frozenset in bm_meteor_contest)
    kind 20:    45127 5.5%
    kind 21:    13893 1.7%
    kind 22:   262472 32.0%  (Note: mostly scimark_lu and scimark_fft doing float compared to int)
    kind 23:     1502 0.2%

[kumaraditya303]

Suggested change

	if (PySet_CheckExact(lhs) || PyFrozenSet_CheckExact(lhs)) {
	if (PyAnySet_CheckExact(lhs)) {

[sweeneyde]

I'd personally rather not -- IMO the || is more clear and self-explanatory.

[markshannon]

Thanks for this. The extra data is going to be useful.

[markshannon]

Could you rewrite this as SPECIALIZATION_FAIL(COMPARE_OP, compare_op_fail_kind(lhs, rhs)); following the pattern used in the other specialization functions?

[markshannon]

As above, could you move the classification of the failure into a helper function?
SPECIALIZATION_FAIL(COMPARE_OP, compare_op_fail_kind(lhs, rhs));
Otherwise, it obscures the logic of the specialization function.

[bedevere-bot]

A Python core developer has requested some changes be made to your pull request before we can consider merging it. If you could please address their requests along with any other requests in other reviews from core developers that would be appreciated.

Once you have made the requested changes, please leave a comment on this pull request containing the phrase I have made the requested changes; please review again. I will then notify any core developers who have left a review that you're ready for them to take another look at this pull request.

[markshannon]

Looks good, thanks.