bpo-38328: Speed up the creation time of constant list and set literals.

[brandtbucher]

At @methane's request, here is an alternate PR that takes place in the compiler, not the peephole optimizer. It would replace both #16498 and #16878.

I think I like this one better.

https://bugs.python.org/issue38328

[brandtbucher]

@benjaminp, care to take a look at this when you have a chance?

[methane]

Would you merge master and make regen-all?

[brandtbucher]

Thanks @methane. This branch is no longer stale.

[methane]

Is n > 1 really good threshold? How about n > 2?

$ python3 -m timeit -s "a=frozenset((1,2))" -- "{*a}" 5000000 loops, best of 5: 63.4 nsec per loop $ python3 -m timeit "{1,2}" 5000000 loops, best of 5: 61.5 nsec per loop $ python3 -m timeit "[*(1,2)]" 5000000 loops, best of 5: 46.5 nsec per loop $ python3 -m timeit "[1,2]" 5000000 loops, best of 5: 46.6 nsec per loop 

[brandtbucher]

I only chose n > 1 because it always emits fewer opcodes than before the change. If you think n > 2 is a better heuristic, I'm fine with that.

[brandtbucher]

I think your first example is only slower because it pays the price of a LOAD_NAME instead of a LOAD_CONST (just a guess, though). It's not a direct comparison, unfortunately.

[methane]

The fewer opcodes is not always more efficient.
BUILD_LIST_UNPACK calls PyList_New(0) and PyList_Extend(iterable). It causes over-allocation.
On the other hand, BUILD_LIST just calls PyList_New(n).

So I think we should be conservative about choosing the threshold.

[brandtbucher]

Thanks for the feedback @methane. I think this makes sense.

I've bumped the threshold to your recommended value of n > 2. Anything else?

[brandtbucher]

Thanks for the guidance and merge @methane!