Streamline list.append for the common case

[sweeneyde]

BPO	47009
Nosy	@tiran, @methane, @markshannon, @sweeneyde
PRs	bpo-47009: Streamline list.append for the common case #31864 bpo-47009: Let PRECALL_NO_KW_LIST_APPEND do its own POP_TOP #32239 bpo-47009: Fix assert on big endian (GH-32332) #32332
Files	_PyList_AppendTakeRef.diff

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2022-04-06.03:19:51.278>
created_at = <Date 2022-03-14.04:57:38.385>
labels = ['interpreter-core', '3.11', 'performance']
title = 'Streamline list.append for the common case'
updated_at = <Date 2022-04-06.03:19:51.277>
user = 'https://github.com/sweeneyde'

bugs.python.org fields:

activity = <Date 2022-04-06.03:19:51.277>
actor = 'Dennis Sweeney'
assignee = 'none'
closed = True
closed_date = <Date 2022-04-06.03:19:51.278>
closer = 'Dennis Sweeney'
components = ['Interpreter Core']
creation = <Date 2022-03-14.04:57:38.385>
creator = 'Dennis Sweeney'
dependencies = []
files = ['50674']
hgrepos = []
issue_num = 47009
keywords = ['patch']
message_count = 8.0
messages = ['415116', '415121', '415127', '415327', '416481', '416766', '416786', '416841']
nosy_count = 4.0
nosy_names = ['christian.heimes', 'methane', 'Mark.Shannon', 'Dennis Sweeney']
pr_nums = ['31864', '32239', '32332']
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'performance'
url = 'https://bugs.python.org/issue47009'
versions = ['Python 3.11']

[sweeneyde]

list_resize is a long function that probably won't get inlined. But for the vast majority of cases in list.append, we just need to check whether the list is big enough (not whether it's small enough, or whether it's null or the wrong type), then insert and update the size. This can be inlined, with an actual call only taking place whenever we need to resize.

We can also add a reference-consuming version of PyList_Append to elide an INCREF/DECREF pair.

[methane]

Hmm. Would you measure benefit from inlining and skipping incref/decref separately?

If benefit of inlining is very small, making _PyList_AppendTakeRef() as regular internal API looks better to me.

[sweeneyde]

The attached _PyList_AppendTakeRef.diff has the ceval.c, but this implementation:

int
_PyList_AppendTakeRef(PyListObject *self, PyObject *newitem)
{
 assert(self != NULL && newitem != NULL);
 assert(PyList_Check(self));
 Py_ssize_t len = PyList_GET_SIZE(self);
 Py_ssize_t allocated = self->allocated;
 assert((size_t)len + 1 < PY_SSIZE_T_MAX);
 if (allocated > len) {
 PyList_SET_ITEM(self, len, newitem);
 Py_SET_SIZE(self, len + 1);
 return 0;
 }
 if (list_resize(self, len + 1) < 0) {
 Py_DECREF(newitem);
 return -1;
 }
 PyList_SET_ITEM(self, len, newitem);
 return 0;
}

Results:

Benchmark	main	PR 31864	_PyList_AppendTakeRef.diff
listcomp 100	1.61 us	1.33 us: 1.21x faster	1.55 us: 1.04x faster
append 100	2.11 us	1.82 us: 1.15x faster	2.05 us: 1.03x faster
listcomp 1000	12.6 us	9.83 us: 1.28x faster	11.9 us: 1.06x faster
append 1000	18.1 us	15.3 us: 1.18x faster	17.6 us: 1.03x faster
listcomp 10000	121 us	93.2 us: 1.29x faster	114 us: 1.06x faster
append 10000	175 us	150 us: 1.17x faster	172 us: 1.02x faster
listcomp 100000	1.17 ms	923 us: 1.26x faster	1.15 ms: 1.02x faster
append 100000	1.70 ms	1.49 ms: 1.14x faster	not significant
Geometric mean	(ref)	1.21x faster	1.03x faster

[methane]

Thank you. I agree that inlining is worth enough.

But we already inlined too many functions in ceval and there is an issue caused by it... (bpo-45116)

[markshannon]

New changeset a0ea7a1 by Dennis Sweeney in branch 'main':
bpo-47009: Streamline list.append for the common case (GH-31864)
a0ea7a1

[markshannon]

New changeset 6c6e040 by Dennis Sweeney in branch 'main':
bpo-47009: Let PRECALL_NO_KW_LIST_APPEND do its own POP_TOP (GH-32239)
6c6e040

[tiran]

New changeset 9e88b57 by Christian Heimes in branch 'main':
bpo-47009: Fix assert on big endian (GH-32332)
9e88b57

[sweeneyde]

Buildbots are passing, so I'm closing this. Thanks for the catch and fix!