Improve ByteBuffer.setSubstring(_:at:) performance

[gwynne]

Substring.UTF8View began implementing Collection.withContiguousStorageIfAvailable(_:) starting in Swift 5.3.

Motivation:

The conversion from Substring to String is usually inexpensive, but not always, and can be avoided in most cases when using Swift 5.3 or later; the gain is minor, but worthwhile.

Modifications:

Updates ByteBuffer.setSubstring(_:at:) to try using Substring.UTF8View.withContiguousStorageIfAvailable(_:) before falling back to setString(_:at:)

Result:

ByteBuffer.setSubstring(_:at:) (and callers such as writeSubstring(_:)) may see a minor performance improvement. (I have no measurements handy, unfortunately.)

Big shout-out to @Lukasa for a ton of help understanding what was going on in the existing code!

[Lukasa]

We should aim to add #1976 first, so that we have some perf numbers to compare with.

[Joannis]

@Lukasa and @gwynne , the mentioned PR has been merged. Time for a review of this PR?

[glbrntt]

@swift-nio-bot test perf please

[swift-server-bot]

performance report

build id: 102

timestamp: Fri Feb 4 17:39:31 UTC 2022

results

name	min	max	mean	std
write_http_headers	0.00405492	0.004175885	0.0040923003	4.232484518590841e-05
http_headers_canonical_form	0.086829406	0.089229209	0.0872961439	0.0007254444090792389
http_headers_canonical_form_trimming_whitespace	0.164904235	0.167407728	0.165662173	0.0007392950851129911
http_headers_canonical_form_trimming_whitespace_from_short_string	0.150754263	0.151376239	0.1511362885	0.00024851314829881237
http_headers_canonical_form_trimming_whitespace_from_long_string	0.235363974	0.236960889	0.235844609	0.0005197167599084876
bytebuffer_write_12MB_short_string_literals	0.514408006	0.520360082	0.5160775989	0.0017970149311249631
bytebuffer_write_12MB_short_calculated_strings	0.513224878	0.516033351	0.5146252976000001	0.0009513815786854592
bytebuffer_write_12MB_medium_string_literals	0.171625226	0.172280846	0.1719082817	0.00023730494257157743
bytebuffer_write_12MB_medium_calculated_strings	0.222987744	0.224650451	0.22358942399999998	0.0006044460630384385
bytebuffer_write_12MB_large_calculated_strings	0.143114963	0.143941991	0.1434703673	0.00023252303351474358
bytebuffer_lots_of_rw	0.450310764	0.4532688	0.451633731	0.0008314060352463263
bytebuffer_write_http_response_ascii_only_as_string	0.039458577	0.040001683	0.0395931796	0.00021372751450396925
bytebuffer_write_http_response_ascii_only_as_staticstring	0.030172709	0.030754806	0.0303394176	0.000173581558559786
bytebuffer_write_http_response_some_nonascii_as_string	0.038961372	0.050291023	0.0408093706	0.0034068322635947366
bytebuffer_write_http_response_some_nonascii_as_staticstring	0.030550324	0.031101807	0.0306433691	0.00016267072731984233
no-net_http1_10k_reqs_1_conn	0.102762967	0.104401503	0.10350213109999999	0.0004439550327282804
http1_10k_reqs_1_conn	0.594324334	0.601293987	0.5965681171	0.0021142389222182567
http1_10k_reqs_100_conns	0.583574152	0.587999262	0.5861890068	0.0014514911563618335
future_whenallsucceed_100k_immediately_succeeded_off_loop	0.071489075	0.072948061	0.071935213	0.0005001057711369094
future_whenallsucceed_100k_immediately_succeeded_on_loop	0.07228386	0.07964346	0.0736384688	0.002215333233283215
future_whenallsucceed_100k_deferred_off_loop	0.336391252	0.339732339	0.3377501602	0.0009983112646162699
future_whenallsucceed_100k_deferred_on_loop	0.123474959	0.126013901	0.1241963569	0.0008160016336525683
future_whenallcomplete_100k_immediately_succeeded_off_loop	0.030595095	0.030951022	0.030736747600000003	0.0001279711412990359
future_whenallcomplete_100k_immediately_succeeded_on_loop	0.03032962	0.03118259	0.0305202368	0.00027766565928340664
future_whenallcomplete_100k_deferred_off_loop	0.263713757	0.268633514	0.2656487625	0.0015445597274160027
future_whenallcomplete_100k_deferred_on_loop	0.062730434	0.068046366	0.0635416465	0.001601846175109294
future_reduce_10k_futures	0.036502099	0.037524324	0.0369687145	0.00031367593710397165
future_reduce_into_10k_futures	0.03676992	0.037650628	0.037252489300000004	0.00029169582515774147
channel_pipeline_1m_events	0.095156916	0.096207718	0.0954280299	0.00033910516940404247
websocket_encode_50b_space_at_front_1m_frames_cow	0.490869603	0.49207664	0.491406412	0.0004731045057797556
websocket_encode_50b_space_at_front_1m_frames_cow_masking	0.064270262	0.064796463	0.0644412634	0.00021281545403533598
websocket_encode_1kb_space_at_front_100k_frames_cow	0.051583124	0.052069575	0.0517571498	0.0002014406838946776
websocket_encode_50b_no_space_at_front_1m_frames_cow	0.485733729	0.487317834	0.4864989187	0.000611180448615627
websocket_encode_1kb_no_space_at_front_100k_frames_cow	0.051677264	0.052165175	0.051797388099999994	0.00019153585746419562
websocket_encode_50b_space_at_front_10k_frames	0.006341487	0.006398943	0.006379225299999999	2.0056074785183972e-05
websocket_encode_50b_space_at_front_10k_frames_masking	0.079065871	0.079565846	0.0792950106	0.00022796180794851538
websocket_encode_1kb_space_at_front_1k_frames	0.000740093	0.000777751	0.0007508603	1.3510225699652679e-05
websocket_encode_50b_no_space_at_front_10k_frames	0.006324753	0.006821112	0.0064198962	0.00014449439031771739
websocket_encode_1kb_no_space_at_front_1k_frames	0.000686982	0.000736435	0.0007054184	1.620670049495989e-05
websocket_decode_125b_100k_frames	0.11225767	0.113513326	0.11283606530000001	0.00040608414472784274
websocket_decode_125b_with_a_masking_key_100k_frames	0.115424102	0.116744775	0.1158322101	0.0003864769944397853
websocket_decode_64kb_100k_frames	0.117540276	0.118081649	0.11782260950000001	0.00022892417060107277
websocket_decode_64kb_with_a_masking_key_100k_frames	0.117918775	0.119213343	0.1183949553	0.0004026487355682086
websocket_decode_64kb_+1_100k_frames	0.115029435	0.115783834	0.1154126159	0.00025439306770099574
websocket_decode_64kb_+1_with_a_masking_key_100k_frames	0.118352001	0.13334149	0.12200683080000001	0.004316688467806617
circular_buffer_into_byte_buffer_1kb	0.040355591	0.041228143	0.0405690907	0.0002880015379812529
circular_buffer_into_byte_buffer_1mb	0.080619998	0.08154882	0.08090993799999999	0.0003105172347473018
byte_buffer_view_iterator_1mb	0.02005411	0.020573808	0.0201288079	0.00015877817587330902
byte_to_message_decoder_decode_many_small	0.173008061	0.179262609	0.1739959392	0.0018566000356277147
generate_10k_random_request_keys	0.089266058	0.090042958	0.08961197439999999	0.0002146249248242691
bytebuffer_rw_10_uint32s	0.298640306	0.300053861	0.2992350047	0.0005225954446880207
bytebuffer_multi_rw_10_uint32s	0.056738289	0.058401222	0.05739364500000001	0.0005783985823301749
lock_1_thread_10M_ops	0.155997604	0.15625471	0.1561369056	7.833430859562799e-05
lock_2_threads_10M_ops	0.906209402	0.95259591	0.9254757802	0.016579498142993764
lock_4_threads_10M_ops	0.893144464	0.933440425	0.9183880926000001	0.013278631733202735
lock_8_threads_10M_ops	0.9113968	0.929785921	0.9244142674999999	0.005976499589544957
schedule_10000_tasks	0.005072241	0.007488474	0.0054446022999999994	0.0007272359983777313
schedule_and_run_10000_tasks	0.031479341	0.032399223	0.031728509599999996	0.00026216309361421777
execute_10000	0.017135981	0.017581668	0.0172382222	0.00012719997554140256
bytebufferview_copy_to_array_1000_times_1kb	0.001431499	0.001482454	0.0014422794	1.5563585606579664e-05

comparison

name	current	previous	winner	diff
write_http_headers	0.00405492	0.004056786	current	0%
http_headers_canonical_form	0.086829406	0.086034723	previous	0%
http_headers_canonical_form_trimming_whitespace	0.164904235	0.161985731	previous	1%
http_headers_canonical_form_trimming_whitespace_from_short_string	0.150754263	0.148023866	previous	1%
http_headers_canonical_form_trimming_whitespace_from_long_string	0.235363974	0.229145041	previous	2%
bytebuffer_write_12MB_short_string_literals	0.514408006	0.516527711	current	0%
bytebuffer_write_12MB_short_calculated_strings	0.513224878	0.516246142	current	0%
bytebuffer_write_12MB_medium_string_literals	0.171625226	0.171232466	previous	0%
bytebuffer_write_12MB_medium_calculated_strings	0.222987744	0.223226254	current	0%
bytebuffer_write_12MB_large_calculated_strings	0.143114963	0.142444361	previous	0%
bytebuffer_lots_of_rw	0.450310764	0.427145566	previous	5%
bytebuffer_write_http_response_ascii_only_as_string	0.039458577	0.039642268	current	0%
bytebuffer_write_http_response_ascii_only_as_staticstring	0.030172709	0.030910666	current	-2%
bytebuffer_write_http_response_some_nonascii_as_string	0.038961372	0.039018965	current	0%
bytebuffer_write_http_response_some_nonascii_as_staticstring	0.030550324	0.03018811	previous	1%
no-net_http1_10k_reqs_1_conn	0.102762967	0.104314793	current	-1%
http1_10k_reqs_1_conn	0.594324334	0.594263813	previous	0%
http1_10k_reqs_100_conns	0.583574152	0.582956895	previous	0%
future_whenallsucceed_100k_immediately_succeeded_off_loop	0.071489075	0.070942767	previous	0%
future_whenallsucceed_100k_immediately_succeeded_on_loop	0.07228386	0.0718041	previous	0%
future_whenallsucceed_100k_deferred_off_loop	0.336391252	0.33674867	current	0%
future_whenallsucceed_100k_deferred_on_loop	0.123474959	0.122631832	previous	0%
future_whenallcomplete_100k_immediately_succeeded_off_loop	0.030595095	0.030588381	previous	0%
future_whenallcomplete_100k_immediately_succeeded_on_loop	0.03032962	0.030240294	previous	0%
future_whenallcomplete_100k_deferred_off_loop	0.263713757	0.263586644	previous	0%
future_whenallcomplete_100k_deferred_on_loop	0.062730434	0.062170827	previous	0%
future_reduce_10k_futures	0.036502099	0.037413122	current	-2%
future_reduce_into_10k_futures	0.03676992	0.037258358	current	-1%
channel_pipeline_1m_events	0.095156916	0.098732162	current	-3%
websocket_encode_50b_space_at_front_1m_frames_cow	0.490869603	0.487616549	previous	0%
websocket_encode_50b_space_at_front_1m_frames_cow_masking	0.064270262	0.064452606	current	0%
websocket_encode_1kb_space_at_front_100k_frames_cow	0.051583124	0.05145914	previous	0%
websocket_encode_50b_no_space_at_front_1m_frames_cow	0.485733729	0.487476365	current	0%
websocket_encode_1kb_no_space_at_front_100k_frames_cow	0.051677264	0.051416387	previous	0%
websocket_encode_50b_space_at_front_10k_frames	0.006341487	0.006390503	current	0%
websocket_encode_50b_space_at_front_10k_frames_masking	0.079065871	0.080231077	current	-1%
websocket_encode_1kb_space_at_front_1k_frames	0.000740093	0.000734463	previous	0%
websocket_encode_50b_no_space_at_front_10k_frames	0.006324753	0.006351988	current	0%
websocket_encode_1kb_no_space_at_front_1k_frames	0.000686982	0.000678678	previous	1%
websocket_decode_125b_100k_frames	0.11225767	0.111094187	previous	1%
websocket_decode_125b_with_a_masking_key_100k_frames	0.115424102	0.113983527	previous	1%
websocket_decode_64kb_100k_frames	0.117540276	0.113875236	previous	3%
websocket_decode_64kb_with_a_masking_key_100k_frames	0.117918775	0.116626998	previous	1%
websocket_decode_64kb_+1_100k_frames	0.115029435	0.11384766	previous	1%
websocket_decode_64kb_+1_with_a_masking_key_100k_frames	0.118352001	0.116651685	previous	1%
circular_buffer_into_byte_buffer_1kb	0.040355591	0.040483449	current	0%
circular_buffer_into_byte_buffer_1mb	0.080619998	0.080568961	previous	0%
byte_buffer_view_iterator_1mb	0.02005411	0.020052445	previous	0%
byte_to_message_decoder_decode_many_small	0.173008061	0.172836061	previous	0%
generate_10k_random_request_keys	0.089266058	0.089073831	previous	0%
bytebuffer_rw_10_uint32s	0.298640306	0.295447307	previous	1%
bytebuffer_multi_rw_10_uint32s	0.056738289	0.055208787	previous	2%
lock_1_thread_10M_ops	0.155997604	0.155925283	previous	0%
lock_2_threads_10M_ops	0.906209402	0.839418602	previous	7%
lock_4_threads_10M_ops	0.893144464	0.910298511	current	-1%
lock_8_threads_10M_ops	0.9113968	0.94410947	current	-3%
schedule_10000_tasks	0.005072241	0.005022541	previous	0%
schedule_and_run_10000_tasks	0.031479341	0.031619862	current	0%
execute_10000	0.017135981	0.017408752	current	-1%
bytebufferview_copy_to_array_1000_times_1kb	0.001431499	0.001438482	current	0%

significant differences found

[Lukasa]

LGTM, thanks!