Re: CopyReadLineText optimization

Andrew Dunstan wrote:
> Another question that occurred to me - did you try using strpbrk() to
> look for the next interesting character rather than your homegrown
> searcher gadget? If so, how did that perform?

It looks like strpbrk() performs poorly:

unpatched:
testname | min duration
----------+-----------------
all | 00:00:08.099656
1/2 | 00:00:06.734241
1/4 | 00:00:06.016946
1/8 | 00:00:05.622122
1/16 | 00:00:05.304252
none | 00:00:05.155755
(6 rows)

strpbrk:

testname | min duration
----------+-----------------
all | 00:00:22.980997
1/2 | 00:00:13.724834
1/4 | 00:00:08.980246
1/8 | 00:00:06.582357
1/16 | 00:00:05.291485
none | 00:00:06.239468
(6 rows)

memchr:

testname | min duration
----------+-----------------
all | 00:00:13.684479
1/2 | 00:00:09.509213
1/4 | 00:00:06.921959
1/8 | 00:00:05.654761
1/16 | 00:00:04.719027
none | 00:00:03.718361
(6 rows)

Attached is the test script and patches I used, if someone wants to test
this on another platform.

--
Heikki Linnakangas
EnterpriseDB http://www.enterprisedb.com