buffer: make indexOf faster #1059
Conversation
Improve the performance of IndexOf() for Buffer#indexOf() searches.
trevnorris
force-pushed
the
supa-fast-indexof
branch
from
36cea64
to
e3cdaab
Compare
|
okay. so it seems that V8's string match is still crazy faster than this. still have a ways to go. |
| do { | ||
| ptr = memchr(inc, *needle, h_length - static_cast<int32_t>(inc - haystack)); | ||
| if (ptr != nullptr) { | ||
| const void* ptr2 = memmem(inc, |
There was a problem hiding this comment.
Guess it isn't: https://msdn.microsoft.com/en-us/library/aa246468(v=vs.60).aspx
There was a problem hiding this comment.
suck, seriously? alright, back to the drawing board.
There was a problem hiding this comment.
Maybe just pick a "good" algorithm from here to use on Windows (or everywhere for consistency?). Start with Turbo Boyer-Moore or Boyer-Moore-Horspool perhaps?
There was a problem hiding this comment.
@mscdex TBH this implementation was thrown together in 10 mins, and is by no means a good implementation. i'll take the time to read up more on proper Boyer-Moore implementations. Thanks for the links.
|
@trevnorris can we close this and re-open a new one when you come up with a good way of doing it? |
|
@Fishrock123 Sure. |
Improve the performance of IndexOf() for Buffer#indexOf() searches.
NOTE: This is very ugly right now, but sooo much faster. Looking for feedback at the moment.
R=@bnoordhuis