develooper Front page | perl.perl5.porters | Postings from August 2014

Suspected reason for the COW performance regression on short strings

Thread Next
From:
Reini Urban
Date:
August 27, 2014 04:23
Subject:
Suspected reason for the COW performance regression on short strings
Message ID:
CAHiT=DH6QZG_BouxJ8yf=ahNnb0F0zhhi28Ym=s8wyKq8Vpo6A@mail.gmail.com
This might interest you
https://code.google.com/p/perl-compiler/issues/detail?id=372

I think I've found out why short COW strings are so slow now.
COW strings now allocate 10 byte instead of 8 for short strings len < 6.
Even for CUR=0. It should be 8 of course.

Rather align to wordsize after adding the 2 slack bytes (first fir \0,
2nd for the COW refcount), not align first and then add the 2 bytes to
the LEN

I didn't patch core or verify it yet, as I'm traveling to europe for a
week, and
will have no time there, but it fixed my 5.18-5.20 issues with perlcc.
-- 
Reini Urban
http://cpanel.net/   http://www.perl-compiler.org/

Thread Next


nntp.perl.org: Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at ask@perl.org | Group listing | About