You are welcome! ;)
I'm sorry it was pos and type, not al_src and al_dst...
Anyway I didn't understand how strcmp may be visualized in one graph...

I've just prepared some graphs from my PowerBook G4 1.67 GHz (for bmove512, memcpy, memset, strcpy, strlen):
http://shaos.net/blog/2008/04/some-vect ... ok-g4.html

_________________
Alexander Shabarshin

Well, if I got all the cases it would need many special graphs, but I only included the case where comparing equal strings. Perhaps I should include more graphs :-/
Wow, the G4 at that speed performs very nicely. If you could send me the CSV data, I'd gladly include the charts in freevec.org with a link to your blog :)

Thanks! This is archive with CSVs, GIFs and sqlite3's benchmarks.db (and also dumped SQL). I run these benchmarks only for 6 functions (including strcmp, but this one was not good probably because of my random.dat had zeros sometimes and I didn't render image for it):

http://shaos.net/blog/files/g4_1670_tests_080425.zip

Later I will publish the same tests for PowerMac G4 400MHz and will run full benchsuite from your SVN on PowerBook G4 1.67GHz.

_________________
Alexander Shabarshin

The same functions (bmove512, memcpy, memset, strcpy, strlen) on my PowerMac G4 400MHz:

http://shaos.net/blog/2008/04/some-vect ... ac-g4.html

Original data files: http://shaos.net/blog/files/g4_400_tests_080426.zip

P.S. On my and your data I see strange peak for G4 vectorized unaligned (!) strcpy for size 6144. Why is it so strange?

_________________
Alexander Shabarshin

More than 2 days (and nights ;) of running benchsuite on my PowerBook G4 1.67 GHz:

http://shaos.net/blog/files/g4_1670_080428.zip (560K)

_________________
Alexander Shabarshin