Vectorized memcpy on G4 Cube

Shaos March 02, 2009 altivec, cube, linux, opensource, ppc No Comments

VN:F [1.9.11_1134]
Rating: 0 (from 0 votes)
VN:F [1.9.11_1134]
Rating: 0.0/5 (0 votes cast)

Again I tested new hardware using my memcpy benchmark – now for my 1st desktop on Intel Core 2 Duo 2.6 GHz and my 2nd desktop on PowerPC G4 450 MHz (black G4 Cube). Here you can see not-aligned (nal) and aligned (al) results and also “vectorized” results (G4-450v-nal and G4-450v-al) where memcpy was AltiVec optimized version from libfreevec v1.0.4.

P.S. If you want optimized AltiVec versions of such functions as bmove512, memccpy, memchr, memcmp, memcpy, memfrob, memmove, mempcpy, memrchr, memset, strcmp, strcpy, strlen, strncmp, strncpy, strnlen and swab for everything in your Linux system, then just put (of course after successful libfreevec installation) line /usr/local/lib/libfreevec_libc.so to file /etc/ld.so.preload (you can create this file if it’s not yet available) – thanks to Konstantinos Margaritis (author of libfreevec)!

You can follow any responses to this entry through the RSS 2.0 feed. You can skip to the end and leave a response. Pinging is currently not allowed.

Leave a Reply

You must be logged in to post a comment.