I was trying out my new 2 x 1Ghz G4, and discovered a number of things. The benchmark was > hilbert<-function(n) 1/(outer(seq(n),seq(n),"+")-1) > system.time(eigen(hilbert(xxx))) where xxx is 500 or 1000. I compare 4 implementations of R on the same machine (running Mac OS X 10.1.3 and OS 9.2.2). In the first place the Carbon version uses about 19 secs for 500 and 160 secs for 1000. This is about 10% faster than on my previous Mac (Dual 800 MHz). But the Darwin versions (both development 1.5.0 and official 1.4.1) use 4.7 seconds for 500 and 50 seconds for 1000 - i.e. the Darwin version is three times as fast as the Carbon version (with this benchmark and with this machine). The 1.5.0 uses ATLAS, the 1.4.1 does not, and that does not seem to make a difference. But (weirdness) R-1.4.1 patched (compiled in exactly the same way as R-devel on the same machine) uses 13.5 seconds for 500 and 115 seconds for 1000, i.e. only 30% faster than Carbon, and more than twice as slow as the official 1.4.1 and as R-devel. === Jan de Leeuw; Professor and Chair, UCLA Department of Statistics; US mail: 9432 Boelter Hall, Box 951554, Los Angeles, CA 90095-1554 phone (310)-825-9550; fax (310)-206-5658; email: deleeuw@stat.ucla.edu homepage: http://www.stat.ucla.edu/~deleeuw ======================================================== No matter where you go, there you are. --- Buckaroo Banzai http://www.stat.ucla.edu/~deleeuw/sounds/nomatter.au ========================================================
some tests with new Mac
2 messages · Jan de Leeuw, Roger D. Peng
1 day later
For using eigen() and ATLAS, maybe La.eigen() would be faster with ATLAS? Haven't tried it yet on the Mac.... -roger _______________________________ UCLA Department of Statistics rpeng@stat.ucla.edu http://www.stat.ucla.edu/~rpeng
On Sat, 2 Feb 2002, Jan de Leeuw wrote:
I was trying out my new 2 x 1Ghz G4, and discovered a number of things. The benchmark was
> hilbert<-function(n) 1/(outer(seq(n),seq(n),"+")-1) > system.time(eigen(hilbert(xxx)))
where xxx is 500 or 1000. I compare 4 implementations of R on the same machine (running Mac OS X 10.1.3 and OS 9.2.2). In the first place the Carbon version uses about 19 secs for 500 and 160 secs for 1000. This is about 10% faster than on my previous Mac (Dual 800 MHz). But the Darwin versions (both development 1.5.0 and official 1.4.1) use 4.7 seconds for 500 and 50 seconds for 1000 - i.e. the Darwin version is three times as fast as the Carbon version (with this benchmark and with this machine). The 1.5.0 uses ATLAS, the 1.4.1 does not, and that does not seem to make a difference. But (weirdness) R-1.4.1 patched (compiled in exactly the same way as R-devel on the same machine) uses 13.5 seconds for 500 and 115 seconds for 1000, i.e. only 30% faster than Carbon, and more than twice as slow as the official 1.4.1 and as R-devel. === Jan de Leeuw; Professor and Chair, UCLA Department of Statistics; US mail: 9432 Boelter Hall, Box 951554, Los Angeles, CA 90095-1554 phone (310)-825-9550; fax (310)-206-5658; email: deleeuw@stat.ucla.edu homepage: http://www.stat.ucla.edu/~deleeuw ======================================================== No matter where you go, there you are. --- Buckaroo Banzai http://www.stat.ucla.edu/~deleeuw/sounds/nomatter.au ========================================================
_______________________________________________ StatCompute mailing list StatCompute@lists.stat.ucla.edu http://lists.stat.ucla.edu/mailman/listinfo.cgi/statcompute