Performance vs. Pathscale EKO on AMD64

> They do not cover Sun Studio 11 alpha compilers there.

Sure, they use product quality releases only

> I would be happy with anecdotal info.

We are performing our own benchmarking so comparison results will be published soon

But now... I hope you will be satisfied with data below. You can see that in some tests Sun wins, in other tests Pathscale wins, and geometric mean times is just slightly different. I interpret it as performance parity

Pathscale EKO, Sun Java Workstation W2100z ( http://www.sun.com/desktop/workstation/w2100z/index.jsp )

Test name - Run time

-

ac-9.57

aermod -34.79

air- 13.19

capacita -56.73

channel -16.21

doduc- 36.83

fatigue-5.84

gas_dyn -8.04

induct -28.51

linpk-24.96

mdbx -18.38

nf- 22.96

protein -53.52

rnflow-26.70

test_fpu-16.85

tfft-7.15

Geometric Mean Execution Time =19.27 seconds

--

Sun Studio eXpress 2, Sun Java Workstation W2100z ( http://www.sun.com/desktop/workstation/w2100z/index.jsp )

Test name - Run time

-

ac-19.30

aermod - 29.97

air- 12.32

capacita-81.78

channel-10.99

doduc- 43.17

fatigue-8.57

gas_dyn -9.97

induct -39.66

linpk-22.45

mdbx -16.92

nf-25.86

protein -63.04

rnflow -41.21

test_fpu -14.51

tfft- 7.11

Geometric Mean Execution Time =21.73 seconds

alexey_shubin at 2007-7-7 0:38:51 >