benchmark

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

benchmark

Francesco Potorti`-9
Something strange happens when benchmarking octave 2.0.5 on Sun and
Alpha machines, because the Schur decomposition is clearly slower than
in octave 1.1.1, by around 30%.  This slowdown does not show on Intel
platforms.  Don't know about other platforms.

John Eaton says nothing should have changed that could justify this
behaviour.

Given the state of things, I will delay the publication of benchmark
results for octave version 2 until this issue becomes clearer.

--
Francesco Potorti` (researcher)        Voice:    +39-50-593203
Computer Network Division              Operator: +39-50-593211
CNUCE-CNR, Via Santa Maria 36          Fax:      +39-50-904052
56126 Pisa - Italy                     Email:    [hidden email]

Reply | Threaded
Open this post in threaded view
|

benchmark

John W. Eaton-6
On 17-Mar-1997, Francesco Potorti` <[hidden email]> wrote:

| Something strange happens when benchmarking octave 2.0.5 on Sun and
| Alpha machines, because the Schur decomposition is clearly slower than
| in octave 1.1.1, by around 30%.  This slowdown does not show on Intel
| platforms.  Don't know about other platforms.
|
| John Eaton says nothing should have changed that could justify this
| behaviour.
|
| Given the state of things, I will delay the publication of benchmark
| results for octave version 2 until this issue becomes clearer.

Well, something apparently did change, but I don't know what (yet).

Can someone verify that there were not significant changes to the
src/schur.cc file (it was called src/f-schur.cc in 1.1.1) and the
Lapack subroutines that are used to compute the Schur factorization?

If there were no significant changes in the code, can someone try
doing some profiling to try to find out what's going on?

Thanks,

jwe