[GiNaC-devel] Univariate GCD benchmark (Was: some questions)

[Alexei Sheplyakov]

Dear Jens,

On Tue, Feb 03, 2009 at 02:27:36PM +0100, Jens Vollinga wrote:

> The timings
>
> timing univariate GCD
> GCD of relatively prime polynomials a and b
> degree(a) = 100, degree(b) = 50
> mod_gcd         : 1 1.36e-02
> heur_gcd        : 1 1.82e-03
> sr_gcd:         : 1 1.43e+00
> Non-trivial GCD, degree(a) = 100, degree(b) = 100
> mod_gcd         : 51 1.36e-02
> heur_gcd        : 51 3.64e-03
> heur_gcd (old)  : 51 5.73e-02
> sr_gcd:         : 51 2.31e-01
>
> seem to stop at that last output ... I didn't put it through a debugger  
> yet, maybe the test is just too expensive?

The benchmark works for me just fine, and it doesn't take too much time:

timing univarite GCD
GCD of relatively prime polynomials a and b
degree(a) = 100, degree(b) = 50
mod_gcd		: 1 2.18e-03
heur_gcd	: 1 3.08e-04
sr_gcd:		: 1 2.43e-01
Non-trivial GCD, degree(a) = 100, degree(b) = 100
mod_gcd		: 51 5.45e-03
heur_gcd	: 51 3.64e-04
heur_gcd (old)	: 51 1.75e-02
sr_gcd:		: 51 4.65e-02
sr_gcd (old)	: 51 1.62e-01
Non-trivial GCD, degree(a) = 1000, degree(b) = 1000
mod_gcd		: 501 2.17e+00
heur_gcd	: 501 1.24e-02
heur_gcd (old)	: 501 2.57e+00
Non-trivial GCD, degree(a) = 2000, degree(b) = 2000
mod_gcd		: 1001 1.53e+01
heur_gcd	: 1001 4.73e-02
heur_gcd (old)	: 1001 1.34e+01
. 

Actually, it's designed to avoid longish calculations (in order to not
annoy users so they don't skip the `make check' step). So I need more
info to find out what is going on.

Best regards,
	Alexei

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: Digital signature
URL: <http://www.cebix.net/pipermail/ginac-devel/attachments/20090205/1d5c3359/attachment.sig>