Re: make bench trouble with cygwin

[Bill Allombert <allomber@math.u-bordeaux.fr>]

On Fri, Jun 03, 2005 at 06:13:19PM +0200, Alain SMEJKAL wrote:
> Dear list,
> 
> Have compiled
>    GP/PARI CALCULATOR Version 2.2.11 (development)
>    i686 running cygwin (ix86 kernel) 32-bit version
>   compiled: Jun  3 2005, gcc-3.3.1 (cygming special)
>  (readline v4.3 enabled, extended help not available)
> 
> make bench reports
> 
> * Testing objets        for gp-sta..TIME=16     for gp-dyn..TIME=16
> * Testing analyz        for gp-sta..TIME=110    for gp-dyn..TIME=110
> * Testing number        for gp-sta..TIME=94     for gp-dyn..TIME=78
> * Testing polyser       for gp-sta..TIME=47     for gp-dyn..TIME=31
> * Testing linear        for gp-sta..TIME=63     for gp-dyn..TIME=63
> * Testing elliptic      for gp-sta..TIME=78     for gp-dyn..TIME=62
> * Testing sumiter       for gp-sta..TIME=78     for gp-dyn..TIME=62
> * Testing graph         for gp-sta..TIME=47     for gp-dyn..TIME=47
> * Testing program       for gp-sta..BUG [94]    for gp-dyn..TIME=94
> * Testing trans         for gp-sta..TIME=250    for gp-dyn..TIME=235
> * Testing nfields       for gp-sta..TIME=641    for gp-dyn..TIME=657
> +++ [BUG] Total bench for gp-sta is 1005
> +++ Total bench for gp-dyn is 929
> 
> Get this in program-sta.dif : *** addii: bug in GP (Segmentation Fault), please report
> Get the same with some manual function install try
> 
> 1) is this the known function install problem as explained in users doc or something went wrong in compil configuration ?

No, this is the install function. I thought Karim managed to fix this
problem, however.

> 2) gp-stat was expected to be faster. Except nfields, it is not. Is this due to cygwin use ?

No, this is because we do not compile libpari with -fPIC on x86, so
there are no real differences. Anyway due to lazy linking, 'make bench'
is slightly biased in favor gp-dyn.

Cheers,
Bill