Re: GMP 4.3.0 released

[Roberto Bagnara]

Torbjorn Granlund wrote:
> It is my great pleasure to announce the availability a new major release of
> the GNU Multiple Precision Arithmetic Library (GMP).  The new release is
> identified as 4.3.0.

Hi Torbjorn,

congratulations and thank you for the new release.


>   Speedups:
>   * Vastly improved assembly code for x86-64 processors from AMD and Intel.

I have run a couple of application programs based on the Parma Polyhedra
Library (PPL), which heavily depends on GMP.  On an AMD Opteron 2384 "Shanghai"
I got the following running times (best run out of 10):

         Application A                   Application B

     GMP 4.2.4    GMP 4.3.0          GMP 4.2.4    GMP 4.3.0

       238 s        187 s              34.2 s       34.8 s

On an Intel Core2 Q9400, instead, I got:

         Application A                   Application B

     GMP 4.2.4    GMP 4.3.0          GMP 4.2.4    GMP 4.3.0

       339 s        213 s              29.8 s       28.9 s


>   Misc:
>   * The gmp_version variable now always contains three parts.  For this
>     release, it is "4.3.0".

Aaargh!  This broke our GMP detection procedure.  Moreover,
PPL 0.10.1 was released only a few hours before the release of
GMP 4.3.0.  I guess now we have no choice but release PPL 0.10.2
with a revised GMP detection procedure.

Was a release candidate made available and announced somewhere?
If so, I would like to subscribe the list for this kind of
announcements.
Thanks again,

     Roberto

--
Prof. Roberto Bagnara
Computer Science Group
Department of Mathematics, University of Parma, Italy
http://www.cs.unipr.it/~bagnara/
mailto:bagnara@...
_______________________________________________
gmp-discuss mailing list
gmp-discuss@...
http://gmplib.org/mailman/listinfo/gmp-discuss

[Dennis Clarke-2]

> Torbjorn Granlund wrote:
>> It is my great pleasure to announce the availability a new major release
>> of
>> the GNU Multiple Precision Arithmetic Library (GMP).  The new release is
>> identified as 4.3.0.
>
> Hi Torbjorn,
>
> congratulations and thank you for the new release.
>

With the most minor tweak it builds fine on Solaris :

ABI=32

$ ls .libs/lib*
.libs/libgmp.a           .libs/libgmp.so.3        .libs/libgmpxx.lai
.libs/libgmp.la          .libs/libgmp.so.3.5.0    .libs/libgmpxx.so
.libs/libgmp.lai         .libs/libgmpxx.a         .libs/libgmpxx.so.4
.libs/libgmp.so          .libs/libgmpxx.la        .libs/libgmpxx.so.4.1.0
$
$
$ file .libs/libgmp.so.3.5.0
.libs/libgmp.so.3.5.0:  ELF 32-bit MSB dynamic lib SPARC Version 1,
dynamically linked, not stripped
$ file .libs/libgmpxx.so.4.1.0
.libs/libgmpxx.so.4.1.0:        ELF 32-bit MSB dynamic lib SPARC Version
1, dynamically linked, not stripped

and ABI=64

$ file .libs/libgmp.so.3.5.0
.libs/libgmp.so.3.5.0:  ELF 64-bit MSB dynamic lib SPARCV9 Version 1,
dynamically linked, not stripped
$ file .libs/libgmpxx.so.4.1.0
.libs/libgmpxx.so.4.1.0:        ELF 64-bit MSB dynamic lib SPARCV9 Version
1, dynamically linked, not stripped

All tests pass for libgmp.so.3.5.0 however I have to get in and manually
run the tests for the cxx libs.

Dennis Clarke
http://www.blastwave.org


_______________________________________________
gmp-discuss mailing list
gmp-discuss@...
http://gmplib.org/mailman/listinfo/gmp-discuss

[Torbjorn Granlund-6]

Roberto Bagnara <bagnara@...> writes:

  >   Speedups:
  >   * Vastly improved assembly code for x86-64 processors from AMD and Intel.
 
  I have run a couple of application programs based on the Parma Polyhedra
  Library (PPL), which heavily depends on GMP.  On an AMD Opteron 2384 "Shanghai"
  I got the following running times (best run out of 10):
 
          Application A                   Application B
 
      GMP 4.2.4    GMP 4.3.0          GMP 4.2.4    GMP 4.3.0
 
        238 s        187 s              34.2 s       34.8 s
 
  On an Intel Core2 Q9400, instead, I got:
 
          Application A                   Application B
 
      GMP 4.2.4    GMP 4.3.0          GMP 4.2.4    GMP 4.3.0
 
        339 s        213 s              29.8 s       28.9 s
 
You get less speedup than I expect to be typical (and a very slight
slowdown in one case).

I am curious what GMP operations application B performs.  Profile runs
for both GMP versions would be helpful.

  >   Misc:
  >   * The gmp_version variable now always contains three parts.  For this
  >     release, it is "4.3.0".
 
  Aaargh!  This broke our GMP detection procedure.  Moreover,
  PPL 0.10.1 was released only a few hours before the release of
  GMP 4.3.0.  I guess now we have no choice but release PPL 0.10.2
  with a revised GMP detection procedure.

I am sorry about that!  I think version GMP detection will be simpler
with the new scheme, but this slight incompatibility will bite during a
transitional period.
 
  Was a release candidate made available and announced somewhere?

No, we did only internal testing this time, since our testing has been
much extended.

--
Torbjörn
_______________________________________________
gmp-discuss mailing list
gmp-discuss@...
http://gmplib.org/mailman/listinfo/gmp-discuss

[Roberto Bagnara]

Torbjorn Granlund wrote:

> Roberto Bagnara <bagnara@...> writes:
>
>   >   Speedups:
>   >   * Vastly improved assembly code for x86-64 processors from AMD and Intel.
>  
>   I have run a couple of application programs based on the Parma Polyhedra
>   Library (PPL), which heavily depends on GMP.  On an AMD Opteron 2384 "Shanghai"
>   I got the following running times (best run out of 10):
>  
>           Application A                   Application B
>  
>       GMP 4.2.4    GMP 4.3.0          GMP 4.2.4    GMP 4.3.0
>  
>         238 s        187 s              34.2 s       34.8 s
>  
>   On an Intel Core2 Q9400, instead, I got:
>  
>           Application A                   Application B
>  
>       GMP 4.2.4    GMP 4.3.0          GMP 4.2.4    GMP 4.3.0
>  
>         339 s        213 s              29.8 s       28.9 s
>  
> You get less speedup than I expect to be typical (and a very slight
> slowdown in one case).
>
> I am curious what GMP operations application B performs.  Profile runs
> for both GMP versions would be helpful.

Hi Torbjorn,

I was in fact wrong.  While the PPL is, generally speaking, heavily
dependent on GMP, application B running time is dominated by non-GMP
functions.  For it, the most important GMP operation is __gmpn_popcount
which consumes 12% of the running time.
Thanks again,

     Roberto

--
Prof. Roberto Bagnara
Computer Science Group
Department of Mathematics, University of Parma, Italy
http://www.cs.unipr.it/~bagnara/
mailto:bagnara@...
_______________________________________________
gmp-discuss mailing list
gmp-discuss@...
http://gmplib.org/mailman/listinfo/gmp-discuss