View mode: basic / threaded / horizontal-split · Log in · Help
December 09, 2005
Radically different performance on same hardware.
In the "Computer Language Shootout" D does pretty well, except for a couple of
the benchmarks.

One of those that it does poorly on is:

http://shootout.alioth.debian.org/benchmark.php?test=revcomp&lang=dlang&id=0

On their test box, it takes ~10 secs. On my machine (a P4 2.2G, 512 MB RAM,
Fedora Core 3) it takes 0.68 secs. The test machine and mine are evenly matched
except it is running Debian Linux 'unstable', Kernel 2.6.8-1-k7.

They have another box that looks hardware-wise identical to mine (running Gentoo
2005.1 stage 3):

http://shootout.alioth.debian.org/gp4/benchmark.php?test=revcomp&lang=dlang&id=0

On that machine, the same test takes ~9 secs. compared to (again) 0.68 secs on
mine.

There are somethings about the other test machines I'm not privy to like cache
size, swap file size, etc... but that couldn't make that big of difference for
just one test, could it?

Anybody have any idea what could be going on here, or run into a similiar
problem between Fedora and Debian?

Thanks.
December 09, 2005
Re: Radically different performance on same hardware.
It's pretty hard on the heap ... used  ~26MB?


"Dave" <Dave_member@pathlink.com> wrote in message 
news:dnamf5$2ls5$1@digitaldaemon.com...
>
> In the "Computer Language Shootout" D does pretty well, except for a 
> couple of
> the benchmarks.
>
> One of those that it does poorly on is:
>
> http://shootout.alioth.debian.org/benchmark.php?test=revcomp&lang=dlang&id=0
>
> On their test box, it takes ~10 secs. On my machine (a P4 2.2G, 512 MB 
> RAM,
> Fedora Core 3) it takes 0.68 secs. The test machine and mine are evenly 
> matched
> except it is running Debian Linux 'unstable', Kernel 2.6.8-1-k7.
>
> They have another box that looks hardware-wise identical to mine (running 
> Gentoo
> 2005.1 stage 3):
>
> http://shootout.alioth.debian.org/gp4/benchmark.php?test=revcomp&lang=dlang&id=0
>
> On that machine, the same test takes ~9 secs. compared to (again) 0.68 
> secs on
> mine.
>
> There are somethings about the other test machines I'm not privy to like 
> cache
> size, swap file size, etc... but that couldn't make that big of difference 
> for
> just one test, could it?
>
> Anybody have any idea what could be going on here, or run into a similiar
> problem between Fedora and Debian?
>
> Thanks.
>
>
December 09, 2005
Re: Radically different performance on same hardware.
"Kris" <fu@bar.com> wrote in message news:dnar25$2pmq$1@digitaldaemon.com...
> It's pretty hard on the heap ... used  ~26MB?
>

That's a good point. The other test that bombs is one using thread 
synchronization, which is used in the GC also. Hmmm - most of the pthreads 
portion of the Linux threads implementation seems to work fine because the 
tcp test uses threads and runs Ok. The synchronization uses mutexes and part 
of the thread impl. uses semaphores.

Maybe a pattern?

Anybody know of problems with semaphores or mutexes on the unstable/newer 
versions of Debian?

Thanks.

(re-posting to digitalmars.D.bugs)

>
> "Dave" <Dave_member@pathlink.com> wrote in message 
> news:dnamf5$2ls5$1@digitaldaemon.com...
>>
>> In the "Computer Language Shootout" D does pretty well, except for a 
>> couple of
>> the benchmarks.
>>
>> One of those that it does poorly on is:
>>
>> http://shootout.alioth.debian.org/benchmark.php?test=revcomp&lang=dlang&id=0
>>
>> On their test box, it takes ~10 secs. On my machine (a P4 2.2G, 512 MB 
>> RAM,
>> Fedora Core 3) it takes 0.68 secs. The test machine and mine are evenly 
>> matched
>> except it is running Debian Linux 'unstable', Kernel 2.6.8-1-k7.
>>
>> They have another box that looks hardware-wise identical to mine (running 
>> Gentoo
>> 2005.1 stage 3):
>>
>> http://shootout.alioth.debian.org/gp4/benchmark.php?test=revcomp&lang=dlang&id=0
>>
>> On that machine, the same test takes ~9 secs. compared to (again) 0.68 
>> secs on
>> mine.
>>
>> There are somethings about the other test machines I'm not privy to like 
>> cache
>> size, swap file size, etc... but that couldn't make that big of 
>> difference for
>> just one test, could it?
>>
>> Anybody have any idea what could be going on here, or run into a similiar
>> problem between Fedora and Debian?
>>
>> Thanks.
>>
>>
>
>
Top | Discussion index | About this forum | D home