Alex Queiroz | 2 Aug 2012 16:09
Picon
Gravatar

Running the benchmarks

Hallo,

With the advent of clang 4.0 I would like to run the Gambit
benchmarks[1] so as to compare with GCC 4.7. Is there an
easy/convenient way to download and run the benchmarks?

[1] - http://www.iro.umontreal.ca/~gambit/bench.html

Cheers,
--

-- 
-alex
http://www.artisancoder.com/
Marc Feeley | 3 Aug 2012 08:06
Picon
Gravatar

Re: Running the benchmarks

The benchmarks are in misc/bench.tgz .  After untaring the file, you will have the benchmark sources and
some scripts to run the benchmarks and generate HTML tables of the results.

I'm not at my computer now so I can't give you more details.

Marc

Sent from my iPhone

On 2012-08-02, at 4:09 PM, Alex Queiroz <asandroq <at> gmail.com> wrote:

> Hallo,
> 
> With the advent of clang 4.0 I would like to run the Gambit
> benchmarks[1] so as to compare with GCC 4.7. Is there an
> easy/convenient way to download and run the benchmarks?
> 
> [1] - http://www.iro.umontreal.ca/~gambit/bench.html
> 
> Cheers,
> -- 
> -alex
> http://www.artisancoder.com/
> _______________________________________________
> Gambit-list mailing list
> Gambit-list <at> iro.umontreal.ca
> https://webmail.iro.umontreal.ca/mailman/listinfo/gambit-list
Alex Queiroz | 3 Aug 2012 15:00
Picon
Gravatar

Re: Running the benchmarks

Hallo,

On Fri, Aug 3, 2012 at 8:06 AM, Marc Feeley <feeley <at> iro.umontreal.ca> wrote:
> The benchmarks are in misc/bench.tgz .  After untaring the file, you will have the benchmark sources and
some scripts to run the benchmarks and generate HTML tables of the results.
>

Doh! I searched for them but clearly I did it poorly. Thanks for
pointing me to the scripts. I have run some benchmarks and it seems
that GCC still is better at compiling Gambit-C code.

Cheers,
--

-- 
-alex
http://www.artisancoder.com/

The following tables contain the execution time of the Gambit benchmarks on various implementations of Scheme. For a given benchmark, the entry in green indicates which Scheme system has the fastest execution and the number given is the time in milliseconds. Other entries give the execution time relative to the green entry. Blank entries indicate that this benchmark was not executed (possibly because the system did not accept to compile the program).

The first table gives CPU time and the second gives real time.

CPU time

(Continue reading)

Marc Feeley | 3 Aug 2012 16:35
Picon
Gravatar

Re: Running the benchmarks

Could you report your findings?

Marc

On 2012-08-03, at 3:00 PM, Alex Queiroz <asandroq <at> gmail.com> wrote:

> Hallo,
> 
> On Fri, Aug 3, 2012 at 8:06 AM, Marc Feeley <feeley <at> iro.umontreal.ca> wrote:
>> The benchmarks are in misc/bench.tgz .  After untaring the file, you will have the benchmark sources and
some scripts to run the benchmarks and generate HTML tables of the results.
>> 
> 
> Doh! I searched for them but clearly I did it poorly. Thanks for
> pointing me to the scripts. I have run some benchmarks and it seems
> that GCC still is better at compiling Gambit-C code.
> 
> Cheers,
> -- 
> -alex
> http://www.artisancoder.com/
> <bench-r5rs.html>
> <bench-r6rs-fixflo-unsafe.html>
> <bench-r6rs-fixflo.html>
> <bench-r6rs.html>
Alex Queiroz | 3 Aug 2012 16:44
Picon
Gravatar

Re: Running the benchmarks

Hallo,

On Fri, Aug 3, 2012 at 4:35 PM, Marc Feeley <feeley <at> iro.umontreal.ca> wrote:
> Could you report your findings?
>

I tested GCC 4.7.1 from MacPorts and Clang 4.0 from Xcode 4.4. This
machine is a Macbook Pro running OS X 10.8, has a Corei7 at 2GHz with
8GB of RAM and a SSD. I used the following options for both compilers:

-Wno-unused -Wno-write-strings -O3 -fno-math-errno -fschedule-insns2
-fno-trapping-math -fno-strict-aliasing -fwrapv -fmodulo-sched
-freschedule-modulo-scheduled-loops -march=core2 -mssse3
-fomit-frame-pointer -fPIC -fno-common -mieee-fp

Clang complained about '-fmodulo-sched' and
'-freschedule-modulo-scheduled-loops' but proceeded anyway. GCC built
Gambit-C at around 6m and Clang took 14m. GCC also built the
benchmarks noticeably faster. It seems that Clang uses a little less
memory now, because I also tested building Gambit-C with 'make -j 4',
and the system did not start swapping. The benchmarks timings were
annexed to my previous email.

Cheers,
--

-- 
-alex
http://www.artisancoder.com/
Bradley Lucier | 3 Aug 2012 21:04
Picon
Favicon

Re: Running the benchmarks

On Fri, 2012-08-03 at 08:06 +0200, Marc Feeley wrote:
The benchmarks are in misc/bench.tgz . After untaring the file, you will have the benchmark sources and some scripts to run the benchmarks and generate HTML tables of the results.

I'm curious, does Clang+LLVM use computed gotos like gcc does?

Brad
_______________________________________________
Gambit-list mailing list
Gambit-list <at> iro.umontreal.ca
https://webmail.iro.umontreal.ca/mailman/listinfo/gambit-list
Alex Queiroz | 3 Aug 2012 21:31
Picon
Gravatar

Re: Running the benchmarks

Hallo,

>
> I'm curious, does Clang+LLVM use computed gotos like gcc does?
>

Yes. Clang tries to support all GCC extensions.

--

-- 
-alex
http://www.artisancoder.com/

Program        Gambit-C-Clang-r5rs Gambit-C-r5rs