[R-br] Fwd: Re: benchmarking

Leonard de Assis assis.leonard em gmail.com
Sexta Outubro 28 16:37:15 BRST 2011



-------- Mensagem original --------
Assunto: 	Re: [R-br] benchmarking
Data: 	Fri, 28 Oct 2011 16:31:34 -0200
De: 	Leonard de Assis <assis.leonard em gmail.com>
Responder a: 	assis.leonard em gmail.com
Para: 	assis.leonard em gmail.com



Mesma máquina, R 2.13.2, Ubuntu 11.10 x64
ATLAS instalado (Com blas, provavelmente)


   R Benchmark 2.5
   ===============
Number of times each test is run__________________________:  3

   I. Matrix calculation
   ---------------------
Creation, transp., deformation of a 2500x2500 matrix (sec): 
0.679999999999997
2400x2400 normal distributed random matrix ^1000____ (sec): 
3.66666666666667
Sorting of 7,000,000 random values__________________ (sec): 
0.919999999999999
2800x2800 cross-product matrix (b = a' * a)_________ (sec):  1.77
Linear regr. over a 3000x3000 matrix (c = a \ b')___ (sec): 
1.05333333333333
                      --------------------------------------------
                 Trimmed geom. mean (2 extremes eliminated): 
1.19704085704037

   II. Matrix functions
   --------------------
FFT over 2,400,000 random values____________________ (sec): 
0.756666666666665
Eigenvalues of a 640x640 random matrix______________ (sec): 
4.31333333333333
Determinant of a 2500x2500 random matrix____________ (sec): 
1.05666666666667
Cholesky decomposition of a 3000x3000 matrix________ (sec): 
1.29333333333333
Inverse of a 1600x1600 random matrix________________ (sec): 
1.11666666666666
                      --------------------------------------------
                Trimmed geom. mean (2 extremes eliminated): 
1.15130575853477

   III. Programmation
   ------------------
3,500,000 Fibonacci numbers calculation (vector calc)(sec): 
5.47666666666667
Creation of a 3000x3000 Hilbert matrix (matrix calc) (sec): 
0.689999999999998
Grand common divisors of 400,000 pairs (recursion)__ (sec): 
1.08000000000001
Creation of a 500x500 Toeplitz matrix (loops)_______ (sec): 
1.15333333333333
Escoufier's method on a 45x45 matrix (mixed)________ (sec): 
0.840000000000003
                      --------------------------------------------
                Trimmed geom. mean (2 extremes eliminated): 
1.01520238306255


Total time for all 15 tests_________________________ (sec): 
25.8666666666667
Overall mean (sum of I, II and III trimmed means/3)_ (sec): 
1.11845219568472
                      --- End of test ---

[]s
Leonard de Assis
assis <dot> leonard <at> gmail <dot> com


Em 28/10/2011 14:49, Leonard de Assis escreveu:
> Segue meu teste na minha 1a config:
>
> SAMSUNG RF411
> Core i5 2a Geração 6 GB RAM
> Windows 7 Pro x64
> R 2.13.1 x64 Sem BLAS (Tem BLAS pra Intel64?)
>
>    R Benchmark 2.5
>    ===============
> Number of times each test is run__________________________:  3
>
>    I. Matrix calculation
>    ---------------------
> Creation, transp., deformation of a 2500x2500 matrix (sec): 
> 0.563333333333333
> 2400x2400 normal distributed random matrix ^1000____ (sec): 
> 3.33333333333333
> Sorting of 7,000,000 random values__________________ (sec): 
> 0.840000000000001
> 2800x2800 cross-product matrix (b = a' * a)_________ (sec): 
> 14.0966666666667
> Linear regr. over a 3000x3000 matrix (c = a \ b')___ (sec): 
> 6.73333333333333
>                       --------------------------------------------
>                  Trimmed geom. mean (2 extremes eliminated): 
> 2.66151784278081
>
>    II. Matrix functions
>    --------------------
> FFT over 2,400,000 random values____________________ (sec): 
> 0.486666666666669
> Eigenvalues of a 640x640 random matrix______________ (sec): 
> 0.946666666666658
> Determinant of a 2500x2500 random matrix____________ (sec): 
> 3.56000000000001
> Cholesky decomposition of a 3000x3000 matrix________ (sec): 
> 5.14666666666668
> Inverse of a 1600x1600 random matrix________________ (sec): 
> 3.53666666666667
>                       --------------------------------------------
>                 Trimmed geom. mean (2 extremes eliminated): 
> 2.28426807887244
>
>    III. Programmation
>    ------------------
> 3,500,000 Fibonacci numbers calculation (vector calc)(sec): 
> 4.11333333333333
> Creation of a 3000x3000 Hilbert matrix (matrix calc) (sec): 
> 0.506666666666661
> Grand common divisors of 400,000 pairs (recursion)__ (sec): 
> 0.770000000000001
> Creation of a 500x500 Toeplitz matrix (loops)_______ (sec): 
> 0.816666666666663
> Escoufier's method on a 45x45 matrix (mixed)________ (sec):  0.5
>                       --------------------------------------------
>                 Trimmed geom. mean (2 extremes eliminated): 
> 0.682997786213496
>
>
> Total time for all 15 tests_________________________ (sec):  45.95
> Overall mean (sum of I, II and III trimmed means/3)_ (sec): 
> 1.60730606307823
>                       --- End of test ---
>
> Warning messages:
> 1: package 'Matrix' was built under R version 2.13.2
> 2: package 'SuppDists' was built under R version 2.13.2
>
> []s
> Leonard de Assis
> assis <dot> leonard <at> gmail <dot> com
>
>
> Em 28/10/2011 11:33, Benilton Carvalho escreveu:
>> Aos interessados em fazer algum benchmarking, aqui uma sugestao:
>>
>> 1) Instale o pacote SuppDists
>> 2) Execute:
>>
>> source('http://r.research.att.com/benchmarks/R-benchmark-25.R')
>>
>> 3) Reporte aqui:
>>
>> a) Informacao de Hardware
>> b) Versao do R
>> c) Informacoes adicionais relevantes (BLAS / LAPACK / etc)
>> d) Resultados do benchmark...
>>
>> Para o meu caso:
>>
>> Mac Pro 4,1
>> Quad-Core Xeon 2.93GHz x 2 HT
>> 16G RAM
>> R 2.13.2 (binario padrao) + ATLAS BLAS
>>
>> Resultados:
>>
>>
>>    R Benchmark 2.5
>>    ===============
>> Number of times each test is run__________________________:  3
>>
>>    I. Matrix calculation
>>    ---------------------
>> Creation, transp., deformation of a 2500x2500 matrix (sec):  0.617666666666613
>> 2400x2400 normal distributed random matrix ^1000____ (sec):  0.267000000000053
>> Sorting of 7,000,000 random values__________________ (sec):  0.821333333333352
>> 2800x2800 cross-product matrix (b = a' * a)_________ (sec):  0.37166666666667
>> Linear regr. over a 3000x3000 matrix (c = a \ b')___ (sec):  0.316000000000031
>>                       --------------------------------------------
>>                  Trimmed geom. mean (2 extremes eliminated):  0.417059757745061
>>
>>    II. Matrix functions
>>    --------------------
>> FFT over 2,400,000 random values____________________ (sec):  0.489666666666684
>> Eigenvalues of a 640x640 random matrix______________ (sec):  0.65166666666668
>> Determinant of a 2500x2500 random matrix____________ (sec):  0.430999999999964
>> Cholesky decomposition of a 3000x3000 matrix________ (sec):  0.331666666666706
>> Inverse of a 1600x1600 random matrix________________ (sec):  0.471333333333329
>>                       --------------------------------------------
>>                 Trimmed geom. mean (2 extremes eliminated):  0.463342341088548
>>
>>    III. Programmation
>>    ------------------
>> 3,500,000 Fibonacci numbers calculation (vector calc)(sec):  0.721333333333367
>> Creation of a 3000x3000 Hilbert matrix (matrix calc) (sec):  0.472333333333268
>> Grand common divisors of 400,000 pairs (recursion)__ (sec):  2.30133333333337
>> Creation of a 500x500 Toeplitz matrix (loops)_______ (sec):  0.733999999999999
>> Escoufier's method on a 45x45 matrix (mixed)________ (sec):  0.533999999999992
>>                       --------------------------------------------
>>                 Trimmed geom. mean (2 extremes eliminated):  0.656333300289129
>>
>>
>> Total time for all 15 tests_________________________ (sec):  9.53200000000008
>> Overall mean (sum of I, II and III trimmed means/3)_ (sec):  0.502429238538774
>>                       --- End of test ---
>> _______________________________________________
>> R-br mailing list
>> R-br em listas.c3sl.ufpr.br
>> https://listas.inf.ufpr.br/cgi-bin/mailman/listinfo/r-br
>> Leia o guia de postagem (http://www.leg.ufpr.br/r-br-guia) e forneça código mínimo reproduzível.

-------------- Próxima Parte ----------
Um anexo em HTML foi limpo...
URL: <http://listas.inf.ufpr.br/pipermail/r-br/attachments/20111028/aaa9a906/attachment.html>


Mais detalhes sobre a lista de discussão R-br