All Downloads are FREE. Search and download functionalities are using the official Maven repository.

cern.colt.matrix.doc-files.allColt1.0.1ibm1.3LxPIII.txt Maven / Gradle / Ivy

Go to download

Parallel Colt is a multithreaded version of Colt - a library for high performance scientific computing in Java. It contains efficient algorithms for data analysis, linear algebra, multi-dimensional arrays, Fourier transforms, statistics and histogramming.

The newest version!
Colt Matrix benchmark running on

java.vm.vendor  IBM Corporation    
java.vm.version 1.3.0              
java.vm.name    Classic VM         
os.name         Linux              
os.version      2.2.12-20smp       
os.arch         x86                
java.version    1.3.0              
java.vendor     IBM Corporation    
java.vendor.url http://www.ibm.com/

Colt Version is Version 1.0.1.51 (Mon May 15 20:23:08 CEST 2000)
Please report problems to [email protected] or http://nicewww.cern.ch/~hoschek/colt/index.htm



Executing command = [dgemm, dense, 2, 2, 0.999, false, true, 5, 10, 25, 50, 100, 250, 500, 1000] ...

@x.x.x.x.x.x.x.x.*
Performance of Blas matrix-matrix mult dgemm(false, true, 1, A, B, 0, C)
type=dense
        | size
        | 5      10     25      50      100     250     500     1000   
-----------------------------------------------------------------------
d 0.999 | 36.528 82.766 112.261 157.978 198.849 212.999 204.165 175.932
Run took a total of Time=37.602 secs. End of run.


Executing command = [dgemm, dense, 1, 2, 0.999, false, true, 5, 10, 25, 50, 100, 250, 500, 1000] ...

@x.x.x.x.x.x.x.x.*
Performance of Blas matrix-matrix mult dgemm(false, true, 1, A, B, 0, C)
type=dense
        | size
        | 5      10     25      50      100     250     500     1000  
----------------------------------------------------------------------
d 0.999 | 36.213 74.518 102.758 105.095 119.012 116.333 102.124 87.409
Run took a total of Time=49.605 secs. End of run.


Executing command = [dgemm, rowCompressed, 1, 2, 0.01, false, false, 5, 10, 25, 50, 100, 250, 500, 1000] ...

@x.x.x.x.x.x.x.x.*
Performance of Blas matrix-matrix mult dgemm(false, false, 1, A, B, 0, C)
type=rowCompressed
       | size
       | 5      10      25      50         100       250        500        1000      
-------------------------------------------------------------------------------------
d 0.01 | 32.525 138.719 703.833 1.845E+003 3.27E+003 2.848E+003 2.772E+003 2.951E+003
Run took a total of Time=32.202 secs. End of run.


Executing command = [dgemm, sparse, 1, 2, 0.01, false, false, 5, 10, 25, 50, 100, 250, 500, 1000] ...

@x.x.x.x.x.x.x.x.*
Performance of Blas matrix-matrix mult dgemm(false, false, 1, A, B, 0, C)
type=sparse
       | size
       | 5      10      25      50         100       250        500        1000      
-------------------------------------------------------------------------------------
d 0.01 | 31.128 135.907 637.538 1.767E+003 3.15E+003 2.751E+003 2.817E+003 2.883E+003
Run took a total of Time=32.277 secs. End of run.


Executing command = [dgemv, dense, 2, 2, 0.01, false, 5, 10, 25, 50, 100, 250, 500, 1000, 2000] ...

@x.x.x.x.x.x.x.x.x.*
Performance of Blas matrix-vector mult dgemv(false, 1, A, B, 0, C)
type=dense
       | size
       | 5      10     25      50      100     250     500    1000   2000  
---------------------------------------------------------------------------
d 0.01 | 20.828 52.811 103.611 104.354 109.498 101.451 67.187 66.341 54.554
Run took a total of Time=39.415 secs. End of run.


Executing command = [dgemv, dense, 1, 2, 0.01, false, 5, 10, 25, 50, 100, 250, 500, 1000, 2000] ...

@x.x.x.x.x.x.x.x.x.*
Performance of Blas matrix-vector mult dgemv(false, 1, A, B, 0, C)
type=dense
       | size
       | 5      10    25      50      100     250    500    1000   2000  
-------------------------------------------------------------------------
d 0.01 | 30.574 65.79 114.877 101.596 111.633 47.494 31.702 31.778 31.793
Run took a total of Time=40.283 secs. End of run.


Executing command = [dgemv, rowCompressed, 1, 2, 0.01, false, 5, 10, 25, 50, 100, 250, 500, 1000, 2000] ...

@x.x.x.x.x.x.x.x.x.*
Performance of Blas matrix-vector mult dgemv(false, 1, A, B, 0, C)
type=rowCompressed
       | size
       | 5      10      25      50         100       250        500        1000       2000      
------------------------------------------------------------------------------------------------
d 0.01 | 31.796 145.363 527.966 1.143E+003 2.33E+003 3.737E+003 4.776E+003 5.992E+003 3.448E+003
Run took a total of Time=41.138 secs. End of run.


Executing command = [dgemv, sparse, 1, 2, 0.01, false, 5, 10, 25, 50, 100, 250, 500, 1000, 2000] ...

@x.x.x.x.x.x.x.x.x.*
Performance of Blas matrix-vector mult dgemv(false, 1, A, B, 0, C)
type=sparse
       | size
       | 5      10     25      50      100     250     500     1000    2000   
------------------------------------------------------------------------------
d 0.01 | 23.507 84.216 341.435 598.555 717.508 733.817 768.899 756.811 573.721
Run took a total of Time=38.784 secs. End of run.


Executing command = [assign, dense, 1, 2, 0.999, 5, 10, 25, 50, 100, 250, 500, 1000] ...

@x.x.x.x.x.x.x.x.*
Performance of A.assign(B) [Mops/sec]
type=dense
        | size
        | 5      10      25      50    100    250    500    1000 
-----------------------------------------------------------------
d 0.999 | 79.408 164.319 309.899 35.58 44.551 20.849 21.326 21.78
Run took a total of Time=36.992 secs. End of run.


Executing command = [assignGetSet, dense, 1, 2, 0.999, 5, 10, 25, 50, 100, 250, 500, 1000] ...

@x.x.x.x.x.x.x.x.*
Performance of A.assign(B) via get and set [Mops/sec]
type=dense
        | size
        | 5     10    25    50    100   250  500   1000 
--------------------------------------------------------
d 0.999 | 8.567 7.898 8.028 7.567 7.823 5.57 5.289 5.275
Run took a total of Time=32.556 secs. End of run.


Executing command = [assignGetSetQuick, dense, 1, 2, 0.999, 5, 10, 25, 50, 100, 250, 500, 1000] ...

@x.x.x.x.x.x.x.x.*
Performance of A.assign(B) via getQuick and setQuick [Mops/sec]
type=dense
        | size
        | 5      10     25    50    100    250   500   1000 
------------------------------------------------------------
d 0.999 | 11.381 11.172 11.51 10.19 10.943 7.344 7.516 7.495
Run took a total of Time=32.857 secs. End of run.


Executing command = [elementwiseMult, dense, 1, 2, 0.999, 5, 10, 25, 50, 100, 250, 500, 1000] ...

@x.x.x.x.x.x.x.x.*
Performance of A.assign(F.mult(0.5)) via Blas [Mflops/sec]
type=dense
        | size
        | 5      10     25     50     100    250    500    1000 
----------------------------------------------------------------
d 0.999 | 34.886 64.726 94.636 43.634 52.731 25.993 22.014 22.75
Run took a total of Time=37.829 secs. End of run.


Executing command = [elementwiseMultB, dense, 1, 2, 0.999, 5, 10, 25, 50, 100, 250, 500, 1000] ...

@x.x.x.x.x.x.x.x.*
Performance of A.assign(B,F.mult) via Blas [Mflops/sec]
type=dense
        | size
        | 5      10     25     50     100    250    500    1000  
-----------------------------------------------------------------
d 0.999 | 44.274 55.451 60.475 35.507 26.752 14.646 14.925 14.599
Run took a total of Time=34.514 secs. End of run.


Executing command = [assignLog, dense, 1, 2, 0.999, 5, 10, 25, 50, 100, 250, 500, 1000] ...

@x.x.x.x.x.x.x.x.*
Performance of A[i,j] = log(A[i,j]) via Blas.assign(fun) [Mflops/sec]
type=dense
        | size
        | 5     10    25    50    100   250   500   1000 
---------------------------------------------------------
d 0.999 | 4.697 4.703 4.777 4.694 4.804 4.333 4.171 4.204
Run took a total of Time=33.089 secs. End of run.


Executing command = [assignLog, dense, 2, 2, 0.999, 5, 10, 25, 50, 100, 250, 500, 1000] ...

@x.x.x.x.x.x.x.x.*
Performance of A[i,j] = log(A[i,j]) via Blas.assign(fun) [Mflops/sec]
type=dense
        | size
        | 5     10    25    50    100   250   500   1000 
---------------------------------------------------------
d 0.999 | 4.351 4.727 4.881 4.775 4.894 7.295 6.991 7.246
Run took a total of Time=32.783 secs. End of run.


Executing command = [assignPlusMult, dense, 1, 2, 0.999, 5, 10, 25, 50, 100, 250, 500, 1000] ...

@x.x.x.x.x.x.x.x.*
Performance of A[i,j] = A[i,j] + s*B[i,j] via Blas.assign(fun) [Mflops/sec]
type=dense
        | size
        | 5      10     25     50     100    250    500    1000  
-----------------------------------------------------------------
d 0.999 | 52.745 79.516 94.163 74.865 44.036 31.974 29.661 31.826
Run took a total of Time=34.536 secs. End of run.


Executing command = [SOR5, dense, 1, 2, 0.999, 5, 10, 25, 50, 100, 250, 500, 1000] ...

@x.x.x.x.x.x.x.x.*
Performance of A.zAssign8Neighbors(5 point function) [Mflops/sec]
type=dense
        | size
        | 5       10     25     50     100   250   500    1000 
---------------------------------------------------------------
d 0.999 | 117.043 74.677 59.492 55.529 53.61 38.48 39.857 37.43
Run took a total of Time=32.801 secs. End of run.


Executing command = [SOR8, dense, 1, 2, 0.999, 5, 10, 25, 50, 100, 250, 500, 1000] ...

@x.x.x.x.x.x.x.x.*
Performance of A.zAssign8Neighbors(9 point function) [Mflops/sec]
type=dense
        | size
        | 5       10     25     50     100    250    500    1000  
------------------------------------------------------------------
d 0.999 | 128.176 82.446 64.465 55.579 55.606 42.713 43.439 42.849
Run took a total of Time=32.433 secs. End of run.


Executing command = [LUDecompose, dense, 1, 2, 0.999, 5, 10, 25, 50, 100, 250, 500, 1000] ...

@x.x.x.x.x.x.x.x.*
Performance of LU.decompose(A) [Mflops/sec]
type=dense
        | size
        | 5     10     25     50     100    250    500    1000  
----------------------------------------------------------------
d 0.999 | 3.482 10.428 27.323 51.485 67.178 88.586 35.401 30.844
Run took a total of Time=47.515 secs. End of run.


Executing command = [LUSolve, dense, 1, 2, 0.999, 5, 10, 25, 50, 100, 250, 500, 1000] ...

@x.x.x.x.x.x.x.x.*
Performance of LU.solve(A) [Mflops/sec]
type=dense
        | size
        | 5      10     25    50     100    250    500    1000  
----------------------------------------------------------------
d 0.999 | 12.488 30.537 66.43 78.551 73.841 56.787 26.843 25.915
Run took a total of Time=134.412 secs. End of run.


Executing command = [pow, dense, 1, 2, 0.999, -1, 5, 10, 25, 50, 100, 250, 500, 1000] ...

@x.x.x.x.x.x.x.x.*
Performance of matrix to the power of an exponent pow(A,-1)
type=dense
        | size
        | 5     10     25     50     100    250    500    1000  
----------------------------------------------------------------
d 0.999 | 5.407 15.041 41.556 60.984 70.523 63.082 30.962 28.746
Run took a total of Time=127.618 secs. End of run.


Command file name used: /afs/cern.ch/user/h/hoschek/coltb4/cmd2
To reproduce and compare results, here it's contents:
// matrix-matrix mult with 1 and with 2 CPUs
dgemm dense 2 2.0 0.999 false true 5 10 25 50 100 250 500 1000
dgemm dense 1 2.0 0.999 false true 5 10 25 50 100 250 500 1000
dgemm rowCompressed 1 2.0 0.01 false false 5 10 25 50 100 250 500 1000
dgemm sparse 1 2.0 0.01 false false 5 10 25 50 100 250 500 1000

// matrix-vector mult
dgemv dense 2 2.0 0.01 false 5 10 25 50 100 250 500 1000 2000
dgemv dense 1 2.0 0.01 false 5 10 25 50 100 250 500 1000 2000
dgemv rowCompressed 1 2.0 0.01 false 5 10 25 50 100 250 500 1000 2000
dgemv sparse 1 2.0 0.01 false 5 10 25 50 100 250 500 1000 2000

assign dense 1 2.0 0.999 5 10 25 50 100 250 500 1000
assignGetSet dense 1 2.0 0.999 5 10 25 50 100 250 500 1000
assignGetSetQuick dense 1 2.0 0.999 5 10 25 50 100 250 500 1000

// with 1 and with 2 CPUs
elementwiseMult dense 1 2.0 0.999 5 10 25 50 100 250 500 1000
elementwiseMultB dense 1 2.0 0.999 5 10 25 50 100 250 500 1000
assignLog dense 1 2.0 0.999 5 10 25 50 100 250 500 1000
assignLog dense 2 2.0 0.999 5 10 25 50 100 250 500 1000
assignPlusMult dense 1 2.0 0.999 5 10 25 50 100 250 500 1000
assignPlusMult dense 2 2.0 0.999 5 10 25 50 100 250 500 1000

SOR5 dense 1 2.0 0.999 5 10 25 50 100 250 500 1000
SOR8 dense 1 2.0 0.999 5 10 25 50 100 250 500 1000

LUDecompose dense 1 2.0 0.999 5 10 25 50 100 250 500 1000
LUSolve dense 1 2.0 0.999 5 10 25 50 100 250 500 1000
pow dense 1 2.0 0.999 -1 5 10 25 50 100 250 500 1000 




© 2015 - 2025 Weber Informatics LLC | Privacy Policy