Performance information for the MASS libraries for Blue Gene/L and Blue Gene/P

Product documentation


Abstract

This document provides performance data for the MASS scalar and vector libraries for Blue Gene/L and Blue Gene/P, and comparison with the standard system math library.

Content

The following table provides approximate performance data for the MASS libraries running on Blue Gene/P. (The results are also representative of Blue Gene/L MASS.) The data was obtained by timing many repetitions of a length-1000 loop (for scalar functions) or calls for length-1000 vectors (for vector functions) over 1000 random vectors and includes all overheads. Timing in this way brings the input and output vectors into the on-chip cache (because the loop is short enough for the vectors to fit in the cache). Performance may deteriorate significantly when the input and output vectors are not in the cache. Performance may also deteriorate for arguments at or near the end-points of the valid argument ranges. The system library measurements were made with the versions of the library available on the test system. They may vary from the versions timed for previous versions of MASS. Users may experience performance that differs from that found in this table.


The column labelled libm give the results from using the functions in the MASS C/C++ source code library libmassv.c to call the functions in the system math library. The C source code was compiled with the IBM Visual Age C/C++ compiler using the -O option. The columns labelled libmass and libmassv obtain the results using the libmass.a and libmassv.a library respectively. The columns labelled libmass speedup and libmassv speedup give the ratio of the corresponding entries from libm and libmass, libmassv respectively. Vector function names are used to indicate the corresponding scalar functions in libm and libmass. A blank entry in any column indicates that the corresponding function is not available.

The range key at the end of this section describes the ranges from which random arguments were uniformly generated.

MASS Performance (cycles per call, length 1000 vector/loop)

          R

           a
           n
           g  libm   libmass   libmass  libmassv  libmassv
function   e                   speedup            speedup
----------------------------------------------------------
vacos      B 205.66   172.18     1.19    41.73     4.93
vacosh     G 477.03   203.22     2.35    45.62    10.46
vasin      B 208.98   170.89     1.22    41.64     5.02
vasinh     D 476.88   187.64     2.54    43.42    10.98
vatan      B 207.60   120.11     1.73
           D 231.65   147.85     1.57
vatan2     D 418.40   171.02     2.45    48.45     8.64
vatanh     B 356.64   169.98     2.10    33.42    10.67
vcbrt      D 267.13   125.91     2.12    17.88    14.94
vcopysign  D  43.41    48.03     0.90
vcos       B 152.31    71.03     2.14    29.24     5.21
           D 246.61   101.43     2.43    29.24     8.43
vcosh      D 261.29    96.04     2.72    18.66    14.00
vcosisin   B 325.94   122.08     2.67    63.10     5.17
           D 519.34   164.14     3.16    69.08     7.52
vdint      D                              8.51
vdiv       D                              3.36
vdnint     D  64.97    73.06     0.89     7.68     8.46
verf       C  98.23    65.64     1.50
verfc      C 192.62   137.89     1.40
vexp       D 200.50    90.03     2.23    14.77    13.57
vexpm1     D 192.66    95.14     2.03    16.10    11.97
vhypot     D 283.92   108.07     2.63
vlgamma    H 493.77   280.90     1.76
vlog       C 329.11   120.14     2.74    16.04    20.52
vlog10     C 391.41   114.08     3.43    15.38    25.45
vlog1p     H 209.71   115.99     1.81    22.92     9.15
vpow       C 651.42   198.01     3.29    67.96     9.59
vqdrt      C                             22.92
vrcbrt     D                             17.08
vrec       D                              2.51
vrqdrt     C                             23.09
vrsqrt     C 214.18    60.11     3.56     5.08    42.16
vsacos     B 240.16   108.94     2.20    30.00     8.01
vsacosh    G 337.93   156.26     2.16    32.77    10.31
vsasin     B 220.88   109.52     2.02    29.38     7.52
vsasinh    D 313.06   164.63     1.90    35.16     8.90
vsatan     B 144.32    84.12     1.72
           D 151.50   129.57     1.17
vsatan2    D 285.82   145.80     1.96    57.15     5.00
vsatanh    B 311.06   145.14     2.14    30.63    10.16
vscbrt     D 250.94   101.96     2.46    14.05    17.86
vscopysign D  43.44    46.24     0.94
vscos      B 129.74    61.18     2.12    24.53     5.29
           D 240.74    96.08     2.51    24.38     9.87
vscosh     D 294.08    84.22     3.49    28.47    10.33
vscosisin  B                             60.42
           D                             67.09
vsdiv      D                              9.86
vserf      C  99.38    54.13     1.84
vserfc     C 183.33    70.14     2.61
vsexp      D 242.95    83.28     2.92    14.05    17.29
vsexpm1    D 181.54    77.62     2.34    13.86    13.10
vshypot    D 241.18   109.94     2.19
vsin       B 173.32    66.23     2.62    29.21     5.93
           D 273.44   104.90     2.61    29.21     9.36
vsincos    B 317.60    94.21     3.37    50.31     6.31
           D 511.76   135.25     3.78    50.32    10.17
vsinh      D 264.30   101.14     2.61    27.18     9.72
vslgamma   H 312.57   270.11     1.16
vslog      C 200.60    88.05     2.28    13.82    14.52
vslog10    C 248.32    88.04     2.82    11.66    21.30
vslog1p    H 200.78    97.76     2.05    16.84    11.92
vspow      C 664.22   126.30     5.26    38.31    17.34
vsqdrt     C                             15.34
vsqrt      C 182.18    64.94     2.81     8.49    21.46
vsrcbrt    D                             14.22
vsrec      D                              4.57
vsrint     D  51.28    52.11     0.98
vsrqdrt    C                             15.29
vsrsqrt    C                              5.75
vssin      B 112.71    69.09     1.63    24.11     4.67
           D 238.69    95.39     2.50    24.10     9.90
vssincos   B                             44.46
           D                             44.63
vssinh     D 285.09    85.14     3.35    25.31    11.26
vssqrt     C                              7.01
vstan      D 292.65   129.82     2.25    31.25     9.36
vstanh     F 329.74    79.97     4.12    27.65    11.93
vtan       D 369.43   114.11     3.24    43.64     8.47
vtanh      F 347.13   110.31     3.15    24.53    14.15

Range Key

 A      0,1
 B     -1,1
 C      0,100
 D   -100,100
 E    -10,10
 F    -20,20
 G      1,100
 H     -1,101
 I      0,10


Original publication date

2005/4/7

Rate this page:

(0 users)Average rating

Add comments

Document information


More support for:

Mathematical Acceleration Subsystem
Libraries

Software version:

4.4

Operating system(s):

Linux

Reference #:

7006977

Modified date:

2009-08-14

Translate my page

Machine Translation

Content navigation