HPC Challenge 1.0.0 review
DownloadThe HPC Challenge benchmark consists of basically 7 benchmarks: 1
|
|
The HPC Challenge benchmark consists of basically 7 benchmarks:
1. HPL - the Linpack TPP benchmark which measures the floating point rate of execution for solving a linear system of equations.
2. DGEMM - measures the floating point rate of execution of double precision real matrix-matrix multiplication.
3. STREAM - a simple synthetic benchmark program that measures sustainable memory bandwidth (in GB/s) and the corresponding computation rate for simple vector kernel.
4. PTRANS (parallel matrix transpose) - exercises the communications where pairs of processors communicate with each other simultaneously. It is a useful test of the total communications capacity of the network.
5. RandomAccess - measures the rate of integer random updates of memory (GUPS).
6. FFTE - measures the floating point rate of execution of double precision complex one-dimensional Discrete Fourier Transform (DFT).
7. Communication bandwidth and latency - a set of tests to measure latency and bandwidth of a number of simultaneous communication patterns; based on b_eff (effective bandwidth benchmark).
Compiling:
The first step is to create a configuration file that reflects characteristics of your machine. The configuration file should be created in the hpl directory. This directory contains instructions (the files README and INSTALL) on how to create the configuration file. The directory hpl/setup contains many examples of configuration files. A good approach is to copy one of them to the hpl directory and if it doesn't work then change it. This file is reused by all the components of the HPC Challange suite.
When configuration is done, a file should exist in the hpl directory whose name starts with Make. and ends with the name for the system used for tests. For example, if the name of the system is Unix, the file should be named Make.Unix.
To build the benchmark executable (for the system named Unix) type: make arch=Unix. This command should be run in the top directory (not in the hpl directory). It will look in the hpl directory for the configuration file and use it to build the benchmark executable.
Configuration:
The HPC Challange is driven by a short input file named hpccinf.txt that is almost the same as the input file for HPL (customarily called HPL.dat). Refer to the file hpl/www/tuning.html for details about the input file for HPL. A sample input file is included with the HPC Challange distribution.
The differences between HPL input file and HPC Challange input file can be summarized as follows:
* Lines 3 and 4 are ignored. The output always goes to the file named hpccoutf.txt.
* There are additional lines (starting with line 33) that may (but do not have to) be used to customize the HPC Challenge benchmark. They are described below.
The additional lines in the HPC Challenge input file (compared to the HPL input file) are:
Lines 33 and 34 describe additional matrix sizes to be used for running the PTRANS benchmark (one of the components of the HPC Challange benchmark).
* Lines 35 and 36 describe additional blocking factors to be used for running PTRANS benchmark.
Just for completeness, here is the list of lines of the HPC Challange's input file with brief descriptions of their meaning:
* Line 1: ignored
* Line 2: ignored
* Line 3: ignored
* Line 4: ignored
* Line 5: number of matrix sizes for HPL (and PTRANS)
* Line 6: matrix sizes for HPL (and PTRANS)
* Line 7: number of blocking factors for HPL (and PTRANS)
* Line 8: blocking factors for HPL (and PTRANS)
* Line 9: type of process ordering for HPL
* Line 10: number of process grids for HPL (and PTRANS)
* Line 11: numbers of process rows of each process grid for HPL (and
PTRANS)
* Line 12: numbers of process columns of each process grid for HPL
(and PTRANS)
* Line 13: threshold value not to be exceeded by scaled residual for
HPL (and PTRANS)
* Line 14: number of panel factorization methods for HPL
* Line 15: panel factorization methods for HPL
* Line 16: number of recursive stopping criteria for HPL
* Line 17: recursive stopping criteria for HPL
* Line 18: number of recursion panel counts for HPL
* Line 19: recursion panel counts for HPL
* Line 20: number of recursive panel factorization methods for HPL
* Line 21: recursive panel factorization methods for HPL
* Line 22: number of broadcast methods for HPL
* Line 23: broadcast methods for HPL
* Line 24: number of look-ahead depths for HPL
* Line 25: look-ahead depths for HPL
* Line 26: swap methods for HPL
* Line 27: swapping threshold for HPL
* Line 28: form of L1 for HPL
* Line 29: form of U for HPL
* Line 30: value that specifies whether equilibration should be used
by HPL
* Line 31: memory alignment for HPL
* Line 32: ignored
* Line 33: number of additional problem sizes for PTRANS
* Line 34: additional problem sizes for PTRANS
* Line 35: number of additional blocking factors for PTRANS
* Line 36: additional blocking factors for PTRANS
HPC Challenge 1.0.0 keywords