☰
Publisher's description
HPC Challenge is a series of high-performance reference. HPC Challenge is a series of high-performance reference. HPC Challenge is basically 7 benchmarks: 1. HPL - the Linpack TPP benchmark which measures the rate of floating point performance to solve a linear system of equations. 2. DGEMM - measures the rate of implementation in floating point double precision real matrix matrix multiplication. 3. STREAM - a simple synthetic benchmark program that measures sustainable memory bandwidth (in GB / s) and the corresponding rate for the kernel of a simple vector calculation. PTrans 4. (Parallel matrix transpose) - exercises the communications where pairs of processors communicate with each other simultaneously. It is a useful test of the capacity of network communications. 5. RandomAccess - measures the rate of updates over random memory (GUP). 6. FFTE - measures the rate of implementation in floating point 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 models, based on b_eff (effective reference bandwidth). Compilation: The first step is to create a configuration file that reflects the characteristics of your machine. The configuration file must be created in the directory of HPL. This directory contains instructions (README and INSTALL files) on how to create the configuration file. The HPL directory / facility contains numerous examples of configuration files. A good approach is to copy one of them to the directory of HPL and if this does not change it. This file is reused by all components of the HPC Challange result. Once the setup is complete, a file should exist in the directory HPL whose name begins with Make. and ends with the name for the system used for testing. For example, if the name of the system is Unix, the file must be called Make. Unix. To build the executable file reference (for the system under the name Unix) type: make arch = Unix. This command should be launched in the root directory (not in the directory HPL). It will search the directory for the HPL configuration file and use it to build the benchmark executable. Setting: The Challange HPC is driven by an input file called short hpccinf.txt which is almost the same as the input file for HPL (usually called HPL.dat). Refer to HPL file / www / tuning.html for more details on the input file for HPL. A sample input file is included with the HPC Challange. The differences between HPL input file and HPC Challange input file can be summarized as follows: * Lines 3 and 4 are ignored. The production is still in the file named hpccoutf.txt .* There are additional lines (from line 33) that may (but need not) be used to customize the baseline HPC Challenge. They are described below. The additional lines in the HPC Challenge input file (from the input file HPL) are: lines 33 and 34 describe additional matrix sizes to be used to establish the benchmark PTrans (one of components of the benchmark Challange HPC) .* Lines 35 and 36 describe additional factors blocking to be used to launch PTrans reference. Just for completeness, here is the list of lines in the input file of HPC Challange with a brief description 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 ordering process for HPL * Line 10: number of process grids HPL (and PTrans) * Line 11: number of rows to process each process grid for HPL (and PTrans) * Line 12: number of columns to process each process grid for HPL (and PTrans) * Line 13: the threshold value not to be exceeded by the residual scale for HPL (and PTrans) * Line 14: number of methods factorization 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 digits of recursive panel for HPL * Line 19: the number of recursive panel for HPL * Line 20: number of methods of recursive panel factorization for HPL * Line 21: recursive factorization methods for HPL * Line 22: number of delivery methods for HPL * Line 23 : Available methods for HPL * Line 24: Number of look-ahead depths HPL * Line 25: look-ahead depths HPL * Line 26: Clearing HPL * Line 27: exchange threshold for HPL * Line 28: Form L1 for HPL * Line 29: U-shaped for HPL * Line 30: A value that specifies whether the balance should be used by HPL * Line 31: alignment of memory for HPL * Line 32: ignored * Line Number 33: sizes additional problem for PTrans * Line 34: additional problem sizes PTrans * Line 35: number of other inhibiting factors for PTrans * Line 36: Other factors hampering a PTRANSWhat New in this version: This version · contains many fixes, major features and minor enhancements, many of which were provided by users. · The main objective of this release was to improve the accuracy of the state performance results and ensure scalability of the code on the largest supercomputer installations with hundreds of thousands of computing cores.