OpenDwarfs/README at master · aminems/OpenDwarfs · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
OpenDwarfs
=============
The OpenDwarfs project provides a benchmark suite consisting of different
computation/communication idioms, i.e., dwarfs, for state-of-art multicore and
GPU systems.  The first instantiation of the OpenDwarfs has been realized in
OpenCL, as described in part in "OpenCL and the 13 Dwarfs: A Work in Progress"
by Feng, Lin, Scogland, and Zhang in the 3rd ACM/SPEC International Conference
on Performance Engineering, April 2012.


The computation/communication idioms are based on the 13 Berkeley Dwarfs:
(http://view.eecs.berkeley.edu/wiki/Dwarf_Mine).

Benchmark status
----------------

Stable:
gem

Beta:
bfs
cfd
crc
fft
kmeans
lud
nw
spmv
srad
swat
bwa_hmm
nqueens

Alpha:
astar
tdm

Requirements
------------

Packages and libraries needed to build and run the applications.

To build:

    opencl >= 1.0 (some apps require 1.1, but we do not yet guarantee support for 1.2 in all applications.)
    autoconf >= 2.63
    autoheader
    automake
    libtool
    gcc
    maker

To run:

    opencl libs

Building
--------

To build all of the included applications:

    $ ./autogen.sh
    $ mkdir build
    $ cd build
    $ ../configure
    $ make

To build only the applications you select, call configure with the --with-apps
option:

    $ ../configure --with-apps=srad,gem,cfd

To see a full list of options and applications:

    $ ../configure --help

Running
-------

See the application-specific README file in each application's directory.
All the dwarf applications support a common list of options for optionally specifying the OpenCL platform ID (-p) and OpenCL device ID (-d), or alternatively, the device type (-t). These options, if supplied, must follow the executable name and be delimited from the application-specific options by double dashes (--).

General format: ./<executable> [-p <platform> -d <device> | -t <type> --] [app-specific options]

    <platform>	:integer ID of platform to use
    <device>    :integer ID of device in <platform> to use
    <type>	: device type to use (0:CPU, 1:GPU, 2:MIC, 3:FPGA)

Example1: ./astar -p 0 -d 0 -- (selects device with device ID 0 on platform with platform ID 0)
Example2: ./astar -t 0 -- (selects CPU device type on default platform with platform ID 0, if available)

Notes:	If no parameters are supplied, default platform ID is 0 and default device type is CPU.
	If -t parameter is given, default platform ID 0 is searched for supplied device type <type>. If not available, CPU device type selection will be attempted.
	If device ID is unknown, a combination of -p and -t is available to search for device of selected <type> on platform ID <platform>.

Acknowledgements
----------------

This project was supported in part by AMD, Air Force Research Lab, Department of
Defense, Harris, Los Alamos National Laboratory, and Xilinx via the NSF Center
for High-Performance Reconfigurable Computing (CHREC) under NSF grant
IIP-0804155 and indirectly by NSF grants CNS-0916719 and MRI-0960081.

Integration for Altera FPGA support for crc and csr, as well as extensions for these
benchmarks, have been contributed by Tyler Kenney at IBM.