-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathINSTALL
79 lines (61 loc) · 2.85 KB
/
INSTALL
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
This OpenCL package requires at least one OpenCL implementation to be
installed in order for OpenCL to work.
OpenCL is a royalty-free standard (http://www.khronos.org/opencl/) and
there are several implementations based on this standard. Each
implementation is usually specific to a brand of a computing device
(such as GPU, CPU or accelerator).
*** OpenCL IMPLEMENTATIONS ***
Most common implementations are (in no particular order):
Type: GPU
Vendor: NVIDIA
URL: http://developer.nvidia.com/opencl
Mostly supporting NVIDIA GPUs for computation. The OpenCL package has
been successfully tested with "NVIDIA GPU computing SDK" on Linux and
depending on the GPU device both single and double-precision is
supported.
The libraries can be either installed from NVIDIA directly or on very
recent Debian and Ubuntu (from the non-free section) via
apt-get install nvidia-opencl-dev
Type: GPU, CPU
Vendor: Apple
This OpenCL implementation is part of Mac OS X 10.6 (Snow Leopard) and
Max OS X 10.7 (Lion) operating systems. The OpenCL package has been
successfully tested with nVidia GPU devices (only single-precision is
supported on the devices available to us).
The OpenCL framework is part of the OS so no additional libraries need
to be installed.
Type: GPU, CPU, ACC
Vendor: ATI / AMD
URL: http://developer.amd.com/zones/OpenCLZone/Pages/default.aspx
Mostly supporting AMD (formerly ATI) GPUs and AMD CPUs for computation.
Since I don't own any AMD GPUs or CPUs I could not test this
implementation. However, Kate Cowles reported success with AMD APP SDK
on Ubuntu.
Type: CPU, ACC
Vendor: Intel
URL: http://www.intel.com/go/opencl/
Supports Intel CPUs. Given the ubiquity of Intel CPUs this is a good
implementation to install for testing. The OpenCL has been
successfully tested with this implementation and RForge.net uses this
configuration for checks of the package (both single and double
precision is supported).
For Debian the installation is fairly simple, download from
http://software.intel.com/en-us/articles/download-intel-opencl-sdk/
apt-get install libnuma1
fakeroot alien intel_ocl_sdk_1.1_lnx_64.rpm
dpkg -i intel-ocl_1.1-13786_amd64.deb
(alternatively rpm2cpio can be used to unpack the rpm directly)
*** INSTALLATION ***
Once you have an OpenCL implementation installed you can install the
OpenCL R package using the regular install.packages() or R CMD INSTALL
methods.
The OpenCL package assumes the availability of OpenCL headers
<CL/opencl.h> and OpenCL library at compile time. If you installed the
OpenCL implementation in a non-standard location, you may have to
set the PKG_CFLAGS and/or PKG_LIBS environment variables accordingly
(and possibly LD_LIBRARY_PATH on some unix systems), see R
documentation on this.
---
DISCLAIMER: Product or corporate names may be trademarks or registered
trademarks, and are used only for identification and explanation
wihtout intent to infringe.