|
9 | 9 | | Windows | [](http://ci.arrayfire.org/view/All/job/arrayfire-wrappers/job/python-windows/) | |
10 | 10 | | OSX | [](http://ci.arrayfire.org/view/All/job/arrayfire-wrappers/job/python-osx/) | |
11 | 11 |
|
12 | | -## Example |
13 | | - |
14 | | -```python |
15 | | -import arrayfire as af |
16 | | - |
17 | | -# Display backend information |
18 | | -af.info() |
19 | | - |
20 | | -# Generate a uniform random array with a size of 5 elements |
21 | | -a = af.randu(5, 1) |
22 | | - |
23 | | -# Print a and its minimum value |
24 | | -af.display(a) |
25 | | - |
26 | | -# Print min and max values of a |
27 | | -print("Minimum, Maximum: ", af.min(a), af.max(a)) |
28 | | -``` |
29 | | - |
30 | | -## Sample outputs |
| 12 | +## Documentation |
31 | 13 |
|
32 | | -On an AMD GPU: |
| 14 | +Documentation for this project can be found [over here](http://arrayfire.org/arrayfire-python/). |
33 | 15 |
|
34 | | -``` |
35 | | -Using opencl backend |
36 | | -ArrayFire v3.0.1 (OpenCL, 64-bit Linux, build 17db1c9) |
37 | | -[0] AMD : Spectre |
38 | | --1- AMD : AMD A10-7850K Radeon R7, 12 Compute Cores 4C+8G |
39 | | - |
40 | | -[5 1 1 1] |
41 | | -0.4107 |
42 | | -0.8224 |
43 | | -0.9518 |
44 | | -0.1794 |
45 | | -0.4198 |
46 | | - |
47 | | -Minimum, Maximum: 0.17936542630195618 0.9517996311187744 |
48 | | -``` |
49 | | - |
50 | | -On an NVIDIA GPU: |
| 16 | +## Example |
51 | 17 |
|
| 18 | +```python |
| 19 | +# Monte Carlo estimation of pi |
| 20 | +def calc_pi_device(samples): |
| 21 | + # Simple, array based API |
| 22 | + # Generate uniformly distributed random numers |
| 23 | + x = af.randu(samples) |
| 24 | + y = af.randu(samples) |
| 25 | + # Supports Just In Time Compilation |
| 26 | + # The following line generates a single kernel |
| 27 | + within_unit_circle = (x * x + y * y) < 1 |
| 28 | + # Intuitive function names |
| 29 | + return 4 * af.count(within_unit_circle) / samples |
52 | 30 | ``` |
53 | | -Using cuda backend |
54 | | -ArrayFire v3.0.0 (CUDA, 64-bit Linux, build 86426db) |
55 | | -Platform: CUDA Toolkit 7, Driver: 346.46 |
56 | | -[0] Tesla K40c, 12288 MB, CUDA Compute 3.5 |
57 | | --1- GeForce GTX 750, 1024 MB, CUDA Compute 5.0 |
58 | | - |
59 | | -Generate a random matrix a: |
60 | | -[5 1 1 1] |
61 | | -0.7402 |
62 | | -0.9210 |
63 | | -0.0390 |
64 | | -0.9690 |
65 | | -0.9251 |
66 | | - |
67 | | -Minimum, Maximum: 0.039020489901304245 0.9689629077911377 |
68 | | -``` |
69 | | - |
70 | | -Fallback to CPU when CUDA and OpenCL are not availabe: |
71 | 31 |
|
72 | | -``` |
73 | | -Using cpu backend |
74 | | -ArrayFire v3.0.0 (CPU, 64-bit Linux, build 86426db) |
75 | | - |
76 | | -Generate a random matrix a: |
77 | | -[5 1 1 1] |
78 | | -0.0000 |
79 | | -0.1315 |
80 | | -0.7556 |
81 | | -0.4587 |
82 | | -0.5328 |
83 | | - |
84 | | -Minimum, Maximum: 7.825903594493866e-06 0.7556053400039673 |
85 | | -``` |
86 | 32 |
|
87 | 33 | Choosing a particular backend can be done using `af.backend.set( backend_name )` where backend_name can be one of: "_cuda_", "_opencl_", or "_cpu_". The default device is chosen in the same order of preference. |
88 | 34 |
|
|
0 commit comments