pavanky

-DDEBUG
-I/home/pyalamanchili/Workspace/source/foo/bar/include/
-I/usr/include/c++/7.3.0/

pavanky

import arrayfire as af
import numpy as np
N = []; err_af = []; err_np = []
eps = 1E-16
for i in range(2, 64):
    in_array_np = np.random.rand(i)
    in_array_af = af.to_array(in_array_np)
    try:
        out_array_af_1 = af.ifft(af.fft(in_array_af), scale=1) / i
        out_array_af_2 = af.ifft(af.fft(in_array_af))

pavanky

```
/*******************************************************
 * Copyright (c) 2014, ArrayFire
 * All rights reserved.
 *
 * This file is distributed under 3-clause BSD license.
 * The complete license agreement can be obtained at:
 * http://arrayfire.com/licenses/BSD-3-Clause
 ********************************************************/

pavanky

Python 3.6.1 (default, Mar 27 2017, 00:27:06) 
[GCC 6.3.1 20170306] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import arrayfire as af
ERROR: GLFW wasn't able to initalize
>>> a = af.randu(5)
>>> af.print_mem_info()
Memory Info
---------------------------------------------------------
|     POINTER      |    SIZE    |  AF LOCK  | USER LOCK |

pavanky

2
single: 0.22531747817993164
100 af.time solves        [s]: 8.939504623413086e-05
4
single: 0.002744913101196289
100 af.time solves        [s]: 7.341623306274414e-05
6
single: 0.0005710124969482422
100 af.time solves        [s]: 5.37109375e-05
8

pavanky

$ ./test/binary_cuda 
Running main() from gtest_main.cc
[==========] Running 191 tests from 2 test cases.
[----------] Global test environment set-up.
[----------] 107 tests from BinaryTests
[ RUN      ] BinaryTests.Test_add_float_float
[       OK ] BinaryTests.Test_add_float_float (567 ms)
[ RUN      ] BinaryTests.Test_add_float_float_left
[       OK ] BinaryTests.Test_add_float_float_left (183 ms)
[ RUN      ] BinaryTests.Test_add_float_float_right

pavanky

>>> a = af.randu(5,5)
>>> af.print_mem_info()
Memory Info
---------------------------------------------------------
|     POINTER      |    SIZE    |  AF LOCK  | USER LOCK |
---------------------------------------------------------
|     0x706400000  |       1 KB |       Yes |        No |
---------------------------------------------------------
>>> b = af.randu(5,5)
>>> af.print_mem_info()

pavanky

local cudaAvailable, _ = pcall(require, 'cunn')

local function benchmark(opt)
   local isize = opt.inputSize or 100000
   local osize = opt.outputSize or 1
   local weightDecay = opt.weightDecay or 0
   local nnzMin = opt.featuresMinNumber or 1
   local nnzMax = opt.featuresMaxNumber or 10000
   local idxMin = 1
   local idxMax = isize

pavanky

#include <mutex>
#include <thread>
#include <cstdio>
#include <vector>
#include <iostream>

class Node
{
private:
    int m_a, m_b, m_c;

pavanky

Params [Size: 10000 x 1, Batch: 1, nnz: (25,50), features: 385]
Speedup (SL vs IL): GPU: 1.4440615290605
----------------------------------------------------------------------------------

Params [Size: 10000 x 1, Batch: 4, nnz: (25,50), features: 1457]
Speedup (SL vs IL): GPU: 1.9608264852205
----------------------------------------------------------------------------------

Params [Size: 10000 x 1, Batch: 32, nnz: (25,50), features: 12087]
Speedup (SL vs IL): GPU: 5.6983979608533