Since there is no built-in random function in OpenCL (not that I know of, please correct me if this is not true). To generate a random list that put into kernel will not work for the purpose of my work. It has to be a random generator running on GPU (kernel). I intended to write my own function that generate random number in the range from 0 to 1. The code below is what i have run on CPU, and it seem to work well.
array_of_random_numbers = [0.0] * N
seed = 19073486328125
multiplier = 19073486328125
adder = 0
random_number = 19073486328125
modulus = 2**48.0
RAND_MAX = 2**48.0
for i in range( 0 , (N-1) ):
random_number = (multiplier * random_number + adder ) % (modulus)
print(multiplier * random_number)
array_of_random_numbers[i] = (random_number / RAND_MAX)
However, I have hard time migrate the code to kernel since I cannot set the random_number in a loop and let it change over iterations.
kernel = """__kernel void get_new_rand(__global float* c)
{
random_number = (19073486328125 * 19073486328125 + 0) % (281474976710656);
c[thread_id] = (random_number / 281474976710656.0);
}"""
Is there a way I can write the random generator on kernel?
Thank you in advance!
I intended to write my own function that generate random number in the range from 0 to 1. The code below is what i have run on CPU, and it seem to work well. However, I have hard time migrate the code to kernel since I cannot set the random_number in a loop and let it change over iterations.
The simplest approach is to use a linear congruence generator function (like you already have), that takes in a seed value and outputs a pseudo-random number that can be normalized to the range [0,1[. The remaining problem is how to get the seed.
Solution: you can pass a seed value as a global parameter from the host side, and change the seed value for every call of the kernel on the host side. This does not have any performance impact. Finally, to get different seeds for the different GPU threads, add the global ID to the seed passed over from the host before calling the LCG function.
This way you don't need any array of numbers stored. Also you have full control over the seed value on the host side, and it all remains fully deterministic.
Related
I'm intending to implement a random number generator via Swift 3. I have three different methods for generating an integer (between 0 and 50000) ten thousand times non-stop.
Do these generators use the same math principles of generating a value or not?
What generator is less CPU and RAM intensive at runtime (having 10000 iterations)?
method A:
var generator: Int = random() % 50000
method B:
let generator = Int(arc4random_uniform(50000))
method C:
import GameKit
let number: [Int] = [0, 1, 2... 50000]
func generator() -> Int {
let random = GKRandomSource.sharedRandom().nextIntWithUpperBound(number.count)
return number[random]
}
All of these are pretty well documented, and most have published source code.
var generator: Int = random() % 50000
Well, first of all, this is modulo biased, so it certainly won't be equivalent to a proper uniform random number. The docs for random explain it:
The random() function uses a non-linear, additive feedback, random number generator, employing a default table of size 31 long integers. It returns successive pseudo-random numbers in the range
from 0 to (2**31)-1. The period of this random number generator is very large, approximately 16*((2**31)-1).
But you can look at the full implementation and documentation in Apple's source code for libc.
Contrast the documentation for arc4random_uniform (which does not have modulo bias):
These functions use a cryptographic pseudo-random number generator to generate high quality random bytes very quickly. One data pool is used for all consumers in a process, so that consumption
under program flow can act as additional stirring. The subsystem is re-seeded from the kernel random number subsystem on a regular basis, and also upon fork(2).
And the source code is also available. The important thing to note from arc4random_uniform is that it avoids modulo bias by adjusting the modulo correctly and then generating random numbers until it is in the correct range. In principle this could require generating an unlimited number of random values; in practice it is incredibly rare that it would need to generate more than one, and rare-to-the-point-of-unbelievable that it would generate more than that.
GKRandomSource.sharedRandom() is also well documented:
The system random source shares state with the arc4random family of C functions. Generating random numbers with this source modifies the outcome of future calls to those functions, and calling those functions modifies the sequence of random values generated by this source. As such, this source is neither deterministic nor independent—use it only for trivial gameplay mechanics that do not rely on those attributes.
For performance, you would expect random() to be fastest since it never seeds itself from the system entropy pool, and so it also will not reduce the entropy in the system (though arc4random only does this periodically, I believe around every 1.5MB or so of random bytes generated; not for every value). But as with all things performance, you must profile. Of course since random() does not reseed itself it is less random than arc4random, which is itself less random than the source of entropy in the system (/dev/random).
When in doubt, if you have GameplayKit available, use it. Apple selected the implementation of sharedRandom() based on what they think is going to work best in most cases. Otherwise use arc4random. But if you really need to minimize impact on the system for "pretty good" (but not cryptographic) random numbers, look at random. If you're willing to take "kinda random if you don't look at them too closely" numbers and have even less impact on the system, look at rand. And if you want almost no impact on the system (guaranteed O(1), inlineable), see XKCD's getRandomNumber().
Xorshift generators are among the fastest non-cryptographically-secure random number generators, requiring very small code and state.
an example of swift implementation of xorshift128+
func xorshift128plus(seed0 : UInt64, seed1 : UInt64) -> () -> UInt64 {
var s0 = seed0
var s1 = seed1
if s0 == 0 && s1 == 0 {
s1 = 1 // The state must be seeded so that it is not everywhere zero.
}
return {
var x = s0
let y = s1
s0 = y
x ^= x << 23
x ^= x >> 17
x ^= y
x ^= y >> 26
s1 = x
return s0 &+ s1
}
}
// create random generator, seed as needed!!
let random = xorshift128plus(seed0: 0, seed1: 0)
for _ in 0..<100 {
// and use it later
random()
}
to avoid modulo bias, you could use
func random_uniform(bound: UInt64)->UInt64 {
var u: UInt64 = 0
let b: UInt64 = (u &- bound) % bound
repeat {
u = random()
} while u < b
return u % bound
}
in your case
let r_number = random_uniform(bound: 5000) // r_number from interval 0..<5000
I try to run the following function in julia command, but when timing the function I see too much memory allocations which I can't figure out why.
function pdpf(L::Int64, iters::Int64)
snr_dB = -10
snr = 10^(snr_dB/10)
Pf = 0.01:0.01:1
thresh = rand(100)
Pd = rand(100)
for m = 1:length(Pf)
i = 0
for k = 1:iters
n = randn(L)
s = sqrt(snr) * randn(L)
y = s + n
energy_fin = (y'*y) / L
#inbounds thresh[m] = erfcinv(2Pf[m]) * sqrt(2/L) + 1
if energy_fin[1] >= thresh[m]
i += 1
end
end
#inbounds Pd[m] = i/iters
end
#thresh = erfcinv(2Pf) * sqrt(2/L) + 1
#Pd_the = 0.5 * erfc(((thresh - (snr + 1)) * sqrt(L)) / (2*(snr + 1)))
end
Running that function in the julia command on my laptop, I get the following shocking numbers:
julia> #time pdpf(1000, 10000)
17.621551 seconds (9.00 M allocations: 30.294 GB, 7.10% gc time)
What is wrong with my code? Any help is appreciated.
I don't think this memory allocation is so surprising. For instance, consider all of the times that the inner loop gets executed:
for m = 1:length(Pf) this gives you 100 executions
for k = 1:iters this gives you 10,000 executions based on the arguments you supply to the function.
randn(L) this gives you a random vector of length 1,000, based on the arguments you supply to the function.
Thus, just considering these, you've got 100*10,000*1000 = 1 billion Float64 random numbers being generated. Each one of them takes 64 bits = 8 bytes. I.e. 8GB right there. And, you've got two calls to randn(L) which means that you're at 16GB allocations already.
You then have y = s + n which means another 8GB allocations, taking you up to 24GB. I haven't looked in detail on the remaining code to get you from 24GB to 30GB allocations, but this should show you that it's not hard for the GB allocations to start adding up in your code.
If you're looking at places to improve, I'll give you a hint that these lines can be improved by using the properties of normal random variables:
n = randn(L)
s = sqrt(snr) * randn(L)
y = s + n
You should easily be able to cut down the allocations here from 24GB to 8GB in this way. Note that y will be a normal random variable here as you've defined it, and think up a way to generate a normal random variable with an identical distribution to what y has now.
Another small thing, snr is a constant inside your function. Yet, you keep taking its sqrt 1 million separate times. In some settings, 'checking your work' can be helpful, but I think that you can be confident the computer will get it right the first time and thus you don't need to make it keep re-doing this calculation ; ). There are other similar places you can improve your code to avoid duplicate computations here that I'll leave to you to locate.
aireties gives a good answer for why you have so many allocations. You can do more to reduce the number of allocations. Using this property we know that y = s+n is really y = sqrt(snr) * randn(L) + randn(L) and so we can instead do y = rvvar*randn(L) where rvvar= sqrt(1+sqrt(snr)^2) is defined outside the loop (thanks for the fix!). This will halve the number of random variables needed.
Outside the loop you can save sqrt(2/L) to cut down a little bit of time.
I don't think transpose is special-cased yet, so try using dot(y,y) instead of y'*y. I know dot for sure is just a loop without having to transpose, while the other may transpose depending on the version of Julia.
Something that would help performance (but not allocations) would be to use one big randn(L,iters) and loop through that. The reason is because if you make all of your random numbers all at once it's faster since it can use SIMD and a bunch of other goodies. If you want to implicitly do that without changing your code much, you can use ChunkedArrays.jl where you can use rands = ChunkedArray(randn,L) to initialize it and then everytime you want a randn(L), you instead use next(rands). Inside the ChunkedArray it actually makes bigger vectors and replenishes them as needed, but like this you can just get your randn(L) without having to keep track of all of that.
Edit:
ChunkedArrays probably only save time when L is smaller. This gives the code:
function pdpf(L::Int64, iters::Int64)
snr_dB = -10
snr = 10^(snr_dB/10)
Pf = 0.01:0.01:1
thresh = rand(100)
Pd = rand(100)
rvvar= sqrt(1+sqrt(snr)^2)
for m = 1:length(Pf)
i = 0
for k = 1:iters
y = rvvar*randn(L)
energy_fin = (y'*y) / L
#inbounds thresh[m] = erfcinv(2Pf[m]) * sqrt(2/L) + 1
if energy_fin[1] >= thresh[m]
i += 1
end
end
#inbounds Pd[m] = i/iters
end
end
which runs in half the time as using two randn calls. Indeed from the ProfileViewer we get:
#profile pdpf(1000, 10000)
using ProfileView
ProfileView.view()
I circled the two parts for the line y = rvvar*randn(L), so the vast majority of the time is random number generation. Last time I checked you could still get a decent speedup on random number generation by changing to to VSL.jl library, but you need MKL linked to your Julia build. Note that from the Google Summer of Code page you can see that there is a project to make a repo RNG.jl with faster psudo-rngs. It looks like it already has a few new ones implemented. You may want to check them out and see if they give speedups (or help out with that project!)
I'm looking for a efficient, uniformly distributed PRNG, that generates one random integer for any whole number point in the plain with coordinates x and y as input to the function.
int rand(int x, int y)
It has to deliver the same random number each time you input the same coordinate.
Do you know of algorithms, that can be used for this kind of problem and also in higher dimensions?
I already tried to use normal PRNGs like a LFSR and merged the x,y coordinates together to use it as a seed value. Something like this.
int seed = x << 16 | (y & 0xFFFF)
The obvious problem with this method is that the seed is not iterated over multiple times but is initialized again for every x,y-point. This results in very ugly non random patterns if you visualize the results.
I already know of the method which uses shuffled permutation tables of some size like 256 and you get a random integer out of it like this.
int r = P[x + P[y & 255] & 255];
But I don't want to use this method because of the very limited range, restricted period length and high memory consumption.
Thanks for any helpful suggestions!
I found a very simple, fast and sufficient hash function based on the xxhash algorithm.
// cash stands for chaos hash :D
int cash(int x, int y){
int h = seed + x*374761393 + y*668265263; //all constants are prime
h = (h^(h >> 13))*1274126177;
return h^(h >> 16);
}
It is now much faster than the lookup table method I described above and it looks equally random. I don't know if the random properties are good compared to xxhash but as long as it looks random to the eye it's a fair solution for my purpose.
This is what it looks like with the pixel coordinates as input:
My approach
In general i think you want some hash-function (mostly all of these are designed to output randomness; avalanche-effect for RNGs, explicitly needed randomness for CryptoPRNGs). Compare with this thread.
The following code uses this approach:
1) build something hashable from your input
2) hash -> random-bytes (non-cryptographically)
3) somehow convert these random-bytes to your integer range (hard to do correctly/uniformly!)
The last step is done by this approach, which seems to be not that fast, but has strong theoretical guarantees (selected answer was used).
The hash-function i used supports seeds, which will be used in step 3!
import xxhash
import math
import numpy as np
import matplotlib.pyplot as plt
import time
def rng(a, b, maxExclN=100):
# preprocessing
bytes_needed = int(math.ceil(maxExclN / 256.0))
smallest_power_larger = 2
while smallest_power_larger < maxExclN:
smallest_power_larger *= 2
counter = 0
while True:
random_hash = xxhash.xxh32(str((a, b)).encode('utf-8'), seed=counter).digest()
random_integer = int.from_bytes(random_hash[:bytes_needed], byteorder='little')
if random_integer < 0:
counter += 1
continue # inefficient but safe; could be improved
random_integer = random_integer % smallest_power_larger
if random_integer < maxExclN:
return random_integer
else:
counter += 1
test_a = rng(3, 6)
test_b = rng(3, 9)
test_c = rng(3, 6)
print(test_a, test_b, test_c) # OUTPUT: 90 22 90
random_as = np.random.randint(100, size=1000000)
random_bs = np.random.randint(100, size=1000000)
start = time.time()
rands = [rng(*x) for x in zip(random_as, random_bs)]
end = time.time()
plt.hist(rands, bins=100)
plt.show()
print('needed secs: ', end-start)
# OUTPUT: needed secs: 15.056888341903687 -> 0,015056 per sample
# -> possibly heavy-dependence on range of output
Possible improvements
Add additional entropy from some source (urandom; could be put into str)
Make a class and initialize to memorize preprocessing (costly if done for each sampling)
Handle negative integers; maybe just use abs(x)
Assumptions:
the ouput-range is [0, N) -> just shift for others!
the output-range is smaller (bits) than the hash-output (may use xxh64)
Evaluation:
Check randomness/uniformity
Check if deterministic regarding input
You can use various randomness extractors to achieve your goals. There are at least two sources you can look for a solution.
Dodis et al, "Randomness Extraction and Key Derivation
Using the CBC, Cascade and HMAC Modes"
NIST SP800-90 "Recommendation for the Entropy Sources Used for
Random Bit Generation"
All in all, you can preferably use:
AES-CBC-MAC using a random key (may be fixed and reused)
HMAC, preferably with SHA2-512
SHA-family hash functions (SHA1, SHA256 etc); using a random final block (eg use a big random salt at the end)
Thus, you can concatenate your coordinates, get their bytes, add a random key (for AES and HMAC) or a salt for SHA and your output has an adequate entropy.
According to NIST, the output entropy relies on the input entropy:
Assuming you use SHA1; thus n = 160bits. Let's suppose that m = input_entropy (your coordinates' entropy)
if m >= 2n then output_entropy=n=160 bits
if 2n < m <= n then maximum output_entropy=m (but full entropy is not guaranteed).
if m < n then maximum output_entropy=m (this is your case)
see NIST sp800-90c (page 11)
To make sure that this is not a duplicate, I have already checked this and this out.
I want to generate random numbers in a specific range including step size (not continuous distribution).
For example, I want to generate random numbers between -2 and 3 in which the step between two consecutive numbers is 0.02. (e.g. [-2 -1.98 -1.96 ... 2.69 2.98 3] so a generated number should be 2.96 not 2.95).
I have tried this:
a=-2*100;
b=3*100;
r = (b-a).*rand(5,1) + a;
for i=1:length(r)
if r(i) >= 0
if mod(fix(r(i)),2)
r(i)=ceil(r(i))/100;
else
r(i)=floor(r(i))/100;
end
else
if mod(fix(r(i)),2)
r(i)=floor(r(i))/100;
else
r(i)=ceil(r(i))/100;
end
end
end
and it works.
there is an alternative way to do this in MATLAB which is :
y = datasample(-2:0.02:3,5,'Replace',false)
I want to know:
How can I make my own implementation faster (improve the
performance)?
If the second method is faster (it looks faster to me), how can I
use similar implementation in C++?
Those previous answers do cover your case if you read carefully. For example, this one produces random numbers between limits with a step size of one. But let's generalize this to an arbitrary step size in case you can't figure out how to get there. There are several different ways. Here's one using randi where we use the default step size of one and the range from one to the number possible values as indices:
lo = 2;
hi = 3;
step = 0.02;
v = lo:step:hi;
r = v(randi(length(v),[5 1]))
If you look inside datasample (type edit datasample in your command window to view the code) you'll see that it's doing something very similar to this answer. In the case of the 'Replace' option being true see around line 135 (in R2013a at least).
If the 'Replace' option is false, as in your use of datasample above, then randperm actually needs to be used instead (see around line 159):
lo = 2;
hi = 3;
step = 0.02;
v = lo:step:hi;
r = v(randperm(length(v),51))
Because there is no replacement in this case, 51 is the maximum number of values that can be requested in a call and all values of r will be unique.
In C++ you should not use rand() if you're doing scientific computing and generating large numbers of random variates. Instead you should use a large period random number generator such as Mersenne Twister (the default in Matlab). C++11 includes a version of this generator as part of . More here in rand(). If you want something fast, you should try the Double precision SIMD-oriented Fast Mersenne Twister. You'll have to ask another question if you want to implement your code in C++.
The distribution you want is a simple transform of integers, so how about:
step = 0.02
r = randi([-2 3] / step, [5, 1]) * step;
In C++, rand() generates integers too, so it should be pretty obvious how to take a similar approach there.
I have been looking at this lib Random123 and associated quote:
One mysterious man came to my booth and asked what I knew about generating random numbers with OpenCL. I told him about implementations of the Mersenne Twister, but he wasn't impressed. He told me about a new technical paper that explains how to generate random numbers on GPUs by combining integer counters and block ciphers. In reverential tones, he said that counter-based random number generators (CBRNGs) produce numbers with greater statistical randomness than the MT and with much greater speed.
I was able to get a demo running using this kernel:
__kernel void counthits(unsigned n, __global uint2 *hitsp) {
unsigned tid = get_global_id(0);
unsigned hits = 0, tries = 0;
threefry4x32_key_t k = {{tid, 0xdecafbad, 0xfacebead, 0x12345678}};
threefry4x32_ctr_t c = {{0, 0xf00dcafe, 0xdeadbeef, 0xbeeff00d}};
while (tries < n) {
union {
threefry4x32_ctr_t c;
int4 i;
} u;
c.v[0]++;
u.c = threefry4x32(c, k);
long x1 = u.i.x, y1 = u.i.y;
long x2 = u.i.z, y2 = u.i.w;
if ((x1*x1 + y1*y1) < (1L<<62)) {
hits++;
}
tries++;
if ((x2*x2 + y2*y2) < (1L<<62)) {
hits++;
}
tries++;
}
hitsp[tid].x = hits;
hitsp[tid].y = tries;
}
My questions are now, will this not generate the same random numbers every time its run, a random number is based on the global id ? How can I generate new random numbers each time. Possible to provide a seed as a parameter for the kernel and then use that somehow?
Anyone who have been using this lib and can give me some more insight in the use of it?
Yes. The example code generates the same sequences of random numbers every time it is called.
To get different streams of random numbers, just initialize any of the values k[1..3] and/or c[1..3] differently. You can initialize them from command line arguments, environment variables, time-of-day, saved state, /dev/urandom, or any other source. Just be aware that:
a) if you initialize all of them exactly the same way in two different runs, then those two runs will get the same stream of random numbers
b) if you initialize them differently in two different runs, then those two runs will get different streams of random numbers.
Sometimes you want property a). Sometimes you want property b). Take a moment to think about which you want and be sure that you're doing what you intend.
More generally, the functions in the library, e.g., threefry4x32, have no state. If you change any bit in the input (i.e., any bit in any of the elements of c or k), you'll get a completely different random, statistically independent, uniformly distributed output.
P.S. I'm one of the authors of the library and the paper "Parallel Numbers: As Easy as 1, 2, 3":
http://dl.acm.org/citation.cfm?id=2063405
If you're not a subscriber to the ACM digital library, the link above may hit a pay-wall. Alternatively, you can obtain the paper free of charge by following the link on this page:
http://www.thesalmons.org/john/random123/index.html
I can't help you with the library per se, but I can tell you that the most common way to generate random numbers in OpenCL is to save some state between calls to the kernel.
Random number generators usually use a state, from which a new state and a random number are generated. In practice, this isn't complicated at all: you just pass an extra array that holds state. In my codes, I implement random numbers as follows:
uint rand_uint(uint2* rvec) { //Adapted from http://cas.ee.ic.ac.uk/people/dt10/research/rngs-gpu-mwc64x.html
#define A 4294883355U
uint x=rvec->x, c=rvec->y; //Unpack the state
uint res = x ^ c; //Calculate the result
uint hi = mul_hi(x,A); //Step the RNG
x = x*A + c;
c = hi + (x<c);
*rvec = (uint2)(x,c); //Pack the state back up
return res; //Return the next result
#undef A
}
inline float rand_float(uint2* rvec) {
return (float)(rand_uint(rvec)) / (float)(0xFFFFFFFF);
}
__kernel void my_kernel(/*more arguments*/ __global uint2* randoms) {
int index = get_global_id(0);
uint2 rvec = randoms[index];
//Call rand_uint or rand_float a number of times with "rvec" as argument.
//These calls update "rvec" with new state, and return a random number
randoms[index] = rvec;
}
. . . then, all you do is pass an extra array that holds the RNG's state into random. In practice, you'll want to seed this array differently for each work item.
0xdecafbad, 0xfacebead, 0x12345678 and 0xf00dcafe, 0xdeadbeef, 0xbeeff00d are just arbitrarily chosen numbers, they're not special. Any other number (even 0) could be used in their place -- I'll add a comment to the example code.
You can replace any of them with variables that you pass in; the only requirement for avoiding undesirable repetition in the output random "stream" is that you avoid repeating the (c, k) input tuple. The example code uses the thread id and loop index to ensure uniqueness, but you can easily add more variables to ensure uniqueness -- e.g. count the kernel invocations in the host code and pass that counter in, use that in place of one of the elements of k or c.
By the way, despite the name 'Counter-based random number generator', there's no requirement that the inputs (c, k) be 'counters', it's just that counters happen to be the most convenient idiom for ensuring that inputs don't repeat.