Releases: vxst/qrand
Releases · vxst/qrand
v1.0.0
This is the first release for qrand project.
It can be used as a drop-in replacement for STL random generator like:
#include <random>
#include <iostream>
#include "qrand.h"
int main(){
std::uniform_int_distribution<int> dist(0, 100);
std::uniform_real_distribution<double> distf(0.0, 1.0);
qrand rng;
for(int i = 0; i < 10; i++){
std::cout << dist(rng) << " " << distf(rng) << std::endl;
}
}
Or you can use the callable object as a drop in replacement for rand function:
#include <iostream>
#include "qrand.h"
int main(){
qrand randq;
for(int i = 0; i < 10; i++){
std::cout << randq() << " " << randq() % 100 << std::endl;
}
}
There is no static global state, so it's thread safe as long as it's thread local:
void some_thread_function(){
static thread_local qrand randq;
// call randq
}
Or in OpenMP
extern qrand randq;
#pragma omp threadprivate(randq)
qrand randq;
For the versioning, given a version number MAJOR.MINOR.PATCH,
- "PATCH" is guaranteed to generate same sequence of random number with the same seed
- "MINOR" may change the sequence with the same seed, but will put in release notes if it happens. It is guaranteed to keep the API unchanged
- "MAJOR" may include the API change
It passes the BigCrush(and also SmallCrush, Crush) tests with the following results(with the default eps=1e-300, eps1=1e-15):
========= Summary results of SmallCrush =========
Version: TestU01 1.2.3
Generator: qrand
Number of statistics: 15
Total CPU time: 00:00:04.65
All tests were passed
========= Summary results of Crush =========
Version: TestU01 1.2.3
Generator: qrand
Number of statistics: 144
Total CPU time: 00:18:43.85
All tests were passed
========= Summary results of BigCrush =========
Version: TestU01 1.2.3
Generator: qrand
Number of statistics: 160
Total CPU time: 02:04:35.23
All tests were passed
The reproducibility is tested on following platform:
Intel Xeon E5-2650 (Sandy Bridge EP, AES-NI+SSE)
Intel Xeon E5-2670 v2 (Ivy Bridge EP, AES-NI+SSE)
Intel Xeon E5-2686 v4 (Broadwell EP, AES-NI+AVX2)
Intel Xeon Platinum 8259CL (Cascade Lake, AES-NI+AVX512F)
Intel Xeon Platinum 8375C (Ice Lake, VAES+AVX512F)
AMD EPYC 7571 (Zen/Naples, AES-NI+AVX2)
AMD EPYC 7R32 (Zen2/Rome, AES-NI+AVX2)
AMD EPYC 7R13 (Zen3/Milan, VAES+AVX2)
Apple M1 (Rosetta 2/GCC, AES-NI+SSE)