linbox
Namespaces | Constant Groups | Functions
benchmark-crafixed.C File Reference

Benchmarking fixed CRA routines. More...

#include "benchmarks/benchmark.h"
#include "linbox/util/debug.h"
#include "linbox/field/modular.h"
#include "linbox/field/modular-balanced.h"
#include "linbox/matrix/random-matrix.h"
#include "linbox/algorithms/blas-domain.h"
#include "linbox/algorithms/rns.h"
#include "linbox/algorithms/cra-domain.h"
#include "linbox/algorithms/cra-early-multip.h"
#include "linbox/integer.h"
+ Include dependency graph for benchmark-crafixed.C:

Namespaces

 LinBox
 Namespace in which all linbox code resides.
 

Constant Groups

 LinBox
 Namespace in which all linbox code resides.
 

Functions

template<bool Unsigned>
int bench_cra (index_t n, index_t m, index_t l, LinBox::PlotData< index_t > &Data)
 Bench CRA. More...
 

Detailed Description

Benchmarking fixed CRA routines.

Here we make benchmarks for CRT (Chinese Remaindering Theorem/Algorithm) in the following case. Let $\mathbf{v}$ be a vector of size $n$ and whose entries have at most $l$ bits (signed or unsigned). Suppose that we only know $\mathbf{v} \mod p_i$ for many primes $p_i$. We try and reconstruct $\mathbf{v}$ from these residues.

We benchmark for one vector or $m$ repetitions on different vectors.

We use the implementations in LinBox, Givaro, IML and NTL (if the latter two are available).

Warning
this is not a benchmark for one integer to reconstruct or a for BlasMatrix.

Function Documentation

int bench_cra ( index_t  n,
index_t  m,
index_t  l,
LinBox::PlotData< index_t > &  Data 
)

Bench CRA.

Parameters
nsize of vector to reconstruct (>1)
mnumber of vectors to reconstruct
lsize of the integers
Template Parameters
Unsigneduse >=0 random integers or not.
Parameters
Datacollects timings