Lattice Basis Reduction Experiments
(Under Construction)
Test Environment:
Sun X4100 with two dual core
AMD Opteron 275 at 2.2 GHz and 4 GB of main memory
(donated by Sun Microsystems through their Academic Excellence Grant Program)
M3:
We tested the original Schnorr-Euchner algorithm for M3 lattice
bases only up to dimension 90.
- runtime of the reduction (compared to the Gram Variant)
Our experiments show that for the Schnorr-Euchner LLL and
unimodular lattice bases of type M3
the number of exact scalar products is in the same range than the number of
reduction steps.
The number of step-backs is negliable compared to number of exact scalar
products.
- comparison of exact scalar products (exact sp) and step-backs
- comparison of reduction steps and swap operations
Other types of unimodular lattice bases.
The following data has been taken from previous experimentents.
All numbers have been sorted (unless stated otherwise).
We used a fixed dimension of 50 and performed 103000 LLL reduction using the
Schorr-Euchner algorithm
Test Environment:
HPCF: 4 Compute nodes equipped with two dual core AMD Opteron 265 at 1.8 GHz and 4 GB of main memory
- number of exact scalarproducts, step-backs and runtime for X1
- number of exact scalarproducts for X2 sorted after
corresponding reduction time.
- number of exact scalarproducts for M1 sorted after
corresponding reduction time.
(We used another test setup, but bases were constructed using
M1)
- number of exact scalarproducts for M2 sorted after
corresponding reduction time.
(We used another test setup, but bases were constructed using
M2)
- number of exact scalarproducts for M3 sorted after
corresponding reduction time.
(We used another test setup, but bases were constructed using
M3)
