Trails for MiniMorus[640/1280] and Morus[640/1280]
DEPENDENCIES
- AVX2 (for Morus1280)
- SSE (for Morus640)
- Lib OpenMp
- AES-NI
- Python3
BUILD
A simple make will show the list of default target and build the default one.
Every target correspond to one of the trail fragments:
- alpha,
- beta,
- gamma,
- delta,
- epsil,
- first approximation
- second approximation
- full trail: Combination of first and second approximation
Generating the masks and optimizations
In order to provide flexibility and speed we generate the masks C files through a Python3 script. Once generated they will be found in the newly created Masks folder.
Using C++ with OOP or C with function pointer does not allow the compiler to
predict efficiently which function will be executed next. This induce a factor
2 slowdown. By specifying with #define the version we want to use, we allow the
compiler to inline most of the code.
In order to compute C_0 + C_2 we do not loop over the bits of C. We generate a masks
where bits at positions 0 and 2 are set and & it with C.
The oddness of the Hamming Weight provide us the result.
Bits positions used in the masks are specified in masks.py
Morus implementations
The Morus implementations in SSE and AVX2 are greatly inspired from the one provided for the CAESAR competition: http://www3.ntu.edu.sg/home/wuhj/research/caesar/caesar.html.