Research and development

 

 

LUFAC has a highly qualified and dedicated to R & D on efficient and optimal use of new technologies in different branches such as numerical computation software, programming tools, settings and optimizations for Linux operating systems, use of libraries and new languages for parallel computing CUDA, OpenCL equally in those that are now standard as OpenMP, MPI, PosixThreads.

Our experts are graduates from leading universities in Mexico among which are::

• Computer Science UAM-I, CINVESTAV
• UNAM Faculty of Physics
• UNAM Faculty of Mathematics
• Computer Engineering Faculty of Engineering UNAM
  

Our research interests are focused on new strategies for fine-grained parallelism in numerical algorithms:

• Linear algebra.
• Differential equations
• Finite differences
• Generation of random numbers
• Mesh generation
• Visualization and graph theory
  .
  

Likewise, an ongoing exploration conducted to certify the performance and use of libraries such as:

• BLAS
• ATLAS
• LAPACK
• cuBLAS
• healers
• cuFFT
• GNU Scientific Library
• OpenMM
• FFTW
• CUSP
  

Improving the modification of tools via open source (thus being the same GPL licensing) for an efficient and friendly of all the material available programming tools such as:

• VIM with snnipet's and Highlighting syntax for Fortran and C languages, API's OpenMP, MPI, CUDA, OpenCL
• Eclipse IDE with the same features as the vim but with all the graphic surroundings added features offered by the IDE..
  

Also generate a continuous testing aplicaiciones, experience in its correct and optimal installation, configuration, tuning thus achieving its best performance for both CPU to GPU versions and covers from such products are available:

• NAMD
• Gromacs
• Matlab
  
  

The list is updated continuously weighing the needs of our customers.

As a fundamental part in the adoption of new technologies develop courses for users interested in integrating these technologies in their software and applications, many of the issues that touch on courses that are in daily use and keep updated continually with the development highest scientific development requirements could apply to be shared among the community of customers, our agenda includes:

I. - ADVANCED TOPICS IN PROGRAMMING IN C
II .- INTRODUCTION TO PARALLEL PROGRAMMING
III .- INTRODUCTION AND CUDA GPUs
IV .- PRINCIPLES AND CONCEPTS OF PROGRAMMING CUDA
V. - PROGRAM STRUCTURE CUDA and enthrone
VI .- EXAMPLES OF PROGRAMS CUDA with numerical methods of linear algebra and differential equations.
IV .- COOPERATION BETWEEN WIRES
V. - GPU EVENT MANAGEMENT
VI .- MultiGPU MANAGEMENT
VII .- GPU debugging
Parallelization strategy .- Vlll OF A SERIAL
IX.-NUMERICAL METHODS
X. - THE USE OF LIBRARIES
XI.-MPI-OpenMP hybrid programming, CUDA-OpenCL

Where the trend in the learning curve is as shown in the chart below.

 

With the help of our experienced R & D LUFAC via a thorough analysis can advise prior to porting your code or application to any computer technology trend to use the following levels of parallelism

• CPU
• OpenMP
• MPI
• Pthreads
• OpenCL
• Cluster's
• GPU-NVIDIA
• CUDA
• GPU
• OpenCL
• Multi-GPU
• Clusters Híbridos CPU-GPU
• MPI-openMP-CUDA
• MPI-openMP-OpenCL
• pthreads-CUDA