High Performance Computing

Language : C, Python, MPI, CUDA

Prerequisites

Usage

Power Method

Example

Cholesky

Example

Mandelbrot

Example

Monitor

## Result ##

To do list

• Add a summary to readme.

Power Method

• Correct and upgrade reports.
• Upgrade CUDA version to allow atomicAdd to be applied to double type.
• CUDA : Check old vs new upgrade of X and Y at the end of each iteration.
• Bug CUDA : error still infinite along iteration.
• Bug MPI : error still constant after several iteration.
• Check the good initialization of X ; if we must apply size or size*size.
• MPI : when divide block_size, handle case size/number_processes = int + decimal where decimal non null
• MPI : Optimize reading set_flattened_submatrix_from_file : error in pointer fseek depending on decimal of double values for example.
• MPI : Add alternative to MPI commands and check optimization, maybe like Ssend for example
•  Upgrade README.txt to add metadata of input for PowerMethod.

Mandelbrot

• Restart all codes and upgrade repository.
• Correct and upgrade reports.

Cholesky

• Correct and upgrade reports.
• Check for upgrade, maybe create a generic code.

MPI

• Choose what to do with those files.
• Upgrade files.