The PERFACCT way to speed up your applications

Specialists for Software Optimization and Highly Efficient Software Architectures

In Potsdam, PERFACCT develops efficient and scalable software solutions for highly specialized industrial and research applications with high quality standards. As a development service provider, we offer holistic solutions from mathematical modeling including interactive data visualization, through the development of highly efficient software architectures, to software optimization.

Our mission is to deliver the optimal result for the customer.

Software Optimization

We identify potential performance improvements in your software codes. We provide you with profiling results, an overview of critical code sections, and recommendations for improvement opportunities in a report tailored to your team.

Draw from our experience. Let our specialists optimize your codes for your target architectures. If possible, we improve your algorithms in terms of runtime and memory requirements. Furthermore, in many cases the performance can already be optimized by code changes. As far as the algorithm supports it, our experts will parallelize your application for selected target architectures.

Highly Efficient Software Architectures

A lack of foresight in development can risk the long-term success of software solutions. Highly efficient software architectures must be scalable, otherwise a not insignificant amount of development time is lost in refactoring existing codes. Therefore, our software engineers place special emphasis on ensuring that extensions can be implemented with as few adjustments as possible.

The software we develop is not only performant in execution. Through appropriate modularization and adherence to clean code principles, it is also very easy to customize.

Modeling & Simulations

Efficient computer simulations are still one of the most important means to validate models. Some of them are so complex that entire cluster systems are necessary for the calculation.

We transform your mathematical model into scalable simulations and help you with algorithm design as well as the evaluation of your models. Our specialists use technologies that ensure optimal execution times without having to risk losses in the development speed or maintainability of your software.


Use our solutions of automatic code generation and parallelization for your Python and MATLAB programs. Concentrate entirely on your modeling and save valuable development time. The programs MatMACC and PyMACC generate high-performance binaries from your computationally intensive codes using state-of-the-art compiler techniques. Evaluate your models on a wide range of target architectures, from multicore CPUs to manycore GPUs (NVIDIA, AMD).

Industries of our customers

Our customers appreciate our initiative and open communication in the projects. We always try to think one step ahead in development to be able to offer our customer an optimal result in the long run.

We cooperate with companies and research institutes from various industries:



Manufacturers and suppliers are increasingly turning to software and modeling tools to optimize their vehicles. With the increase in complexity in vehicles, require more and more software solutions to improve the performance of their computationally intensive simulations, as well as tools for simpler development processes.

Advances in GPU hardware have enabled breakthroughs in machine learning in particular computer vision (autonomous driving). However, optimization on GPU machines remains a major challenge for many.


Health Care

New modeling approaches make it possible to better understand interrelationships in the body in order to derive treatment options or therapy plans. In view of the many possible influencing variables and the large number of patients, the calculations can sometimes be very time-consuming.


Geo Science

Researchers are still trying to understand how our planet is built and what interactions take place on it. With the help of software simulations and visualizations, these can be better understood.

Often the calculations are performed on meshes or two- or three-dimensional grids. The algorithms used allow data points at different grid points to be calculated independently of each other. This approach is optimally suited for parallelization on multi-core processors.

Contact us

We will get back to you as soon as possible.