f Cost-effective algorithm prototyping for HPC environments with continuous resource profiling

In current seismic industry, one of the goals of software development is to decrease the “time-to-market” for the deployment of new algorithms while obtaining overall good performances by making better use of current HPC hardware. In this presentation we will discuss the pitfalls and inefficiencies we encountered while developing a fast prototyping framework for seismic data processing over the past few years. The ability to continuously monitor the efficiency of resource usage proved to be a key feature, allowing us to fine tune performances for the different HPC architectures that came and went in these years.