Sign up for our newsletter and get the latest HPC news and analysis.
Send me information from insideHPC:


Four Reasons that Software Development is in the HPC Driving Seat for 2016

David Lecomber, CEO, Allinea

David Lecomber, CEO, Allinea

Sponsored Post

Hardware and the Top 500 ranking always grabs effort and headlines in HPC, but for a sometimes Pyrrhic victory of raw power over utility.

It has often been at the expense of training, tools and code development that would unleash far more scientific output from the same “tin”.

2016 is set to change all that.

Just as representative benchmarks like HPCG are set to replace Linpack, so a focus on software is taking over. From industry analysts to users at SC15 we heard that software is the number one challenge and the number one opportunity to have world-class impact.

The key drivers of this attention?

  1. Code may run slower on newer hardware

Funding bodies and sponsors will not accept a significant HPC investment delivering less scientific output.

The Knights Landing Platform will provide many-core nodes – but without software preparation, anticipated lower clock-speeds will hamper performance.  KNL has great potential but code modernization is vital. Using threads, SIMD vector units and HBM memory should be transformational.

Developers and computational scientists are essential  – and to apply their skills requires tools such as Allinea Forge. Forge finds vectorization, threading and memory access optimizations – as well I/O and MPI bottlenecks – and can debug the changes!

  1. Code may not run at all on new hardware

We’re set for a wider range of viable architectures than at any time in recent memory: NVIDIA GPU on ARM or OpenPOWER, or an Intel Xeon or Intel Xeon Phi.

Performance and correctness on one platform does not necessarily transfer to another. It’s complex work and requires deep software insight that only tools can provide.

Allinea Forge is HPC’s cross-platform development tool suite. Users can switch platforms easily and compare systems like-for-like, and can ensure changes for one architecture do not degrade performance on others.

  1. Awareness of the performance problem is rising

As more codes are modernized, more hero stories emerge – which leads to competition amongst codes and highlights the potential for improvement elsewhere!  Again the developer is the key.

Faster, better codes attract larger user-bases with greater numbers of publications, and, for commercial codes, significant customer wins or business gains.

Our Performance Reports tool helps users, admins and developers to understand performance problems quickly and how to tune them – and helps to target the development efforts quickly when necessary.

  1. Access to HPC is changing and widening

In the tug-of-war between nationwide consolidation, local systems and on-demand cloud deployment, the outcome isn’t clear. Consolidation may be a trend, but initiatives such as Open HPC will enable easier construction of on-demand and physical clusters. Staying relevant for all outcomes means recognizing where the value lies.

At SC15 many centers spoke of a drive to achieve recognition through software excellence on their new machine.

The success of a center is the outstanding science that its members and users create. That means empowering those who develop and use the software to succeed – through training and providing the tools they need to do the job well!

While many thought-leading sites have pushed software excellence alongside excellent systems,  this thinking is finally going mainstream.

For that reason, 2016 will be a great year for HPC.

Allinea – with our performance analysis and integrated development tools – is excited by this future!

This article was submitted by David Lecomber, CEO, Allinea.

Resource Links: