Enabling extreme scale applications on heterogeneous hardware

EPiGRAM-HS is motivated by the increasing presence of heterogeneous technologies on pre-exascale supercomputers and by the need of porting key HPC and emerging applications to these systems on time for exascale.

We are working towards delivering a new validated programming environment, extending the programmability and maximizing the productivity of application development for large-scale heterogeneous computing systems.

Subscribe to our mailing list to get quarterly EPiGRAM-HS news  

Your Email  

Your Name  

"MPI is too High-Level - MPI is too Low-Level". A relevant look at some of the MPI problems and how to go forward tackling them. Certainly, the growth of the MPI standard is a scary 😨. Insightful presentation by Prof. Marc Snir.

Emerging Memory Technologies in HPC: which challenges is HPC facing? Great talk by Prof. Vetter at KTH in conjunction with our kick-off meeting.

Interesting! @TensorFlow provides software prefetcher and caches for potential data staging on NVM. Our experiments-on much smaller scales though-show that TF input pipeline allow for complete overlap of I/O and training. Also,TF has Obj. Store APIs: easy to port DAOS&friends.

HPC Guru@HPC_Guru

#HPC file systems fail for #DeepLearning at scale - lessons from the recent exaop DeepLabv3+ run on #Summit


#AI #Storage via @NicoleHemsoth

Load More...
Image is not available

EPiGRAM-HS is consisted of 6 key European universities and enterprises focused on high performance computing.


Delivering a new validated programming environment for large-scale heterogeneous computing systems for enabling HPC and emerging deep-learning frameworks to run on large-scale heterogeneous systems at maximum performance.

Extending the programmability of large-scale heterogeneous systems with GPUs, FPGAs, HBM and NVM, by introducing new concepts, adding functionalities and carrying out implementations in two widely-used HPC programming systems for large-scale supercomputers (MPI and GASPI)

Maximizing the productivity of application development on heterogeneous supercomputers by providing auto-tuned collective communication, a framework for automatic code generation for FPGAs, a memory abstraction device comprised of APIs and a runtime for automatic data placement on diverse memories, and a DSL for large-scale deep-learning frameworks.