BEST4HEP (Project Scientist in Charge, Borut Kersevan)
======================================================

The existing advanced approaches to programming, from vectorisation to machine
learning, need to become a skill mastered by the new generation of particle
physicists. This knowledge then needs to be incorporated into our HEP algorithms
and frameworks, as they urgently need to be developed to optimally use the
resources on the computing facilities we expect to have available for HL-LHC
computing (e.g. HPC farms with ‘accelerator’ CPU/GPU components, FPGA clusters,
TPUs, and eventually down to Quantum Computing in case of technological
breakthroughs).

HEP has been very successful in the last decade at implementing large scale
distributed computing systems and has been driving the development and adoption
of the close to Exascale Distributed Computing (Grid) Technologies in Europe.
Today, the resources we support and which are at the base of the CERN/LHC
successes, span from hundreds of thousands of computing cores, to Exabyte sized
storage systems. But we need to do more, with a predicted increase in needs more
than 10x during the coming decade; we need to utilize in full the existing
resources (e.g. leadership supercomputers with hardware accelerators), as well
as embrace new technologies, as they become available, and form on them a new
generation of computing-savvy physicists.

The BEST4HEP project will develop and disseminate a knowledge-base about the
optimal accelerator (e.g., Nvidia GPU/Tesla, FPGAs, CSAs , TPUs etc..)
programming and software workflows to be employed in the next generation of HEP
software (WP1). We will develop advanced and innovative algorithms on modern
computing platforms for various kinds of HEP workflows and tasks, from
simulation to data reconstruction, job brokering and data placement (WP2). To
efficiently connect leadership HPC computing facilities and future computing
solutions to large scale scientific data stores and software repositories we
shall develop HEP software processing frameworks that can exploit heterogeneous
computing architectures in a robust manner (WP3). Modern accelerators, GPUs,
FPGAs, Quantum Computing or even unexpected technologies are most probably going
to be a part of any future computing scenario; this needs a long range program
on training, introduction to the new platforms, access to either emulators or
early systems, using knowledge inside and outside the HEP community (WP4).