Speaker
Description
Summary
Two major trends can be identified in the development of trigger and DAQ
systems for particle-physics experiments: the massive use of general-purpose
commodity systems such as commercial PC farms for data acquisition, and
the reduction of trigger levels implemented in hardware, towards a pure
software selection system ("triggerless").
The NA62 experiment at the CERN SPS aims at measuring an ultra-rare decay
of the charged kaon; the signal has to be extracted from a huge background
which is ten orders of magnitude more frequent.
With an input particle rate of 10 MHz, some tens of thousands detector
channels and the requirement of avoiding zero suppression as much as
possible, triggerless readout into PCs is not affordable.
An innovative approach aims at exploiting the parallel computing power of
commercial GPUs to perform fast computations in software in the early
trigger stages.
General-purpose computing on GPUs is emerging as a new paradigm in several
fields of science, although so far applications have been tailored to the
specific strengths of such devices, exploiting parallelization and avoiding
real-time applications.
With the steady reduction of GPU latencies, and the increase in link and
memory throughputs, the use of such devices for real-time applications in
high-energy physics data acquisition and trigger systems is becoming ripe.
A pilot project within NA62 aims at integrating GPUs into the central
L0 trigger processor, and also to use them as a fast online processors
for computing trigger primitives.
Several TDC-equipped sub-detectors with sub-nanosecond time resolution will
participate to the first-level NA62 trigger (L0), fully integrated with the
data-acquisition system, to reduce the readout rate of all sub-detectors to
1 MHz, using multiplicity information asynchronously computed over time
windows of a few ns, both for positive sub-detectors and for vetos.
The online use of GPUs would allow the computation of more complex trigger
primitives already at this first trigger level.
Cheap commercial links can be used to collect trigger primitives, and the
task of the dedicated central processor is to perform time matching of those,
generate the trigger signal and re-align it in time for synchronous
distribution.
We describe the architecture of the proposed system and present the
performances achieved in tests on a real detector data acquisition system,
to perform online recognition of rings from a RICH detector with
sub-nanosecond time resolution.
The challenges and the prospects of this promising approach are discussed.
