Software and firmware co-development using high-level synthesis

Ghanathe, Nikhil P; Madorsky, A.; Lam, Herman; Acosta, D.; George, Alan; Carver, M.; Xia, Yibin; Jyothishwara, A.; Hansen, Maxwell T.

doi:10.1088/1748-0221/12/01/c01083

Cited by 8 publications

(4 citation statements)

References 2 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The hardware implementations of the HL-LHC ATLAS and CMS L1T will use high-bandwidth serial I/O links for data communication and large, modern field-programmable gate arrays (FPGAs) for sophisticated and fast algorithms. The development and synthesis of FPGA firmware incorporating these algorithms is significantly enhanced in reliability, accessibility and performance with Higher Level Synthesis (HLS) tools [31]. The latest developments and expectations for future FPGAs not only include significant increases in the number of logic gates available and high-speed serial links, but also increases in the number of highbandwidth serial links per device, more sophisticated and fast DSPs, embedded Linux, and integration with high speed networking.…”

Section: Discussionmentioning

confidence: 99%

Triggering and High-Level Data Selection

Smith¹

2020

Particle Physics Reference Library

View full text Add to dashboard Cite

The data taken by a particle physics collider detector consists of events, which are snapshots of the detector data at specific intervals in time. Usually these snapshots are taken at the frequency of the crossing of the colliding beams. For HERA this was 96 ns, for the Tevatron Run II this was 396 ns and for the LHC at design luminosity this is 25 ns. An individual bunch crossing may contain either no, one or many interactions between the particles in the colliding beams. The time during which beam collisions take place during a beam crossing is 1-2 ns. Even if there are multiple collisions in a single crossing the detector elements will make only one recording and the events will be superimposed. Therefore, each bunch crossing is individually evaluated. Not all of the detector data from an individual crossing is available immediately. Some may be stored as charge and need digitization. Other digital detector data may be inaccessible until further detector processing is complete. The selection of bunch crossings is a highly complex function that involves a series of levels which take increasing amounts of time, process increasing amounts of data, use increasingly complex algorithms and make increasingly more precise determinations to reject increasing numbers of crossings. The first level(s) of the series usually involve(s) specific custom high-speed electronics. The subsequent level(s) involve more general CPU farms that run code similar to that found in the offline reconstruction. Due to this structure, the first level of trigger decision

show abstract

Section: Discussionmentioning

confidence: 99%

Triggering and High-Level Data Selection

Smith¹

2020

Particle Physics Reference Library

View full text Add to dashboard Cite

show abstract

“…The third stage is firmware design. The inputs to Xilinx Vivado are created [49,50]. The inputs are a combination of HLS and hardware description language (HDL).…”

Section: Jinst 16 P08016mentioning

confidence: 99%

Nanosecond machine learning event classification with boosted decision trees in FPGA for high energy physics

et al. 2021

View full text Add to dashboard Cite

We present a novel implementation of classification using the machine learning/artificial intelligence method called boosted decision trees (BDT) on field programmable gate arrays (FPGA). The firmware implementation of binary classification requiring 100 training trees with a maximum depth of 4 using four input variables gives a latency value of about 10 ns, independent of the clock speed from 100 to 320 MHz in our setup. The low timing values are achieved by restructuring the BDT layout and reconfiguring its parameters. The FPGA resource utilization is also kept low at a range from 0.01% to 0.2% in our setup. A software package called achieves this implementation. Our intended user is an expert in custom electronics-based trigger systems in high energy physics experiments or anyone that needs decisions at the lowest latency values for real-time event classification. Two problems from high energy physics are considered, in the separation of electrons vs. photons and in the selection of vector boson fusion-produced Higgs bosons vs. the rejection of the multijet processes.

show abstract

“…The third stage is firmware design. The inputs to Vivado are created [46,47]. The inputs are a combination of HLS and hardware description language (HDL).…”

Section: Introductionmentioning

confidence: 99%

Nanosecond machine learning event classification with boosted decision trees in FPGA for high energy physics

Hong,

Carlson,

Eubanks

et al. 2021

Preprint

View full text Add to dashboard Cite

We present a novel implementation of classification using the machine learning / artificial intelligence method called boosted decision trees (BDT) on field programmable gate arrays (FPGA). The firmware implementation of binary classification requiring 100 training trees with a maximum depth of 4 using four input variables gives a latency value of about 10 ns, which corresponds to 3 clock ticks at 320 MHz in our setup. The low timing values are achieved by restructuring the BDT layout and reconfiguring its parameters. The FPGA resource utilization is also kept low at a range from 0.01% to 0.2% in our setup. A software package called fwXmachina achieves this implementation. Our intended audience is a user of custom electronics-based trigger systems in high energy physics experiments or anyone that needs decisions at the lowest latency values for real-time event classification. Two problems from high energy physics are considered, in the separation of electrons vs. photons and in the selection of vector boson fusion-produced Higgs bosons vs. the rejection of the multĳet processes.

show abstract

Software and firmware co-development using high-level synthesis

Cited by 8 publications

References 2 publications

Triggering and High-Level Data Selection

Triggering and High-Level Data Selection

Nanosecond machine learning event classification with boosted decision trees in FPGA for high energy physics

Nanosecond machine learning event classification with boosted decision trees in FPGA for high energy physics

Contact Info

Product

Resources

About