In-pixel integration of signal processing and machine learning based data filtering for particle tracking detectors
Explore the source record for details and available documents.
Parpillon, Benjamin [Fermilab]↗
Engineering topics
Tran, Nhan
Publications and source records attributed to Tran, Nhan.
Sensor co-design for Smartpixels
Explore the source record for details and available documents.
Parpillon, Benjamin↗
hls4ml: A Flexible, Open-Source Platform for Deep Learning Acceleration on Reconfigurable Hardware
We present hls4ml, a free and open-source platform that translates machine learning (ML) models from modern deep learning frameworks into high-level synthesis (HLS) code that can be integrated into full designs for field-programmable gate arrays (FPGAs) or application-specific integrated circuits (ASICs). With its flexible and modular design, hls4ml supports a large number of deep learning frameworks and can target HLS compilers from several vendors, including Vitis HLS, Intel oneAPI and Catapult HLS. Together with a wider eco-system for software-hardware co-design, hls4ml has enabled the acceleration of ML inference in a wide range of commercial and scientific applications where low latency, resource usage, and power consumption are critical. In this paper, we describe the structure and functionality of the hls4ml platform. The overarching design considerations for the generated HLS code are discussed, together with selected performance results.
FOS: Computer and information sciences↗
Deployment of inference as a service at the US CMS Tier-2 data centers
Coprocessors, especially GPUs, will be a vital ingredient of data production workflows at the HL-LHC. At CMS, the GPU-as-a-service approach for production workflows is implemented by the SONIC project (Services for Optimized Network Inference on Coprocessors). SONIC provides a mechanism for outsourcing computationally demanding algorithms, such as neural network inference, to remote servers, where requests from multiple clients are intelligently distributed across multiple GPUs by a load-balancing service. This talk highlights the recent progress in deploying SONIC at selected U.S. CMS Tier-2 data centers. Using realistic CMS Run3 data processing workflows, such as those containing transformer-based algorithms, we demonstrate how SONIC is integrated into the production-like environment to enable accelerated inference offloading. We will present developments from both the client and server sides, including production job and data center configurations for NVIDIA and AMD GPUs. We will also present performance scaling benchmarks and discuss the challenges of operating SONIC in CMS production, such as server discovery, GPU saturation, fallback server logic, etc.
Holzman, Burt↗
Radiation-Hard Smart-Pixel Detector ASIC ReadOut with Digital AI in 28nm
Detectors at future high energy colliders will face enormous technical challenges. Disentangling the unprecedented numbers of particles expected in each event will require highly granular silicon pixel detectors with billions of readout channels. With event rates as high as 40 MHz, these detectors will generate petabytes of data per second. To enable discovery within strict bandwidth and latency constraints, future trackers must be capable of fast, power efficient, and radiation hard data-reduction at the source. This effort is pursuing the co-design development of high-performance readout smart pixel ASICs for a future Phase III High Luminosity upgrade of the Large Hadron Collider. A 1.6mm2 ASIC prototype was designed by Fermilab in CMOS 28 nm bulk process and submitted for manufacturing in February 2024. It leverages the analog front-end pixel design of a previous prototype fabricated and tested in 2023, which achieved a simulated detection level of ~400e- with 30fF input capacitance. The ROIC consists of two matrices of 16×16 smart pixels, each 25×25 μm2 in size. Each smart pixel contains a charge-sensitive preamplifier with leakage current compensation and three auto-zero comparators for a 2-bit flash-type ADC. There is digital space for the integration of our fully combinatorial AI that performs momentum classification at the bunch crossing rate. The total power consumption is ∼6μW per pixel, which corresponds to ~1mW/cm2. The ASIC incorporates programmable front-end charge injection circuitry to generate pixel cluster charges during characterization. The cluster profile will be generated to duplicate hit characteristics of various momentum (pT). We will present early results from chip testing.
Parpillon, Benjamin↗