Routing FPGAs is a challenging problem because of the relative scarcity of routing resources, both wires and connection points. This can lead either to slow implementations caused by long wiring paths that avoid congestion or a failure to route all signals. This paper presents PathFinder, a router that balances the goals of performance and routability. PathFinder uses an iterative algorithm that converges to a solution in which all signals are routed while achieving close to the optimal performance allowed by the placement. Routability is achieved by forcing signals to negotiate for a resource and thereby determine which signal needs the resource most. Delay is minimized by allowing the more critical signals a greater say in this negotiation. Because PathFinder requires only a directed graph to describe the architecture of routing resources, it adapts readily to a wide variety of FPGA architectures such as Triptych, Xilinx 3000 and mesh-connected arrays of FPGAs. The results of routing ISCAS benchmarks on the Triptych FPGA architecture show an average increase of only 4.5% in critical path delay over the optimum delay for a placement. Routes of ISCAS benchmarks on the Xilinx 3000 architecture show a greater completion rate than commercial tools, as well as 11% faster implementations.

/pdf/pathfinder-a-negotiation-based-performance-driven-router-for-29xkvzrcmp.pdf

PathFinder: A Negotiation-Based Performance-Driven Router for FPGAs

YOLOv7 surpasses all known object detectors in both speed and accuracy in the range from 5 FPS to 160 FPS and has the highest accuracy 56.8% AP among all known real-time object detectors with 30 FPS or higher on GPU V100. YOLOv7-E6 object detector (56 FPS V100, 55.9% AP) outperforms both transformer-based detector SWIN-L Cascade-Mask R-CNN (9.2 FPS A100, 53.9% AP) by 509% in speed and 2% in accuracy, and convolutional-based detector ConvNeXt-XL Cascade-Mask R-CNN (8.6 FPS A100, 55.2% AP) by 551% in speed and 0.7% AP in accuracy, as well as YOLOv7 outperforms: YOLOR, YOLOX, Scaled-YOLOv4, YOLOv5, DETR, Deformable DETR, DINO-5scale-R50, ViT-Adapter-B and many other object detectors in speed and accuracy. Moreover, we train YOLOv7 only on MS COCO dataset from scratch without using any other datasets or pre-trained weights. Source code is released in https://github.com/WongKinYiu/yolov7.

YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors

This survey presents an overview of recent advances in the state of the art for computer-aided design (CAD) tools for analog and mixed-signal integrated circuits (ICs). Analog blocks typically constitute only a small fraction of the components on mixed-signal ICs and emerging systems-on-a-chip (SoC) designs. But due to the increasing levels of integration available in silicon technology and the growing requirement for digital systems to communicate with the continuous-valued external world, there is a growing need for CAD tools that increase the design productivity and improve the quality of analog integrated circuits. This paper describes the motivation and evolution of these tools and outlines progress on the various design problems involved: simulation and modeling, symbolic analysis, synthesis and optimization, layout generation, yield analysis and design centering, and test. This paper summarizes the problems for which viable solutions are emerging and those which are still unsolved.

/pdf/computer-aided-design-of-analog-and-mixed-signal-integrated-23xbvj9rc2.pdf

Computer-aided design of analog and mixed-signal integrated circuits

An experimental technique is described for observing the effects of switching transients in digital MOS circuits that perturb analog circuits integrated on the same die by means of coupling through the substrate Various approaches to reducing substrate crosstalk (the use of physical separation of analog and digital circuits, guard rings, and a low-inductance substrate bias) are evaluated experimentally for a CMOS technology with a substrate comprising an epitaxial layer grown on a heavily doped bulk wafer Observations indicate that reducing the inductance in the substrate bias is the most effective Device simulations are used to show how crosstalk propagates via the heavily doped bulk and to predict the nature of substrate crosstalk in CMOS technologies integrated in uniform, lightly doped bulk substrates, showing that in such cases the substrate noise is highly dependent on layout geometry A method of including substrate effects in SPICE simulations for circuits fabricated on epitaxial, heavily doped substrates is developed >

Experimental Results and Modeling Techniques for Substrate Noise in Mixed-Signal Integrated Circuits

We describe a new method for determining component values and transistor dimensions for CMOS operational amplifiers (op-amps). We observe that a wide variety of design objectives and constraints have a special form, i.e., they are posynomial functions of the design variables. As a result, the amplifier design problem can be expressed as a special form of optimization problem called geometric programming, for which very efficient global optimization methods have been developed. As a consequence we can efficiently determine globally optimal amplifier designs or globally optimal tradeoffs among competing performance measures such as power, open-loop gain, and bandwidth. Our method, therefore, yields completely automated sizing of (globally) optimal CMOS amplifiers, directly from specifications. In this paper, we apply this method to a specific widely used operational amplifier architecture, showing in detail how to formulate the design problem as a geometric program. We compute globally optimal tradeoff curves relating performance measures such as power dissipation, unity-gain bandwidth, and open-loop gain. We show how the method can he used to size robust designs, i.e., designs guaranteed to meet the specifications for a variety of process conditions and parameters.

/pdf/optimal-design-of-a-cmos-op-amp-via-geometric-programming-4thsejbwi6.pdf

Optimal design of a CMOS op-amp via geometric programming

This paper discusses Voltage Islands, a system architecture and chip implementation methodology, that can be used to dramatically reduce active and static power consumption for System-on-Chip (SoC) designs. As technology scales for increased circuit density and performance, the need to reduce power consumption increases in significance as designers strive to utilize the advancing silicon capabilities. The consumer product market further drives the need to minimize chip power consumption. Effective use of Voltage Islands for meeting SoC power and performance requirements, while meeting Time to Market (TAT) demands, requires novel approaches throughout the design flow as well as special circuit components and chip powering structures. This paper outlines methods being used today to design Voltage Islands in a rapid-TAT product development environment, and discusses the need for industry EDA advances to create an industry-wide Voltage Island design capability.

/pdf/managing-power-and-performance-for-system-on-chip-designs-40v9f44inq.pdf

Managing power and performance for system-on-chip designs using Voltage Islands

The authors describe KOAN and ANAGRAM II, new tools for device-level analog placement and routing. Analog layout tools that merely apply known digital macrocell techniques fall short of achieving the density and performance of handcrafted analog cells. KOAN and ANAGRAM II differ from previous approaches by using general algorithmic techniques to find critical device-level layout optimizations rather than relying on a large library of fixed-topology module generators. New placement algorithms implemented in KOAN handle complex layout symmetries, dynamic merging and abutment of individual devices, and flexible generation of wells and bulk constants. New routing algorithms implemented in ANAGRAM II handle arbitrary gridless design rules in addition to over-the-device, crosstalk-avoiding, mirror-symmetric, and self-symmetric wiring. Examples of CMOS and BiCMOS analog cell layouts produced by these tools are presented. >

KOAN/ANAGRAM II: new tools for device-level analog placement and routing

/pdf/koan-anagram-11-new-tools-for-device-level-analog-placement-zl3r88czzf.pdf

KOAN/ANAGRAM 11: New Tools for Device-Level Analog Placement and Routing

Machine learning on tiny IoT devices based on microcontroller units (MCU) is appealing but challenging: the memory of microcontrollers is 2-3 orders of magnitude smaller even than mobile phones. We propose MCUNet, a framework that jointly designs the efficient neural architecture (TinyNAS) and the lightweight inference engine (TinyEngine), enabling ImageNet-scale inference on microcontrollers. TinyNAS adopts a two-stage neural architecture search approach that first optimizes the search space to fit the resource constraints, then specializes the network architecture in the optimized search space. TinyNAS can automatically handle diverse constraints (i.e.device, latency, energy, memory) under low search costs.TinyNAS is co-designed with TinyEngine, a memory-efficient inference library to expand the search space and fit a larger model. TinyEngine adapts the memory scheduling according to the overall network topology rather than layer-wise optimization, reducing the memory usage by 4.8x, and accelerating the inference by 1.7-3.3x compared to TF-Lite Micro and CMSIS-NN. MCUNet is the first to achieves >70% ImageNet top1 accuracy on an off-the-shelf commercial microcontroller, using 3.5x less SRAM and 5.7x less Flash compared to quantized MobileNetV2 and ResNet-18. On visual&audio wake words tasks, MCUNet achieves state-of-the-art accuracy and runs 2.4-3.4x faster than MobileNetV2 and ProxylessNAS-based solutions with 3.7-4.1x smaller peak SRAM. Our study suggests that the era of always-on tiny machine learning on IoT devices has arrived. Code and models can be found here: this https URL.

/pdf/mcunet-tiny-deep-learning-on-iot-devices-3itd5v6603.pdf

MCUNet: Tiny Deep Learning on IoT Devices

Power dissipation is becoming the most challenging design constraint in nanometer technologies. Among various design implementation schemes, standard cell ASICs offer the best power efficiency for high-performance applications. The flexibility of ASICs allow for the use of multiple voltages and multiple thresholds to match the performance of critical regions to their timing constraints, and minimize the power everywhere else. We explore the trade-off between multiple supply voltages and multiple threshold voltages in the optimization of dynamic and static power. The use of multiple supply voltages presents some unique physical and electrical challenges. Level shifters need to be introduced between the various voltage regions. Several level shifter implementations will be shown. The physical layout needs to be designed to ensure the efficient delivery of the correct voltage to various voltage regions. More flexibility can be gained by using appropriate level shifters. We will discuss optimization techniques such as clock skew scheduling which can be effectively used to push performance in a power neutral way.

/pdf/pushing-asic-performance-in-a-power-envelope-4a5b0cexum.pdf

John M. Cohn

Papers

Managing power and performance for system-on-chip designs using Voltage Islands

KOAN/ANAGRAM II: new tools for device-level analog placement and routing

KOAN/ANAGRAM 11: New Tools for Device-Level Analog Placement and Routing

MCUNet: Tiny Deep Learning on IoT Devices

Pushing ASIC performance in a power envelope