30 March 2021 in IEEE Transactions on Components, Packaging and Manufacturing Technology
With the increasing popularity and applications of 2.5-D integration, both chip and packaging industries are making significant progress in this direction. In advanced high-density 2.5-D packages, package redistribution layers become similar to chiplet back-end-of-line routing layers, and the gap between them scales down with pin density improvement. Chiplet-package interactions become significant and severely affect system performance and reliability. Moreover, 2.5-D integration offers opportunities to apply novel design techniques. The traditional die-by-die design approach neither carefully considers these interactions nor fully exploits the cross-boundary design opportunities. In this article, we present a holistic chiplet-package co-optimization flow for high-density 2.5-D packaging technologies with little performance overhead and zero pipeline-depth increase. Our holistic extraction can capture all parasitics from chiplets and the package and improve system performance through iterative optimizations. Both drop-in and pay-as-you-use design methodologies are implemented for agile development and quick turn-around time. To prove the effectiveness of our flow, we implement several design cases of a microcontroller system in TSMC 65-nm technology. Our design methodologies can reduce the performance gap by 85% with respect to the 2-D reference design after holistic optimizations. We demonstrate design flexibility and development cost-saving by presenting several flavors of a three chiplets system. To validate our flow in silicon, we tape-out a chip in TSMC 65-nm technology with measured data and validated functionality.
11 February 2026 in ACM Transactions on Design Automation of Electronic Systems
2.5D packaging has become a popular alternative to integrate advanced logic and memory chiplets for high-performance computing and artificial intelligence systems. In the conventional design flow, chiplets and packages are independently designed and then integrated at the assembly stage. To bridge the gap between chiplet designs and package designs, existing chiplet-package co-design methods iteratively optimize chiplet layouts to improve the performance of the entire system. However, Redistribution Layer (RDL) routing, which finishes the interconnections between chiplets at the package level and significantly affects the system performance, is neglected in the existing co-design flows. Therefore, this article proposes an effective chiplet-package co-design flow focusing on the RDL routing to optimize the package system performance dynamically. The proposed co-design flow can fill in the missing link, package-level co-optimization, of previous design flows. In the proposed co-design flow, we propose an efficient RDL routing algorithm to iteratively optimize the substrate layout based on the cross-boundary timing context extracted from both chiplets and the package. The proposed RDL routing algorithm has two critical techniques, including (1) a Maximal Independent Set-based (MIS-based) pin assignment method to dynamically optimize the pin positions of nets and (2) a network-flow-based router to generate routing layouts. Experimental results show that the proposed design flow can gradually improve the maximum frequency of a real design to the target performance, 400 MHz.
21 November 2025 in ACM Transactions on Reconfigurable Technology and Systems
The matrix operations that underpin today’s deep learning models are routinely implemented in Single Instruction Multiple Data (SIMD) domain specific accelerators. SIMD accelerators including GPUs and array processors can effectively leverage parallelism in models that are compute-bound, but their effectiveness can be diminished for models that are memory-bound. Processing-in-Memory (PIM) architectures are being explored to provide better energy efficiency and scalable performance for these memory-bound models. Modern Field Programmable Gate Arrays (FPGAs) feature hundreds of megabits of Static Random Access Memory (SRAM) distributed across the device as disaggregated memory resources. This makes FPGAs ideal programmable platforms for developing custom Processor In/Near Memory accelerators. Several PIM array-based accelerator designs have been proposed to leverage this substantial internal bandwidth. However, results reported to date show the FPGA based PIM architectures operating at system clock frequencies well below a chips Block-RAM (BRAM) Fmax clock frequency. Results also show that the compute densities of the designs do not scale linearly with BRAM densities. These results indicate that FPGA PIM architectures will never be competitive with their custom Application-Specific Integrated Circuit (ASIC) counterparts. In this article, we introduce DA-VinCi, a D eep-Learning A ccelerator O v erlay using In -Memory C omput i ng. DA-VinCi is the first scalable FPGA based PIM deep-learning accelerator overlay capable of clocking at the maximum frequency of a device’s BRAM. Further, the architecture of DA-VinCi allows the number of compute units to scale linearly up to the maximum capacity of a devices BRAM, and at the maximum clock frequency of the BRAM. The DA-VinCi overlay has a programmable Instruction Set Architecture (ISA) that allows the same synthesized design to provide low-latency inferencing of a range of memory-bound deep-learning models, including Multilayer Perceptrons, Recurrent Neural Network, Long Short-Term Memory, and Gated Recurrent Unit networks. The scalability and high clocking frequency of DA-VinCi is achieved through a new Processor In Memory (PIM) tile architecture and a highly scalable system-level framework. We present results showing DA-VinCi linearly scaling the number of Processing Elements (PEs) to 100% of the BRAM capacity (over 60K PEs) on an Alveo U55 clocking at 737 MHz, the chips BRAM Fmax. We provide comparative studies on inference latency across multiple deep-learning applications that show DA-VinCi achieves up to a 201 \(\times\) improvement over a state-of-the-art PIM overlay accelerator, up to 87 \(\times\) improvement over existing PIM-based FPGA accelerators, and up to 57 \(\times\) improvement over custom deep-learning accelerators on FPGAs.