Customized accelerators have revolutionized modern computing by delivering substantial gains in energy efficiency and performance through hardware specialization. Field-Programmable Gate Arrays (FPGAs) play a crucial role in this paradigm, offering unparalleled flexibility and high-performance potential. High-Level Synthesis (HLS) and source-to-source compilers have simplified FPGA development by translating high-level programming languages into hardware descriptions enriched with directives. However, achieving high Quality of Results (QoR) remains a significant challenge, requiring intricate code transformations, strategic directive placement, and optimized data communication. This article presents Prometheus , a holistic optimization framework that integrates key optimizations - including task fusion, tiling, loop permutation, computation-communication overlap, and concurrent task execution -into a unified design space. By leveraging Non-Linear Programming (NLP) methodologies , Prometheus explores the optimization space under strict resource constraints, enabling automatic bitstream generation. Unlike existing frameworks, Prometheus considers interdependent transformations and dynamically balances computation and memory access. We evaluate Prometheus across multiple benchmarks, demonstrating its ability to maximize parallelism, minimize execution stalls, and optimize data movement. The results showcase its superior performance compared to state-of-the-art FPGA optimization frameworks, highlighting its effectiveness in delivering high QoR while reducing manual tuning efforts.

High-Level Synthesis compilers and Design Space Exploration tools have greatly advanced the automation of hardware design, improving development time and performance. However, achieving a good Quality of Results still requires extensive manual code transformations, pragma insertion, and tile size selection, which are typically handled separately. The design space is too large to be fully explored by this fragmented approach. It is too difficult to navigate this way, limits the exploration of potential optimizations, and complicates the design generation process.

High-Level Synthesis enables the rapid prototyping of hardware accelerators, by combining a high-level description of the functional behavior of a kernel with a set of micro-architecture optimizations as inputs. Such optimizations can be described by inserting pragmas e.g., pipelining and replication of units, or even higher level transformations for HLS such as automatic data caching using the AMD/Xilinx Merlin compiler. Selecting the best combination of pragmas, even within a restricted set, remains particularly challenging and the typical state-of-practice uses design-space exploration to navigate this space. But due to the highly irregular performance distribution of pragma configurations, typical DSE approaches are either extremely time consuming, or operating on a severely restricted search space. This work proposes a framework to automatically insert HLS pragmas in regular loop-based programs, supporting pipelining, unit replication, and data caching. We develop an analytical performance and resource model as a function of the input program properties and pragmas inserted, using non-linear constraints and objectives. We prove this model provides a lower bound on the actual performance after HLS. We then encode this model as a Non-Linear Program, by making the pragma configuration unknowns of the system, which is computed optimally by solving this NLP. This approach can also be used during DSE, to quickly prune points with a (possibly partial) pragma configuration, driven by lower bounds on achievable latency. We extensively evaluate our end-to-end, fully implemented system, showing it can effectively manipulate spaces of billions of designs in seconds to minutes for the kernels evaluated.

High-Level Synthesis enables the rapid prototyping of hardware accelerators, by combining a high-level description of the functional behavior of a kernel with a set of micro-architecture optimizations as inputs. Such pragmas may describe the pipelining and replication of units, or even higher-level transformations for HLS such as automatic data caching using the AMD/Xilinx Merlin compiler. Selecting the best combination of pragmas, even within a restricted set, remains particularly challenging and the typical state-of-practice uses design-space exploration to navigate this space. But due to the highly irregular performance distribution of pragma configurations, typical DSE approaches are either extremely time consuming, or operating on a severely restricted search space. In this work we propose a framework to automatically insert HLS pragmas in regular loop-based programs, supporting pipelining, unit replication (coarse- and fine-grain), and data caching. We develop a simple analytical performance and resource model as a function of the input program properties and pragmas inserted. We prove this model provides a lower bound on the actual performance for any possible configuration. We then encode this model as a Non-Linear Program, by making the pragma configuration as unknowns of the system, which is computed optimally by solving this NLP. This approach can also be used during DSE, to quickly prune points with a (possibly partial) pragma configuration, employing this latency lower bound property. We extensively evaluate our end-to-end fully implemented system, showing it can effectively manipulate spaces of billions of designs in seconds to minutes for the kernels evaluated.

A wide range of scientific and machine learning applications depend on highly optimized implementations of tensor computations. Exploiting the full capacity of a given processor architecture remains a challenging task, due to the complexity of the microarchitectural features that come into play when seeking near-peak performance. Among the state-of-the-art techniques for loop transformations for performance optimization, AutoScheduler [Zheng et al. 2020a ] tends to outperform other systems. It often yields higher performance as compared to vendor libraries, but takes a large number of runs to converge, while also involving a complex training environment. In this article, we define a structured configuration space that enables much faster convergence to high-performance code versions, using only random sampling of candidates. We focus on two-dimensional convolutions on CPUs. Compared to state-of-the-art libraries, our structured search space enables higher performance for typical tensor shapes encountered in convolution stages in deep learning pipelines. Compared to auto-tuning code generators like AutoScheduler, it prunes the search space while increasing the density of efficient implementations. We analyze the impact on convergence speed and performance distribution, on two Intel x86 processors and one ARM AArch64 processor. We match or outperform the performance of the state-of-the-art oneDNN library and TVM’s AutoScheduler, while reducing the autotuning effort by at least an order of magnitude.

Cloud applications based on the "Functions as a Service" (FaaS) paradigm have become very popular. Yet, due to their stateless nature, they must frequently interact with an external data store, which limits their performance. To mitigate this issue, we introduce OFC, a transparent, vertically and horizontally elastic in-memory caching system for FaaS platforms, distributed over the worker nodes. OFC provides these benefits cost-effectively by exploiting two common sources of resource waste: (i) most cloud tenants overprovision the memory resources reserved for their functions because their footprint is non-trivially input-dependent and (ii) FaaS providers keep function sandboxes alive for several minutes to avoid cold starts. Using machine learning models adjusted for typical function input data categories (e.g., multimedia formats), OFC estimates the actual memory resources required by each function invocation and hoards the remaining capacity to feed the cache. We build our OFC prototype based on enhancements to the OpenWhisk FaaS platform, the Swift persistent object store, and the RAM-Cloud in-memory store. Using a diverse set of workloads, we show that OFC improves by up to 82 % and 60 % respectively the execution time of single-stage and pipelined functions.