Search arXiv⌕ Search
Home/Archive/arXiv

arXiv · 2606.19913

Design and Evaluation of Energy-Efficient Whisper Dot-Product Kernel Offloading on a CGLA Architecture

Takuto Ando·Yu Eto·Ayumu Takeuchi·Yasuhiko Nakashima
View original source ↗Read source document ↗Export metadata ↓

Abstract

In this paper, we implement and evaluate Whisper dot-product kernel offloading on IMAX, a programmable Coarse-Grained Linear Arrays (CGLAs) architecture. Whisper-tiny.en profiling on an ARM Cortex-A72 shows that dot-product operations account for 90.6% of FP16 execution time and 87.1% of Q8_0 execution time. To address this kernel bottleneck, we combine kernel mapping, local-memory sizing, and burst scheduling. The implementation uses inline FP16-to-FP32 conversion, 2-way SIMD FMA on a 64-bit datapath, column-wise multithreading, and mixed execution in which aligned vector segments run on IMAX and residual segments run concurrently on the host CPU. We evaluate the design with an FPGA prototype and a 28nm ASIC projection at 840MHz. For Whisper-tiny.en, 32KB local memory and burst length 16 jointly minimize PDP and EDP. Under a TDP-based cross-platform comparison, the projected IMAX records a PDP of 11.58J for Whisper-tiny.en Q8_0, 2.35x lower than Jetson AGX Orin (27.16J) and 10.48x lower than RTX 4090 (121.38J). The same design extends to Whisper-base.en and Whisper-small.en, where the PDP gap narrows as 32KB local-memory coverage drops from 93.8% for tiny to about 66.5% for base and small. These results position IMAX as a programmable architecture for lower-PDP local ASR in the tiny-model regime.

Explore related subjects

cs.AR

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Takuto Ando, Yu Eto, Ayumu Takeuchi, Yasuhiko Nakashima. 2026-06-18. Design and Evaluation of Energy-Efficient Whisper Dot-Product Kernel Offloading on a CGLA Architecture. https://arxiv.org/abs/2606.19913

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

A System Architecture for Low Latency Multiprogramming Quantum Computing

As quantum systems scale, multiprogramming quantum computing (MPQC) provides a practical way to improve device utilization and throughput. However, because quantum executables are device-dependent, non-portable across qubit regions, and highly susceptible to noise and crosstalk, current MPQC pipelines rely on expensive online compilation to co-optimize concurrently running programs. This online step dominates runtime and impedes low-latency deployments for practical, real-world workloads in the future, such as repeatedly invoked quantum neural network (QNN) services. We present FLAMENCO, a fidelity-aware multi-version compilation system that enables independent offline compilation and low-latency multiprogramming at runtime. \textbf{At the architecture level}, the system abstracts devices into compute units to reduce the search space of region allocation. \textbf{At compile time}, it generates diverse executable versions for each program---each bound to a distinct qubit region---allowing dynamic region selection at runtime and overcoming non-portability. \textbf{At runtime}, it employs a lightweight orchestrator that uses post-compilation fidelity metrics to avoid conflicts and mitigate crosstalk, supporting conflict-free co-execution without online co-optimization. Evaluations show that FLAMENCO achieves over 5$\times$ runtime speedup in post-scheduling execution while maintaining comparable execution fidelity on common-success workloads. When integrated into existing scheduler-coupled systems, it raises workload-level conflict-free orchestration ratio from 0.183 to 1.000 for HyperQ and from 0.050 to 0.400 for QOS.

cs.AR↗
arXiv2026

HBFSim: Fast and Faithful Simulation of High-Bandwidth Flash Under Real GPU Execution

High-Bandwidth Flash (HBF) places high-capacity NAND beside HBM to relieve the memory-capacity bottleneck of LLM inference, yet its system-level behavior cannot be evaluated before hardware becomes available. Cycle-level GPU simulators are too slow for production-scale models. Trace replay has a further shortcoming: it cannot capture the allocation, migration, and execution changes induced by different HBM-HBF configurations. Our key insight is that HBF need not be evaluated by simulating the GPU: only the program-visible effects of HBF need to be modeled. And only a real LLM workload running on real hardware can answer the arguments about HBF. Hence the modeled service has to be injected into that running program, and the injection must not destroy the GPU concurrency that would hide the original I/O latency. We present HBFSim, an open-source HBF simulator that executes LLM workloads on a real GPU while modeling HBF timing, thermal, and other behaviors online. HBFSim rewrites the PTX of the workload's kernels and routes accesses inside a registered address range into the HBF simulator. It supports asynchronous TMA transfers and capacities beyond physical GPU memory. HBFSim leaves the model's run unaffected across ordinary-memory, TMA, and capacity-mode tests. The delay it injects matches the delay requested to within 0.152%. We also design a coupled thermal module that puts HBF, HBM, and the GPU in one advanced package, which is important for answering how severe the hot throttling problem becomes after HBF runs for a long time. Experiments with Qwen3-30B show how package heating, HBM-HBF allocation, and shared MoE demand jointly constrain the design space of future HBF accelerators.

cs.AR↗
arXiv2026

Bio-inspired Learning and Decision-Making with Probabilistic In-Memory Computing Hardware: Part 2

This report extends our previous work (Part 1), which introduced an energy-based model for learning and decision-making under uncertainty. The model leverages stochastic Langevin dynamics to continuously evolve approximate probability distributions over neuron states and model weights. However, as noted in Part 1 and confirmed through GPU-based implementations, large-scale probabilistic energy-based models of this nature face significant scalability challenges due to excessive execution latency. This latency stems from a fundamental mismatch: massively parallel models with low arithmetic intensity (such as energy-based models) are being executed on processor architectures like GPUs that rely on high-bandwidth memory (HBM) interfaces. The HBM imposes brutally sequential execution constraints on inherently parallelizable models, creating the false impression that such models are unscalable. In reality, it is the GPU architecture itself, with its dependence on HBM interfaces, that is not a scalable processor architecture for this class of AI model. In this report, we demonstrate using a detailed transaction-level model (TLM) of a probabilistic analogue in-memory computing (AIMC) processor that the same energy-based model can execute well over 1000x faster than data-center-grade hardware by eliminating the HBM interface and performing computation directly within on-chip memory.

cs.AR↗