Researchers have proposed Mixture of Contexts, a long video generation framework that learns to route each query to the most relevant segments of the video sequence, instead of relying on uniform or static sparse attention or a fixed selection strategy. arxiv.org/pdf/2508.21058

In this paper is presented SwizzlePerf, a LLM workflow that automatically generates spatial optimizations for GPU kernels on disaggregated architectures by giving LLMs explicit hardware-awareness. arxiv.org/pdf/2508.20258

In this paper is presented OmniSim, a framework that extend C-level simulation capability of HLS tools by enabling both functionality and performance simulations for those complex dataflow designs that are currently unsupported or considered infeasible. arxiv.org/pdf/2508.19299

For the first time, researchers reported the realization of multi-value probabilistic computing by leveraging the thermally activated diffusion of magnetic skyrmions through an effectively non-flat energy landscape defined by a discrete number of sites. arxiv.org/pdf/2508.19623

A thin metallic film on germanium, in the presence of water, results in a remarkable pattern-forming system, such as this beautiful spiral spontaneously etched on the surface with a total structure diameter of 680 μm.

Researchers have demonstrated graphene spin-field-effect junctions where the electric field can control ultrafast spin currents and spin dynamics in thin-film ferromagnets. PRL link: journals.aps.org/prl/pdf/10.1...

In this paper, researchers have developed a neural approach for the topology aware interpolation of scalar fields losses based on persistence diagrams, for constraining the topology and geometry of the output interpolations. arxiv.org/pdf/2508.17995

In this paper is presented an ab-initio transistor simulation of unprecedented scale including electron-electron interactions within the self-consistent GW approximation, carefully optimized to take advantage of the Alps and Frontier supercomputers. #HPC arxiv.org/pdf/2508.19138

Nvidia researchers have developed a unified framework for real-time radiance field rendering on light field displays, supporting a wide range of radiance field representations within a shared architecture based on a single-pass plane sweeping strategy. arxiv.org/pdf/2508.18540

In this paper, researchers have developed a comprehensive framework for managing multi-instance GPU partitions that can significantly improve throughput, power consumption, memory utilization, and task response time. arxiv.org/pdf/2508.18556

In this paper is presented an ab-initio transistor simulation of unprecedented scale including electron-electron interactions within the self-consistent GW approximation, carefully optimized to take advantage of the Alps and Frontier supercomputers. #HPC arxiv.org/pdf/2508.19138

In this paper is introduced Targeted Wearout Attack, a stealthy, software-based attack leveraging transistor aging to induce timing faults and resulting threats like data corruption. arxiv.org/pdf/2508.16868

In this paper is proposed Wave Tracing, a new formulation that generalizes its wave-based light transport equation and maintains the crucial features of modeling transport in small regions of space, so that only observable wave effects are accounted for. arxiv.org/pdf/2508.17386

Researchers have demonstrated the first-ever nontrivial, biologically realistic connectome simulated on neuromorphic computing hardware, implementing the whole-brain connectome of the fruit fly on the Intel Loihi 2 neuromorphic platform. arxiv.org/pdf/2508.16792

In this paper is presented a comprehensive analysis of GPT-OSS-20B on a single Nvidia H100 GPU against dense baselines. arxiv.org/pdf/2508.16700

In this paper, researchers have demonstrated that an optical chirality flux signal can solely arise from the contributions of source current and the magnetic induction. arxiv.org/pdf/2508.15803

In this paper is introduced the Hardwired-Neurons Language Processing Unit, a high specialized processor that hardwires LLM weight parameters into its compute fabric. arxiv.org/pdf/2508.16151

In this paper is presented RIROS, a parallel RTL fault simulation framework which integrates structural-level and fault-level parallelism approaches to minimize bubbles that arise from parallel core load imbalances. arxiv.org/pdf/2508.16376

Researchers have presented the first experimental deployment of a hybrid classical–quantum computing environment within an operational HPC datacenter, integrating multiple QPUs and Nvidia GPUs into a multi-user, scheduler-managed cluster. arxiv.org/pdf/2508.16297

The less final decision-making power the current board of directors has, the more secure Intel's future will be. Hopefully, we'll soon see significant changes to the board of directors, and the bad actors will finally be removed. www.intc.com/news-events/...

Today is Intel's anniversary. I sincerely hope that Intel can recover from these difficult times and rid itself of the real culprits for its problems who are still on its board of directors.

In this paper is presented a detailed experimental analysis of NVIDIA’s Blackwell architecture through microbenchmarks with a comparison to the previous Hopper generation GPUs. arxiv.org/pdf/2507.10789

Microsoft Quantum researchers have presented a hardware realization and measurements of a tetron qubit device in a superconductor-semiconductor heterostructure, which support four Majorana zero modes (MZMs) when tuned into the topological phase. arxiv.org/pdf/2507.08795

In this paper is proposed Physics-based ASIC, a transformative paradigm that directly leverages the physical dynamics intrinsic to computation, rather than expending resources to impose idealized digital abstractions. arxiv.org/pdf/2507.10463

In this paper, researchers have demonstrated the ferroelectric behavior of a hybrid superconducting Josephson Fe-FET operating at a cryogenic sub-K temperature. arxiv.org/pdf/2507.04773

In this paper is introduced a hybrid Waveguide Neural Operator, a novel neural operator to solve the EUV diffraction problem, replacing only the most computationally intensive part of the WG method with a neural network. arxiv.org/pdf/2507.04153

In this paper is introduced the HLStrans, the first large-scale dataset of over 23K paired C and HLS programs harvested from academic repositories for LLM-powered HLS code synthesis from software. arxiv.org/pdf/2507.04315

In this paper is proposed NeuroPDE, a hardware design for neuromorphic PDE solvers which exploit the intrinsic physical randomness of MTJ devices to create a scalable neuromorphic architecture capable of performing random walk functions. arxiv.org/pdf/2507.04677

Meta researchers have developed Neuralocks, a novel method that achieves high-performance dynamic neural hair simulation with a compact network size. arxiv.org/pdf/2507.05191

Meta researchers have developed ANY4, a learned 4-bit weight quantization solution for large language models, providing arbitrary numeric representations without requiring pre-processing of weights or activations. arxiv.org/pdf/2507.04610

n this paper is proposed a hybrid ECC architecture that pairs large-size Reed–Solomon codes with fine-grained CRC, allowing to reduce HBM cost by shifting fault tolerance from HBM stack to the system level. arxiv.org/pdf/2507.02654

Researchers have demonstrated a compact, all-solid-state platform for the generation of intense, sub-50fs DUV pulses at 200 nm with pulse energies up to 130 μJ, achieved via cascaded sum-frequency generation in BBO crystals pumped by 800 nm fs pulses. arxiv.org/pdf/2507.02756

In this paper is presented a high-level overview of the threat model and security principles that guide the design of NVIDIA Hopper GPUs operating in GPU Confidential Computing mode. arxiv.org/pdf/2507.02770

Help a journalist out? I'm looking to talk to researchers, engineers, students or entrepreneurs who are considering leaving the U.S. because they lost funding or are worried about visas or anything like that. I'm at mimsical.94 on Signal

In this paper is presented a hardware accelerator-based online scheduler for scheduling tasks with unpredictable timing characteristics targeting heterogeneous computing resources whose availability is emergent and evolves at runtime. arxiv.org/pdf/2507.01113

It's been a long time since I posted anything on my Substack and I would like to apologize to everyone. I hope that in this new month that is approaching I can pull myself together and start publishing more frequently again.

Researchers have demonstrated that qubits based on Majorana Zero Modes (MZMs) will have decoherence because of the presence of 1/f noise that is ubiquitous in semiconductors, even if the nanowires are defect-free and the temperature is very low. arxiv.org/pdf/2506.22394

In this paper presented an optimised, high-speed FPGA implementations of two spectral correlation density estimation techniques. arxiv.org/pdf/2506.18003

In this paper, researchers have demonstrated quantized electron pumping using two parallel-connected FDSOI devices integrated on the same integrated circuit die, fabricated using the commercial 22FDX platform. arxiv.org/pdf/2506.15956

In this paper, researchers have proposed a novel compiler designed specifically for massively-parallel, spatial, dataflow architectures like the Wafer Scale Engine from Cerebras. arxiv.org/pdf/2506.15875

In this paper is presented a new compiler transformation to map iterations of SPMM to GPU threads, emphasizing thread-level optimization while facilitating register reuse. arxiv.org/pdf/2506.15174

In this paper, researchers have demosntrated that the real cause of Physics-Informed NNs failure modes is the early stopping of the optimization process, which can be easily solved by improving the training arithmetic precision of the mode. arxiv.org/pdf/2505.10949

In this paper is presented Prophet, a profile-guided solution for temporal prefetching that maintains the metadata table within hardware temporal prefetchers while leveraging profile-guided methods to improve metadata table management. arxiv.org/pdf/2506.15985

In this paper is presented Sudoku, the first software-based tool to automatically decompose full DRAM address mappings into channel, rank, bank group, and bank functions while identifying row and column bits. arxiv.org/pdf/2506.15918

In this paper is presented a self-stabilized, time-bin encoded QKD system based on a deterministic telecom-wavelength quantum dot source. arxiv.org/pdf/2506.15520

In this paper is proposed a design method of the all-facet illuminator for high NA EUV lithography exposure tool and a matching method based on deep RL for the double facets. arxiv.org/pdf/2506.15558

In this paper is presented GG, the first CISC-to-RISC transpiler built via a custom-trained, architecture-aware language model, achieving a test accuracy of 99.39% on ARMv8 and 89.93% on RISC-V64. arxiv.org/pdf/2506.14606

In this paper is introduced the Tensor Manipulation Unit, a reconfigurable, near-memory hardware block designed to execute data-movement-intensive operators efficiently in AI SoCs. arxiv.org/pdf/2506.14364

Researchers have proposed ReFrame, a novel cache mechanism for neural networks that exploits the temporal coherence of real-time rendering workloads. arxiv.org/pdf/2506.13814

Researchers have demonstrated an Ising machine that combines a hafnium-oxide memristor crossbar and stochastic MTJs, providing an intrinsic, nearly circuit-free annealing solver. arxiv.org/pdf/2506.14676