Related papers: Planck 2015 results. XII. Full Focal Plane simulat…
This paper presents an efficient approach for multiplierless implementation for eight-point DCT approximation, which based on coordinate rotation digital computer (CORDIC) algorithm. The main design objective is to make critical path of…
We present a novel low latency CMOS hardware accelerator for fully connected (FC) layers in deep neural networks (DNNs). The FC accelerator, FC-ACCL, is based on 128 8x8 or 16x16 processing elements (PEs) for matrix-vector multiplication,…
FP8 is a natural progression for accelerating deep learning training inference beyond the 16-bit formats common in modern processors. In this paper we propose an 8-bit floating point (FP8) binary interchange format consisting of two…
Post-training quantization (PTQ) is a powerful technique for model compression, reducing the numerical precision in neural networks without additional training overhead. Recent works have investigated adopting 8-bit floating-point…
We describe in detail the processing of a set of images of the z = 0.42 supercluster MS0302 taken with the UH8K camera at CFHT. The result of this is a pair of seamless combined V- and I-band images of the field, along with a…
We present a new map-making method for CMB measurements. The method is based on the destriping technique, but it also utilizes information about the noise spectrum. The low-frequency component of the instrument noise stream is modelled as a…
We review recent advances in the capabilities of the open source ab initio Quantum Monte Carlo (QMC) package QMCPACK and the workflow tool Nexus used for greater efficiency and reproducibility. The auxiliary field QMC (AFQMC) implementation…
Rate-dependent effects in the electronics used to instrument the tagger focal plane at the MAX IV Laboratory were recently investigated using the novel approach of Monte Carlo simulation to allow for normalization of high-rate experimental…
FPGAs are a promising platform for accelerating Deep Learning (DL) applications, due to their high performance, low power consumption, and reconfigurability. Recently, the leading FPGA vendors have enhanced their architectures to more…
Scientific foundation models hold great promise for advancing nuclear and particle physics by improving analysis precision and accelerating discovery. Yet, progress in this field is often limited by the lack of openly available large scale…
Szapudi et al (2001) introduced the method of estimating angular power spectrum of the CMB sky via heuristically weighted correlation functions. Part of the new technique is that all (co)variances are evaluated by massive Monte Carlo…
We apply state-of-the art data analysis methods to a number of fictitious CMB mapping experiments, including 1/f noise, distilling the cosmological information from time-ordered data to maps to power spectrum estimates, and find that in all…
MADmap is a software application used to produce maximum-likelihood images of the sky from time-ordered data which include correlated noise, such as those gathered by Cosmic Microwave Background (CMB) experiments. It works efficiently on…
The latest industrial inference engines, such as FasterTransformer and TurboTransformers, have verified that half-precision floating point (FP16) and 8-bit integer (INT8) quantization can greatly improve model inference speed. However, the…
This thesis is the result of my work as research fellow at IASF-MI, Milan section of the Istituto di Astrofisica Spaziale e Fisica Cosmica, part of INAF, Istituto Nazionale di Astrofisica. This work started in January 2006 in the context of…
We discuss the effects of inhomogeneous sky-coverage on CMB lens reconstruction, focusing on application to the recently launched Planck satellite. We discuss the mean-field which is induced by noise inhomogeneities, as well as three…
Training tool-calling agents requires large-scale trajectory data with verifiable labels, yet existing approaches either synthesize environments that diverge from real API behavior or generate tasks without ground-truth outcomes for…
Among the algorithms that are likely to play a major role in future exascale computing, the fast multipole method (FMM) appears as a rising star. Our previous recent work showed scaling of an FMM on GPU clusters, with problem sizes in the…
This study explores the use of automatic BLAS offloading and INT8-based emulation for accelerating traditional HPC workloads on modern GPU architectures. Through the use of low-bitwidth integer units and cache-coherent Unified Memory…
We present Overall FLOP Utilization (OFU), a hardware-level, precision-agnostic GPU efficiency metric for AI workloads on HPC systems, derived from two on-chip performance counters: Tensor Pipe Activity and SM clock frequency. OFU requires…