arXiv subjects

Rakesh Kumar

Publications and source records attributed to Rakesh Kumar.

At least 37 records · Page 2Linked to original sources

arXiv2025

Diffusion-Guided Gaussian Splatting for Large-Scale Unconstrained 3D Reconstruction and Novel View Synthesis

Recent advancements in 3D Gaussian Splatting (3DGS) and Neural Radiance Fields (NeRF) have achieved impressive results in real-time 3D reconstruction and novel view synthesis. However, these methods struggle in large-scale, unconstrained environments where sparse and uneven input coverage, transient occlusions, appearance variability, and inconsistent camera settings lead to degraded quality. We propose GS-Diff, a novel 3DGS framework guided by a multi-view diffusion model to address these limitations. By generating pseudo-observations conditioned on multi-view inputs, our method transforms under-constrained 3D reconstruction problems into well-posed ones, enabling robust optimization even with sparse data. GS-Diff further integrates several enhancements, including appearance embedding, monocular depth priors, dynamic object modeling, anisotropy regularization, and advanced rasterization techniques, to tackle geometric and photometric challenges in real-world settings. Experiments on four benchmarks demonstrate that GS-Diff consistently outperforms state-of-the-art baselines by significant margins.

cs.CV↗

arXiv2025

Virtuoso: Enabling Fast and Accurate Virtual Memory Research via an Imitation-based Operating System Simulation Methodology

The unprecedented growth in data demand from emerging applications has turned virtual memory (VM) into a major performance bottleneck. Researchers explore new hardware/OS co-designs to optimize VM across diverse applications and systems. To evaluate such designs, researchers rely on various simulation methodologies to model VM components.Unfortunately, current simulation tools (i) either lack the desired accuracy in modeling VM's software components or (ii) are too slow and complex to prototype and evaluate schemes that span across the hardware/software boundary. We introduce Virtuoso, a new simulation framework that enables quick and accurate prototyping and evaluation of the software and hardware components of the VM subsystem. The key idea of Virtuoso is to employ a lightweight userspace OS kernel, called MimicOS, that (i) accelerates simulation time by imitating only the desired kernel functionalities, (ii) facilitates the development of new OS routines that imitate real ones, using an accessible high-level programming interface, (iii) enables accurate and flexible evaluation of the application- and system-level implications of VM after integrating Virtuoso to a desired architectural simulator. We integrate Virtuoso into five diverse architectural simulators, each specializing in different aspects of system design, and heavily enrich it with multiple state-of-the-art VM schemes. Our validation shows that Virtuoso ported on top of Sniper, a state-of-the-art microarchitectural simulator, models the memory management unit of a real high-end server-grade page fault latency of a real Linux kernel with high accuracy . Consequently, Virtuoso models the IPC performance of a real high-end server-grade CPU with 21% higher accuracy than the baseline version of Sniper. The source code of Virtuoso is freely available at https://github.com/CMU-SAFARI/Virtuoso.

cs.AR↗

arXiv2025

Filling systems of maximum size

Let $S_g$ be a closed orientable surface of genus $g\geq 2$. A collection $Ω= \{ γ_1, \dots, γ_s\}$ of pairwise non-homotopic simple closed curves on $S_g$ such that $γ_i$ and $γ_j$ are in minimal position, is called a \emph{filling system} or a \emph{filling} of $S_g$ if the complement $S_g\setminus Ω$ is a disjoint union of $b$ topological discs for some $b\geq 1$. The \emph{size} of a filling system is defined as the number of its elements. We prove that the maximum size of a filling system on $S_g$ with $ 1 \leq b \leq 2g-2$ boundary components is $2g+b-1$. Furthermore, we give a lower bound on mapping class group orbits of filling systems of maximum size with $ 1 \leq b \leq g-2$ boundary components.

math.GT↗

arXiv2024

Boosting the transparency of metallic SrNbO3 through Ti doping

In recent years, various materials have been developed to reduce the reliance of industries on Indium, a primary component of transparent conducting oxides (TCOs) used in the current generation of devices. The leading candidates for indium free TCOs are strontium vanadates, niobates and molybdates -- strongly correlated perovskite systems that exhibit high intrinsic electrical conductivity and optimal transparency. In this work, we focus on the strontium niobate thin films and manipulate its optical conductivity by Ti doping, which shifts the plasma frequency and reduces electronic correlations. This allows us to achieve a low resistance for Ti doped SNO thin films, while maintaining a high transparency in the visible spectrum. We obtain the optimal figure-of-merit (FOM) of 10.3 ($10^{-3}Ω^{-1}$) for $x = 0.3$. This FOM significantly outperforms the optoelectronic capabilities of Tin-doped Indium oxide (ITO) and several other proposed transparent conductor materials. Our research paves the way for designing the next generation of transparent conductors, guided by insights from density-functional theory (DFT) and dynamical mean-field theory (DMFT).

cond-mat.mtrl-sci↗

arXiv2024

Exact staggered dimer ground state and its stability in a two-dimensional magnet

Finding an exact solution for a realistic interacting quantum many-body problem is often challenging. There are only a few problems where an exact solution can be found, usually in a narrow parameter space. Here, we propose a spin-$1/2$ Heisenberg model on a square lattice with spatial anisotropy and bond depletion for the nearest-neighbor antiferromagnetic interactions but not for the next-nearest-neighbor interactions. This model has an \emph{exact} and \emph{unique} dimer ground state at $J_2/J_1=1/2$; a dimer state is a product state of spin-singlets on dimers (here, staggered nearest-neighbor bonds). We examine this model by employing the bond-operator mean-field theory and exact diagonalization. These analytical and numerical methods precisely affirm the correctness of the dimer ground state at the exact point ($J_2/J_1=1/2$). As one moves away from the exact point, the dimer order melts and vanishes when the spin gap becomes zero. The mean-field theory with harmonic approximation indicates that the dimer order persists for $-0.35\lesssim J_2/J_1\lesssim 1.35$. However, in non-harmonic approximation, the upper critical point lowers by $0.28$ to $1.07$, but the lower critical point remains intact. The exact diagonalization results suggest that the latter approximation fares better. The model reveals Néel order below the lower critical point and stripe magnetic order above the upper critical point. It has a topologically equivalent model on a honeycomb lattice where the nearest-neighbor interactions are still spatial anisotropic, but the bond depletion shifts into the isotropic next-neighbor interactions. Moreover, these models can also be generalized in the three dimensions.

cond-mat.str-el↗

arXiv2024

Victima: Drastically Increasing Address Translation Reach by Leveraging Underutilized Cache Resources

Address translation is a performance bottleneck in data-intensive workloads due to large datasets and irregular access patterns that lead to frequent high-latency page table walks (PTWs). PTWs can be reduced by using (i) large hardware TLBs or (ii) large software-managed TLBs. Unfortunately, both solutions have significant drawbacks: increased access latency, power and area (for hardware TLBs), and costly memory accesses, the need for large contiguous memory blocks, and complex OS modifications (for software-managed TLBs). We present Victima, a new software-transparent mechanism that drastically increases the translation reach of the processor by leveraging the underutilized resources of the cache hierarchy. The key idea of Victima is to repurpose L2 cache blocks to store clusters of TLB entries, thereby providing an additional low-latency and high-capacity component that backs up the last-level TLB and thus reduces PTWs. Victima has two main components. First, a PTW cost predictor (PTW-CP) identifies costly-to-translate addresses based on the frequency and cost of the PTWs they lead to. Second, a TLB-aware cache replacement policy prioritizes keeping TLB entries in the cache hierarchy by considering (i) the translation pressure (e.g., last-level TLB miss rate) and (ii) the reuse characteristics of the TLB entries. Our evaluation results show that in native (virtualized) execution environments Victima improves average end-to-end application performance by 7.4% (28.7%) over the baseline four-level radix-tree-based page table design and by 6.2% (20.1%) over a state-of-the-art software-managed TLB, across 11 diverse data-intensive workloads. Victima (i) is effective in both native and virtualized environments, (ii) is completely transparent to application and system software, and (iii) incurs very small area and power overheads on a modern high-end CPU.

cs.AR↗

arXiv2023

Utopia: Fast and Efficient Address Translation via Hybrid Restrictive & Flexible Virtual-to-Physical Address Mappings

Conventional virtual memory (VM) frameworks enable a virtual address to flexibly map to any physical address. This flexibility necessitates large data structures to store virtual-to-physical mappings, which leads to high address translation latency and large translation-induced interference in the memory hierarchy. On the other hand, restricting the address mapping so that a virtual address can only map to a specific set of physical addresses can significantly reduce address translation overheads by using compact and efficient translation structures. However, restricting the address mapping flexibility across the entire main memory severely limits data sharing across different processes and increases data accesses to the swap space of the storage device, even in the presence of free memory. We propose Utopia, a new hybrid virtual-to-physical address mapping scheme that allows both flexible and restrictive hash-based address mapping schemes to harmoniously co-exist in the system. The key idea of Utopia is to manage physical memory using two types of physical memory segments: restrictive and flexible segments. A restrictive segment uses a restrictive, hash-based address mapping scheme that maps virtual addresses to only a specific set of physical addresses and enables faster address translation using compact translation structures. A flexible segment employs the conventional fully-flexible address mapping scheme. By mapping data to a restrictive segment, Utopia enables faster address translation with lower translation-induced interference. Utopia improves performance by 24% in a single-core system over the baseline system, whereas the best prior state-of-the-art contiguity-aware translation scheme improves performance by 13%.

cs.AR↗

arXiv2023

Vieta-Lucas Wavelet based schemes for the numerical solution of the singular models

In this paper, numerical methods based on Vieta-Lucas wavelets are proposed for solving a class of singular differential equations. The operational matrix of the derivative for Vieta-Lucas wavelets is derived. It is employed to reduce the differential equations into the system of algebraic equations by applying the ideas of the collocation scheme, Tau scheme, and Galerkin scheme respectively. Furthermore, the convergence analysis and error estimates for Vieta-Lucas wavelets are performed. In the numerical section, the comparative analysis is presented among the different versions of the proposed Vieta-Lucas wavelet methods, and the accuracy of the approaches is evaluated by computing the errors and comparing them to the existing findings.

math.NA↗

arXiv2023

Groups generated by Dehn Twists along fillings of surfaces

Let $S_g$ denote a closed oriented surface of genus $g \geq 2$. A set $Ω= \{ c_1, \dots, c_d\}$ of pairwise non-homotopic simple closed curves on $S_g$ is called a filling system or simply a filling of $S_g$, if $S_g\setminus Ω$ is a union of $\ell$ topological discs for some $\ell\geq 1$. For $1\leq i\leq d$, let $T_{c_i}$ denotes the Dehn twist along $c_i$. In this article, we show that for each $d\geq 2$, there exists a filling $Ω=\{c_1,c_2,\dots, c_d\}$ of $S_g$ such that the group $\langle T_{c_1}, T_{c_2},\dots,T_{c_d}\rangle$ is isomorphic to the free group of rank $d$.

math.GT↗

arXiv2023

Existence and uniqueness of weak solutions to the Smoluchowski coagulation equation with source and sedimentation

This article is devoted to a generalized version of Smoluchowski's coagulation equation. This model describes the time evolution of a system of aggregating particles under the effect of external input and output particles. We show that for a large class of coagulation kernels, output rates, and exponentially decaying input rates, there is a weak solution. Moreover, the solution satisfies the mass-conservation property for linear coagulation rate and an additional condition on input and output rates. The uniqueness of weak solutions is also established by applying additional restrictions on the rates.

math.AP↗

arXiv2023

Understanding Interactions Between Chip Architecture and Uncertainties in Semiconductor Supply and Demand

Mitigating losses from supply and demand volatility in the semiconductor supply chain and market has traditionally been cast as a logistics and forecasting problem. We investigate how the architecture of a family of chips influences how it is affected by supply and demand uncertainties. We observe that semiconductor supply chains become fragile, in part, due to single demand paths, where one chip can satisfy only one demand. Chip architects can enable multiple paths to satisfy a chip demand, which improves supply chain resilience. Based on this observation, we study composition and adaptation as architectural strategies to improve resilience to volatility and also introduce a third strategy of dispersion. These strategies allow multiple paths to satisfy a given chip demand. We develop a model to analyze the impact of these architectural techniques on supply chain costs under different regimes of uncertainties and evaluate what happens when they are combined. We present several interesting and even counterintuitive observations about the product configurations and market conditions where these interventions are impactful and where they are not. In all, we show that product redesign supported by architectural changes can mitigate nearly half of the losses caused by supply and demand volatility. As far as we know, this is the first such investigation concerning chip architecture.

cs.AR↗

arXiv2023

Hardware Acceleration of Neural Graphics

Rendering and inverse-rendering algorithms that drive conventional computer graphics have recently been superseded by neural representations (NR). NRs have recently been used to learn the geometric and the material properties of the scenes and use the information to synthesize photorealistic imagery, thereby promising a replacement for traditional rendering algorithms with scalable quality and predictable performance. In this work we ask the question: Does neural graphics (NG) need hardware support? We studied representative NG applications showing that, if we want to render 4k res. at 60FPS there is a gap of 1.5X-55X in the desired performance on current GPUs. For AR/VR applications, there is an even larger gap of 2-4 OOM between the desired performance and the required system power. We identify that the input encoding and the MLP kernels are the performance bottlenecks, consuming 72%,60% and 59% of application time for multi res. hashgrid, multi res. densegrid and low res. densegrid encodings, respectively. We propose a NG processing cluster, a scalable and flexible hardware architecture that directly accelerates the input encoding and MLP kernels through dedicated engines and supports a wide range of NG applications. We also accelerate the rest of the kernels by fusing them together in Vulkan, which leads to 9.94X kernel-level performance improvement compared to un-fused implementation of the pre-processing and the post-processing kernels. Our results show that, NGPC gives up to 58X end-to-end application-level performance improvement, for multi res. hashgrid encoding on average across the four NG applications, the performance benefits are 12X,20X,33X and 39X for the scaling factor of 8,16,32 and 64, respectively. Our results show that with multi res. hashgrid encoding, NGPC enables the rendering of 4k res. at 30FPS for NeRF and 8k res. at 120FPS for all our other NG applications.

cs.AR↗

arXiv2023

Cross-View Visual Geo-Localization for Outdoor Augmented Reality

Precise estimation of global orientation and location is critical to ensure a compelling outdoor Augmented Reality (AR) experience. We address the problem of geo-pose estimation by cross-view matching of query ground images to a geo-referenced aerial satellite image database. Recently, neural network-based methods have shown state-of-the-art performance in cross-view matching. However, most of the prior works focus only on location estimation, ignoring orientation, which cannot meet the requirements in outdoor AR applications. We propose a new transformer neural network-based model and a modified triplet ranking loss for joint location and orientation estimation. Experiments on several benchmark cross-view geo-localization datasets show that our model achieves state-of-the-art performance. Furthermore, we present an approach to extend the single image query-based geo-localization approach by utilizing temporal information from a navigation pipeline for robust continuous geo-localization. Experimentation on several large-scale real-world video sequences demonstrates that our approach enables high-precision and stable AR insertion.

cs.CV↗

arXiv2023

A brief introduction to bulk viscosity of fluids

Fluid flows are typically studied by solving the Navier--Stokes equation. One of the fundamental assumptions of this equation is Stokes' hypothesis. This hypothesis assumes bulk viscosity, to be identically zero. The Stokes' hypothesis is a reasonable approximation for commonly observed fluid flows; therefore, Navier--Stokes equation gives satisfactory results in these situations. However, there are circumstances where this hypothesis does not hold good, and therefore, the classical Navier--Stokes equation becomes inapt. These situations include the absorption of sound waves, hypersonic flows, turbulent flows, and flow of Martian air etc. Reliable analytical and computational studies of these flows requires account of bulk viscosity in the governing equations. In this article, we provide a brief review of the subject of bulk viscosity. We start with a brief background of this topic. Then we discuss the underlying microscopic mechanisms that give rise to bulk viscosity effects. It was followed by a review of methods available in the literature for estimation of this parameter. Finally, a review of the studies that analyze the effects of bulk viscosity in various fluid flows is provided.

physics.flu-dyn↗

arXiv2023

A Storage-Effective BTB Organization for Servers

Many contemporary applications feature multi-megabyte instruction footprints that overwhelm the capacity of branch target buffers (BTB) and instruction caches (L1-I), causing frequent front-end stalls that inevitably hurt performance. BTB capacity is crucial for performance as a sufficiently large BTB enables the front-end to accurately resolve the upcoming execution path and steer instruction fetch appropriately. Moreover, it also enables highly effective fetch-directed instruction prefetching that can eliminate a large portion L1-I misses. For these reasons, commercial processors allocate vast amounts of storage capacity to BTBs. This work aims to reduce BTB storage requirements by optimizing the organization of BTB entries. Our key insight is that storing branch target offsets, instead of full or compressed targets, can drastically reduce BTB storage cost as the vast majority of dynamic branches have short offsets requiring just a handful of bits to encode. Based on this insight, we size the ways of a set associative BTB to hold different number of target offset bits such that each way stores offsets within a particular range. Doing so enables a dramatic reduction in storage for target addresses. Our final design, called BTB-X, uses an 8-way set associative BTB with differently sized ways that enables it to track about 2.24x more branches than a conventional BTB and 1.3x more branches than a storage-optimized state-of-the-art BTB organization, called PDede, with the same storage budget.

cs.AR↗

arXiv2022

Bulk viscosity of dilute gases and their mixtures

In this work, we use the Green-Kubo method to study the bulk viscosity of various dilute gases and their mixtures. First, we study the effects of the atomic mass on the bulk viscosity of dilute diatomic gas by estimating the bulk viscosity of four different isotopes of nitrogen gas. We then study the effects of addition of noble gas on the bulk viscosity of dilute nitrogen gas. We consider mixtures of nitrogen with three noble gases, viz., neon, argon, and krypton at eight different compositions between pure nitrogen to pure noble gas. It is followed by an estimation of bulk viscosity of pure oxygen and mixtures of nitrogen and oxygen for various compositions. In this case, three different composition are considered, viz., 25\% N$_2$ + 75\% O$_2$, 50\% N$_2$ + 50\% O$_2$, and 78\% N$_2$ + 22\% O$_2$. The last composition is aimed to represent the dry air. A brief review of works that study the effects of incorporation of bulk viscosity in analysis of various flow situations has also been provided.

physics.flu-dyn↗

arXiv2022

Bulk viscosity of dilute monatomic gases

Extensive research has been carried out in the past to estimate the bulk viscosity of dense monatomic fluids; however, little attention has been paid to estimate the same in the dilute gas regime. In this work, we perform precise Green-Kubo calculations in molecular dynamics simulations to estimate the bulk viscosity of dilute argon gas. The investigated temperature and pressure range is 300 to 750 K and 0.5 to 1.5 bar respectively. It is observed that the estimated bulk viscosity is $O(10^{-10})$ Pa s, which is several orders of magnitude smaller than that of other diatomic and polyatomic gases, but nonetheless, not an absolute zero as typically assumed. It implies that Stokes' hypothesis is true for dilute monatomic gases only in an approximate rather than absolute sense. The variation of bulk viscosity with pressures at constant temperatures has also been studied and is found to be of quadratic nature. The obtained bulk viscosity values are also reported in the reduced Lennard-Jones units to enable extension of the present results to other noble gases as well. It has also been observed that the bulk viscosity of Lennard-Jones monatomic gases becomes temperature independent at very low densities.

physics.chem-ph↗

arXiv2022

GraphMapper: Efficient Visual Navigation by Scene Graph Generation

Understanding the geometric relationships between objects in a scene is a core capability in enabling both humans and autonomous agents to navigate in new environments. A sparse, unified representation of the scene topology will allow agents to act efficiently to move through their environment, communicate the environment state with others, and utilize the representation for diverse downstream tasks. To this end, we propose a method to train an autonomous agent to learn to accumulate a 3D scene graph representation of its environment by simultaneously learning to navigate through said environment. We demonstrate that our approach, GraphMapper, enables the learning of effective navigation policies through fewer interactions with the environment than vision-based systems alone. Further, we show that GraphMapper can act as a modular scene encoder to operate alongside existing Learning-based solutions to not only increase navigational efficiency but also generate intermediate scene representations that are useful for other future tasks.

cs.CV↗