Search arXiv⌕ Search
Home/Archive/arXiv

arXiv · 2609.35844

A Branch-Free General Renormalization Scheme for Pair Arithmetic and Its Performance Evaluation

Tomonori Kouya
View original source ↗Read source document ↗Export metadata ↓

Abstract

Pair arithmetic omits the renormalization stage performed at the end of each operation of multi-component multiple-precision arithmetic, reducing the operation count and, by eliminating conditional branches, easing SIMD vectorization. Renormalization cannot be omitted entirely, however, because iterative solvers may then fail to converge. We propose RenormBF-pair, a branch-free fixed-trip-count renormalizer parameterized by the word count \(K\) and the input length \(n\): a VecSum sweep, a tail fold, and \(r\) rounds of a FastTwoSum chain, costing \(6(n-1)+(n-K)+3r(K-1)\) flops. Over \(200{,}000\) trials per condition, the smallest round count producing no non-overlap violation was \(r=1\) for \(K \leq 3\) and \(r=2\) for \(K=4\), whereas the same cost spent on a VecSum\(K\)-style TwoSum chain fails for \(K=4\). We delimit with explicit counterexamples both the input condition under which the sum relation holds and the fact that the scheme does not, in general, guarantee a correctly rounded lowest component. Integrating it into CG and BiCGStab over pair arithmetic in binary64 and binary32, we sweep \(240\) paired runs including real matrices screened from SuiteSparse. The proposed scheme and VecSum\(K\) returned bitwise identical words wherever both succeeded, so they are indistinguishable for CG in binary64; differences appear only where VecSum\(K\) breaks, namely binary32 and BiCGStab. There the \(22\)--\(16\) win--loss record is not significant (sign test, \(p=0.42\)); the clear bias is structural, non-overlap violations dropping from \(66\) rows to \(37\). The cost is \(+3.4\%\) for the whole CG solver.

Explore related subjects

math.NAcs.NA

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Tomonori Kouya. 2026-09-25. A Branch-Free General Renormalization Scheme for Pair Arithmetic and Its Performance Evaluation. https://arxiv.org/abs/2609.35844

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

KEEP EXPLORING

Related papers

arXiv2026

Error Estimates for the Arnoldi Approximation of a Matrix Square Root

The Arnoldi process provides an efficient framework for approximating functions of a matrix applied to a vector, i.e., of the form $f(M)\bm{b}$, by repeated matrix-vector multiplications. In this paper, we derive error estimates for approximating the action of a matrix square root using the Arnoldi process, where the integral representation of the error is reformulated in terms of the error for solving the linear system $M\bm{x}=\bm{b}$. The results extend the error analysis of the Lanczos method for Hermitian matrices in [Chen et al., SIAM J. Matrix Anal. Appl., 2022] to non-Hermitian cases and provide an improved bound for the Hermitian case. Furthermore, in practical settings, the matrix may only be available via approximate or structured representations. Motivated by this, we extend the analysis and establish a generalized error bound for perturbed matrices. The numerical results on matrices with different structures demonstrate that our theoretical analysis yields a reliable upper bound. Finally, simulations on large-scale matrices arising in particulate suspensions, represented in hierarchical matrix form, validate the effectiveness and practicality of the approach.

math.NA↗
arXiv2026

Study of a TPFA scheme for the stochastic Allen-Cahn problem with constraint through numerical experiments

We study the properties of a fully discrete numerical scheme for the stochastic Allen-Cahn problem with constraint on a bounded polygonal domain in two or three dimensions with homogeneous Neumann boundary condition. The scheme under consideration is of Two Point Flux Approximation (TPFA) type with respect to space and of semi-implicit Euler-Maruyama type with respect to time. The constraint is implemented by a multivalued, subdifferential operator which makes the problem a differential inclusion. This operator is incorporated into the scheme via its Yosida approximation, at the expense of introducing an additional regularization parameter. From previous studies it is known that the time discretization of the Yosida approximation has to be implicit in order to obtain a convergent scheme. Consequently, a non-linear and non-smooth equation has to be solved at each time step and this is challenging for the computation of approximate solutions. In this contribution, we introduce a splitting method to compute the solutions of the TPFA approximations. This method allows us to solve a linear equation in a first step. Then, a piecewise affine projection operator coming from the Yosida approximation can be computed explicitly in a second step. We quantify the error coming from the splitting method and show that the splitting method is accurate. We study properties of the scheme and provide convergence rates through numerical experiments.

math.NA↗
arXiv2026

Spectral density estimation for normal matrices

The spectral density estimation problem asks for an algorithm that, given an $n\times n$ matrix $A$, outputs a probability measure that is a good approximation to the uniform distribution on the eigenvalues of $A$, called the spectral density of $A$. This paper considers the setting where $A$ is a large normal matrix that is accessible only through matrix-vector product queries. We provide an algorithm that makes just $m$ matrix-vector queries to $A$ and returns, with high probability, a measure within earth mover's distance $O(1/m+\log m/{\sqrt n})$ of the true spectral density of $A$. We provide a complementary lower bound that any algorithm producing an $\varepsilon$-approximation to the true spectral density for large matrices must make $Ω(1/\varepsilon)$ matrix-vector queries. The lower bound holds even for the more restricted case of real symmetric input matrices. In combination with our upper bound, it shows that spectral density estimation is essentially no harder for complex normal matrices than for real symmetric matrices.

math.NA↗