Mixed precision block fused multiply-add: error analysis and application to GPU tensor cores
DOI10.1137/19M1289546zbMATH Open1452.65425MaRDI QIDQ3300847FDOQ3300847
Authors: Pierre Blanchard, Nicholas J. Higham, Florent Lopez, Theo A. Mary, Srikara Pranesh 
Publication date: 30 July 2020
Published in: SIAM Journal on Scientific Computing (Search for Journal in Brave)
Recommendations
- Matrix Multiplication in Multiword Arithmetic: Error Analysis and Application to GPU Tensor Cores
- Rounding error analysis of mixed precision block Householder QR algorithms
- Performance and energy consumption of accurate and mixed-precision linear algebra kernels on GPUs
- Mixed-precision iterative refinement using tensor cores on GPUs to accelerate solution of linear systems
- Mixed precision algorithms in numerical linear algebra
Direct numerical methods for linear systems and matrix inversion (65F05) Preconditioners for iterative methods (65F08) Roundoff error (65G50) Numerical algorithms for specific classes of architectures (65Y10)
Cites Work
- Accuracy and Stability of Numerical Algorithms
- Verification methods: rigorous results using floating-point arithmetic
- The Arithmetic of the Digital Computer: A New Approach
- A New Approach to Probabilistic Rounding Error Analysis
- A new analysis of iterative refinement and its application to accurate solution of ill-conditioned sparse linear systems
- Accelerating the solution of linear systems by iterative refinement in three precisions
- Systolic super summation
- A Class of Fast and Accurate Summation Algorithms
- Squeezing a Matrix into Half Precision, with an Application to Solving Linear Systems
- IEEE754 Precision- k base-β Arithmetic Inherited by Precision- m Base-β Arithmetic for k < m
Cited In (12)
- Mixed precision algorithms in numerical linear algebra
- Mixed-precision iterative refinement using tensor cores on GPUs to accelerate solution of linear systems
- Mixed-precision explicit stabilized Runge-Kutta methods for single- and multi-scale differential equations
- Numerical algorithms for high-performance computational science
- Matrix Multiplication in Multiword Arithmetic: Error Analysis and Application to GPU Tensor Cores
- Double precision is not necessary for LSQR for solving discrete linear ill-posed problems
- Numerical stability of algorithms at extreme scale and low precisions
- Sharper probabilistic backward error analysis for basic linear algebra kernels with random data
- Exploiting lower precision arithmetic in solving symmetric positive definite linear systems and least squares problems
- Rounding error analysis of mixed precision block Householder QR algorithms
- Performance and energy consumption of accurate and mixed-precision linear algebra kernels on GPUs
- Rigorous floating-point mixed-precision tuning
Uses Software
This page was built for publication: Mixed precision block fused multiply-add: error analysis and application to GPU tensor cores
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3300847)
