The increasing availability of very low precisions (tfloat32, fp16, bfloat16, fp8) in hardware pushes modern high performance computing to embrace mixed precision standards. By employing mostly low precision and by making wise use of high precision, mixed precision algorithms can leverage the low precision advantages while preserving the quality of the computed solution. Mixed precision iterative refinement is one of the oldest and most famous representatives of these algorithms; this method was shown to be very effective in reducing the resource consumption of linear solvers while delivering accurate solutions in a robust way.This thesis is dedicated to investigating the use of this algorithm for the solution of large sparse linear systems...
3rd International Workshop on Energy Efficient Supercomputing (E2SC '15)We formulate an implementati...
In this PhD thesis, we address three challenges faced by linear algebra solvers in the perspective o...
AbstractIn this paper we compare two recently proposed methods, FGMRES (Saad, 1993) and GMRESR (van ...
L'accessibilité grandissante des arithmétiques à précision faible (tfloat32, fp16, bfloat16, fp8) da...
International audienceThe standard LU factorization-based solution process for linear systems can be...
Hardware trends have motivated the development of mixed precision algo-rithms in numerical linear al...
summary:With the emergence of mixed precision hardware, mixed precision GMRES-based iterative refine...
With the commercial availability of mixed precision hardware, mixed precision GMRES-based iterative ...
Iterative refinement is a long-standing technique for improving the accuracy of a computed solution ...
Motivated by the demand in machine learning, modern computer hardware is increas- ingly supporting r...
International audienceScientific applications very often rely on solving one or more linear systems....
It is well established that reduced precision arithmetic can be exploited to accelerate the solution...
On many current and emerging computing architectures, single-precision calculations are at least twi...
It is well established that mixed precision algorithms that factorize a matrix at a precision lower...
Mixed-precision algorithms are a class of algorithms that uses low precision in part of the algorith...
3rd International Workshop on Energy Efficient Supercomputing (E2SC '15)We formulate an implementati...
In this PhD thesis, we address three challenges faced by linear algebra solvers in the perspective o...
AbstractIn this paper we compare two recently proposed methods, FGMRES (Saad, 1993) and GMRESR (van ...
L'accessibilité grandissante des arithmétiques à précision faible (tfloat32, fp16, bfloat16, fp8) da...
International audienceThe standard LU factorization-based solution process for linear systems can be...
Hardware trends have motivated the development of mixed precision algo-rithms in numerical linear al...
summary:With the emergence of mixed precision hardware, mixed precision GMRES-based iterative refine...
With the commercial availability of mixed precision hardware, mixed precision GMRES-based iterative ...
Iterative refinement is a long-standing technique for improving the accuracy of a computed solution ...
Motivated by the demand in machine learning, modern computer hardware is increas- ingly supporting r...
International audienceScientific applications very often rely on solving one or more linear systems....
It is well established that reduced precision arithmetic can be exploited to accelerate the solution...
On many current and emerging computing architectures, single-precision calculations are at least twi...
It is well established that mixed precision algorithms that factorize a matrix at a precision lower...
Mixed-precision algorithms are a class of algorithms that uses low precision in part of the algorith...
3rd International Workshop on Energy Efficient Supercomputing (E2SC '15)We formulate an implementati...
In this PhD thesis, we address three challenges faced by linear algebra solvers in the perspective o...
AbstractIn this paper we compare two recently proposed methods, FGMRES (Saad, 1993) and GMRESR (van ...