METHOD OF ADJUSTING THE PRECISION OF A COMPUTER PROGRAM MANIPULATING AT LEAST ONE FLOATING POINT NUMBER

摘要

The invention relates to a method (100) of adjusting the precision provided by a source computer program manipulating at least one floating point number, comprising the following steps: -measurement (106, 110) with initial precision, of at least one value of at least one floating point number; -measurement (108, 110), termed with infinite precision, of at least one value of said floating point number in said source program; -for at least one instruction manipulating, in said source program (200), at least one floating point number, a step (114-118), termed optimization, identifying, for said instruction, a variant providing a better precision, and -modification (120) of said source program by replacing said instruction with said selected variant. It also relates to a computer program and a system implementing such a method.