Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: Motion 52 -- Time to revive the "Expression" subgroup?



On 2013-11-21 15:34:39 -0500, Ian McIntosh wrote:
> An fma in an expression often makes the expression more precise but can
> also make it less precise.  This is most obvious when doing a subtract of
> "equal" values, but the problem can occur with an add of other values.
> 
> In the case of evaluating the floating point expression a*b - a*b you might
> expect the result to be zero.

This is what my SIESTE slides (URL given in the message you replied to)
show. Or see:

  http://www.vinc17.net/research/fptest.en.html#contract2fma

Contraction of floating expressions still allows the compiler to
generate a fma in this case.

-- 
Vincent Lefèvre <vincent@xxxxxxxxxx> - Web: <http://www.vinc17.net/>
100% accessible validated (X)HTML - Blog: <http://www.vinc17.net/blog/>
Work: CR INRIA - computer arithmetic / AriC project (LIP, ENS-Lyon)