Oh dear, sounds like everyone has an opinion, not necessarily well informed
though.
Here are some facts/opinions.
1. If speed is an issue, you should use hardware floats.
2. On some (most?) processors, changing the rounding mode is more expensive
than doing arithmetic. Sometimes MUCH more expensi
On 2013-10-29 15:28, Jori Mantysalo wrote:
On Tue, 29 Oct 2013, Vincent Delecroix wrote:
all rounding are implemented in the CPU (is that true ? perhaps
changing the rounding often makes it slower). Do you have timings ?
Seems to be more complicated than just slow-or-fast -question. See
http:
On Tue, 29 Oct 2013, Vincent Delecroix wrote:
all rounding are implemented in the CPU (is that true ? perhaps changing
the rounding often makes it slower). Do you have timings ?
Seems to be more complicated than just slow-or-fast -question. See
http://www.intel.co.uk/content/dam/doc/manual/64
On 2013-10-29 12:39, Volker Braun wrote:
This is called ball arithmetic and was jut implemented in FLINT by
Fredrik Johannson,
e.g. http://fredrikj.net/blog/2012/04/high-precision-ball-arithmetic/. I
talked to him at the recent LMFDB workshop in Bristol and he thinks that
it is ready for producti
2013/10/29, Jean-Pierre Flori :
>
>
> On Tuesday, October 29, 2013 1:48:55 PM UTC+1, vdelecroix wrote:
>>
>> I do not see why RIF might be slower than RR. By construction, RIF is
>> a pair of real numbers. For arithmetic involving only +,-,*,/ we
>> should get a ratio of 2 for the execution time be
On Tuesday, October 29, 2013 1:48:55 PM UTC+1, vdelecroix wrote:
>
> I do not see why RIF might be slower than RR. By construction, RIF is
> a pair of real numbers. For arithmetic involving only +,-,*,/ we
> should get a ratio of 2 for the execution time because all rounding
> are implemented
I do not see why RIF might be slower than RR. By construction, RIF is
a pair of real numbers. For arithmetic involving only +,-,*,/ we
should get a ratio of 2 for the execution time because all rounding
are implemented in the CPU (is that true ? perhaps changing the
rounding often makes it slower).
This is called ball arithmetic and was jut implemented in FLINT by Fredrik
Johannson,
e.g. http://fredrikj.net/blog/2012/04/high-precision-ball-arithmetic/. I
talked to him at the recent LMFDB workshop in Bristol and he thinks that it
is ready for production. We should definitely wrap it ;-)
