I am programming the most repetitive stuff (except for the part where I
copy/paste into julia to use her parsing algs and then recopy) in Maple --
eventually it would be good to be able to, in the special cases where it is
helpful, macro-gobble whatever is on the line and then pass it along.
munchymacro? And that should fall out nicely from far more substantial
evolution. Whenever.
Re correct rounding: if by correct you mean there is no result that more
than 0.5ulp from the ideal result -- no. For that, crlibm is the play for
Float64 results. Should someone *("not it")* reimplement crlib in Julia and
parameterize the working float type, my stuff would just work, and maybe
crlibm takes care of the correct rounding. I am trying to have everything
be correct up through the penultimate nibble of the extended significand --
at least 102 reliable bits for non-outlandish function calls. The
restriction is throughput. An earlier version usually had the last bit
correct, but it ran slower than BigFloat so I changed my approach.
On Monday, November 2, 2015 at 9:07:28 AM UTC-5, David P. Sanders wrote:
>
>
>
> El lunes, 2 de noviembre de 2015, 7:58:17 (UTC-6), Stefan Karpinski
> escribió:
>>
>> I think we should probably make it possible to access the full string of
>> a numeric literal in a macro but that is a substantial change to the parser.
>>
>
> That would be great.
>
> Off-topic: Jeffrey, will your Float128 library be correctly-rounded?
>
>
>>
>> On Monday, November 2, 2015, David P. Sanders <[email protected]> wrote:
>>
>>>
>>>
>>> El lunes, 2 de noviembre de 2015, 6:35:46 (UTC-6), Milan Bouchet-Valat
>>> escribió:
>>>>
>>>> Le lundi 02 novembre 2015 à 00:58 -0800, Jeffrey Sarnoff a écrit :
>>>> > I have many values like
>>>> > 0.6584871727288045313850172023417636020375045372547107712468813403
>>>> > that come from Maple and I would like to avoid doing this by
>>>> > copy/paste for each one:
>>>> > Float128(parse(BigFloat,"0.658487172728804531385017202341763602037504
>>>> > 5372547107712468813403"))
>>>> >
>>>> > I tried writing a macro that would put quotes around the value and
>>>> > then affix the rest -- without good result.
>>>> > julia>@fromMaple
>>>> > 0.6584871727288045313850172023417636020375045372547107712468813403
>>>> > Float128(parse(BigFloat,"0.658487172728804531385017202341763602037504
>>>> > 5372547107712468813403"))
>>>> >
>>>> > The REPL converts the unenquoted value to a Float64 before I get at
>>>> > it.
>>>> >
>>>> > Help is appreciated.
>>>> I don't think you can work around this at the moment. The best you can
>>>> do is to define a non-standard string literal by creating a @f128_str
>>>> macro, so that you can type these numbers as
>>>> f128"0.6584871727288045313850172023417636020375045372547107712468813403
>>>> ".
>>>>
>>>
>>> There is already a `big` macro:
>>>
>>> Float128(big"0.65848717272880453138501720234176360203750453725471077124688134")
>>>
>>> Instead of copying and pasting in the REPL, couldn't you write these
>>> numbers to a file
>>> and read them in as strings in Julia?
>>>
>>> There has been some discussion in the past about forwarding strings like
>>> this to the parser
>>> already wrapped in a macro (as is done for large integer values); I
>>> don't recall what the
>>> issue with this was.
>>>
>>
