1

Topic: Re: About conversion of number from IEEE-754 in the text

> Though I too was surprised, why among set of commands FPU there is no explicit transfer in scientific the notation in the form of a text line. It is possible to save from the register in a binary-decimal format (only whole) And from it it is already trivial about characters to translate. > article, is written two and a half a year back: > http://files.rsdn.org/122727/pl1ex15.doc Thanks.

2

Re: Re: About conversion of number from IEEE-754 in the text

3

Re: Re: About conversion of number from IEEE-754 in the text

Hello, , you wrote: > In my opinion for log this article not so. More likely for students that they understood and mastered commands FPU. > Though I too was surprised, why among set of commands FPU there is no explicit transfer in scientific the notation in the form of a text line. As far as I understand, the task not so is simple, as it seems at first sight. > article, is written two and a half a year back: > http://files.rsdn.org/122727/pl1ex15.doc > I Remind that the author not I, and already printed here Loaves http://rsdn.org/users/99493.aspx Is a pity that the author did not compare speed of the implementation with existing Dragon4 or Grisu3.

4

Re: Re: About conversion of number from IEEE-754 in the text

Hello, , you wrote: > There are numbers representable in the form of such fraction, but the majority - is not present. Not representable in the form of finite decimal fraction at all it is less than numbers.

5

Re: Re: About conversion of number from IEEE-754 in the text

Hello, , you wrote: > Hello, PM, you wrote: PM>> It is a pity that the author did not compare speed of the implementation with existing Dragon4 or Grisu3. > Really problem uneasy, appears. What works write > https://333341ac-a-698ebdfa-s-sites.goo … edirects=0 the Suspicious any link, did not open. But how much I understood, it is the link to operation Florian Loitsch, author Grisu3. Yes, the task not such simple because, , Dragon4 is already 4th version of the algorithm widely used in functions of type printf () since 80th years (http://www.netlib.org/fp/dtoa.c) Accordingly, Grisu3 - newer and fast algorithm well working in ~95 % cases and using Dragon4 in remaining 5 %. As far as I know, it is implemented by the author and it is used in  V8: https://github.com/google/double-conversion > But the author like wanted as much as possible FPU to involve. And in those algorithms FPU it is explicit as though it is not applied. However, for normal float (i.e. 24 discharges of a mantissa), probably, without FPU will be faster. Can will, and can and is not present. Without comparing with existing implementations (at least sprintf) there is especially nothing to consider. Simply one more bicycle specific for x86, and not the fact that correctly working. Here for example, the person tried, made comparing: https://github.com/miloyip/dtoa-benchmark Or here people, trying to overcome lacks Grisu3, compare the method to others: https://cseweb.ucsd.edu/%7Emandrysc/pub/dtoa.pdf