posithub.org abandonned? Twitter handle doesn't exit

[Tommy Thorn]

I went to posithub.org to find the latest news, but it doesn't appear updated in the past 3-4 years and the linked Twitter handle "posithub_org" doesn't exist on Twitter.

Anyone know what the status is there?

Thanks,

Tommy

[John Gustafson]

Hi Tommy,

Yesterday I went on GoDaddy to see about accessing posithub.org, and they had (without my permission) imposed a two-factor authentication after my login, by sending a code to "c*****@posithub.org". I assume that was Cerlane Leong's email, and she switched the site over to me years ago.

I vaguely recall that we changed the Twitter handle "posithub_org" to just "posithub" but it sounds like something didn't get updated correctly.

I had a volunteer webmaster, a post-doc in Singapore, but he does not seem to be answering emails. We provided access to another volunteer, but I now know that he was similarly defeated by the two-factor authentication scheme.

GoDaddy said I need to apply to remove the 2FA with a form. I filled out the form, and they say it will take three days to process. Once they do that, I plan to (at least) put a headline on the home page that the Posit Standard has been ratified, as of March 2. We plan to post it on arXiv so it becomes a citable reference, finally. It was 5 years in the making, and the Posit Working Group did some amazing work.

If you go to the Wikipedia page for "Unum (number format)" and scroll down to Implementations, the table has 28 entries for the efforts going on worldwide. Instead of trying to keep the table up to date on posithub.org, I encourage people to make edits and additions to that Wikipedia table, per the usual Wikipedia editing rules. Then posithub.org can simply point to that article, and there is no cache-incoherency.

The final, ratified version of the Posit Standard has significant improvements over the earlier versions I have sent to the Unum Computing group as attachments; I'll attach the final version here.

Thanks for raising this issue, Tommy.

John G.

[Tommy Thorn]

In Section 6.4 we read "In converting an integer value to a posit value, the integer representation with its MSB = 1 and all other bits 0 converts to NaR; otherwise, the integer value is rounded, per Section 4.1."

So converting a, say unsigned 8-bit value (uint8_t), all values 0-127,129-255 maps to their expected values, but 128 maps to a NaR?

That seems less than ideal.  At the very least there should be two different notions of integer -> posit conversion.

Tommy

PS: Apologies for misspelling the quire.

[John Gustafson]

Notice that with floats, NaN does not always get propagated. If you convert a (quiet) NaN to an integer, you'll get an integer that has no indication that anything went wrong; convert that integer back to a float and you'll get a real, not a NaN. Converting an integer to a real value indicates the integer is being used as a subset of the reals and not as an enumerator or address.

If you really want to convert unsigned 128 to 128,0 just store the 128 in a 16-bit integer before converting it.

If you have signed 8-bit integers, what happens when you take the absolute value of 10000000 = –128? You get 10000000 = –128. That seems less than ideal, too. Try to negate 10000000 and you'll get back 10000000, still negative. That bit pattern, at any precision, behaves like an error value.

So there is an easy workaround for unsigned integers, but I see no easy way to assure that Indeterminate (NaR) always propagates to a final answer, unless you do it the way the Standard says.

John

--
You received this message because you are subscribed to the Google Groups "Unum Computing" group.
To unsubscribe from this group and stop receiving emails from it, send an email to unum-computin...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/unum-computing/78c16123-dcc4-4869-86f6-d7dd00355c84n%40googlegroups.com.