Re^14: Math::BigFloat to native double?

[BrowserUk]

Thanks syphilis. That's beyond the call.

I'm still digesting Re^12: Math::BigFloat to native double?; but I took "But I'm pretty much out of ideas (and energy) if Math::MPFR is to be excluded. (Without it, there's just too many hoops for me)." at face value, so this is most unexpected :) Thank you.

"(I'll scrutinise it and test it over the next week or so.)" But, please don't expend more of your energies on this. It's just a background project for me; something I do whilst I'm waiting for other stuff to run; or when I hit a mental block in my current work and need a diversion to allow my brain to unencumber itself of the details.

Given where I started with the simple question in the OP, this thread's has gone way beyond anything I could have expected anyone else to give a damn about :)

[syphilis]

Thing is that, with re-charged batteries, I start thinking "Hang on, I *can* improve on that". (And there's still room for improvement.)

Anyway - good luck, and enjoy the exercise. I find the doubledouble to be rather intriguing (not to mention a little exasperating), and I purchased the (second hand) PPC box for the sole purpose of looking more readily at it.
You should be able to pick one up over there for next-to-nothing if you ever get the urge. I had one offered to me for free - it was a genuine offer, too, but would've required a trip to New York to pick it up.
Being big-endian, the PPC box also provides a handy reference to that architecture, too.

Cheers,
Rob

[BrowserUk]

I find the doubledouble to be rather intriguing (not to mention a little exasperating), and I purchased the (second hand) PPC box for the sole purpose of looking more readily at it.

I'm gonna ask this again: are you sure that we are talking about the same thing with respect to doubledouble and PPC?

When I google "PPC doubledouble" (and click past its annoying habit of trying to be helpful instead of searching for what I ask for), the *only* hits I get are those that also include (one of) your handles. Here, or on active state forums etc.

But, when I look at the C code there, I see "long double"; which according to wikipedia: MS map to double; Intel map to 80-bit IEEE (hardware) extended precision; and gcc maps to 128-bit IEEE quad precision. None of which are the doubledouble I'm talking of.

[syphilis]

are you sure that we are talking about the same thing with respect to doubledouble and PPC

Yes - though "being sure" doesn't guarantee that I'm right ;-)

The wikipedia "long double" page mentions a number of ways of implementing long doubles, including the double double:

"On some PowerPC and SPARCv9 machines, long double is implemented as a double-double arithmetic, where a long double value is regarded as the exact sum of two double-precision values, giving at least a 106-bit precision; with such a format, the long double type does not conform to the IEEE floating-point standard."

Also this Wikipedia "double-double arithmetic" entry accurately describes the PPC double-double (and, I think, the double-double that you are wanting to implement).

Cheers,
Rob