How to Determine Highest Precision Available on a PC ??
in [Fortran]
|
Prev: Programmer's Paradise
Next: Reading a .wav file
From: glen herrmannsfeldt on 10 Aug 2010 22:45 monir <monirg(a)rogers.com> wrote: (snip) > I looked at the G95 Manual (2007-1-1), and couldn't find any ref to > "qp". > Also, under "Key G95 Features" on page 2, it reads: > ... REAL(KIND=10) for x86-compatible systems. 19 digits of precision, > value range 10�4931 > But as I showed in my previous reply, kind=10 is associated with 18 > (not 19) decimal digits w g95. It has a 64 bit significand, which works out to about 19.2 decimal digits. But remember that it rounds up to the next higher kind. Double is 53 bits, which comes out as about 15.954 decimal digits, which some might round to 16, but not quite. In any case, the temporary real format is the same from the Intel 8087 from about 1985 through current processors. > Will re-examine more thoroughly in the morning. -- glen
From: Richard Maine on 11 Aug 2010 00:25 e p chandler <epc8(a)juno.com> wrote: > I find it quite hard to believe > that there is some soft of problem that has some basis in reality in > the physical world where you really do need 80 digits. I don't find it hard to believe. I've heard of such problems. However... It is also my observation that the large majority of people asking for such precision are not actually working with the rare problems that actually justify it, but instead are fruitlessly hoping that throwing precision at their ill-formulated problem will solve it. I'm sure that the problems that justify the usage exist. I just observe that those real problems are far outnumbered by people clueless in th ematter. -- Richard Maine | Good judgment comes from experience; email: last name at domain . net | experience comes from bad judgment. domain: summertriangle | -- Mark Twain
From: mecej4 on 11 Aug 2010 08:15 Richard Maine wrote: > > It is also my observation that the large majority of people asking for > such precision are not actually working with the rare problems that > actually justify it, but instead are fruitlessly hoping that throwing > precision at their ill-formulated problem will solve it. > > I'm sure that the problems that justify the usage exist. I just observe > that those real problems are far outnumbered by people clueless in th > ematter. > We can go further and state an extension of what you said: If someone has to ask on C.L.F. how to obtain "the highest precision regardless of any ramifications that arise from...", that person probably is already off the track. Somewhat similar to "If you need to ask the price of this yacht, you cannot afford it." Throughout this long thread I have not yet seen any statement of the context where and why such precision is deemed necessary. Someone with a real need for high precision hardware capability would probably do better by first looking up the CPU specs, perhaps the CPUID instruction (at least on X86/X64), /proc/cpuinfo (on Linux/Unix). Once the CPU capabilities are known and deemed adequate one can look for a compiler/library combination that supports those CPU capabilities that are needed. -- mecej4
From: Richard Maine on 11 Aug 2010 11:40
mecej4 <mecej4.nyetspam(a)opFeramail.com> wrote: > We can go further and state an extension of what you said: If someone has to > ask on C.L.F. how to obtain "the highest precision regardless of any > ramifications that arise from...", that person probably is already off the > track. Somewhat similar to "If you need to ask the price of this yacht, you > cannot afford it." Yes. I had even thought about phrasing it somewhat along those lines. :-) -- Richard Maine | Good judgment comes from experience; email: last name at domain . net | experience comes from bad judgment. domain: summertriangle | -- Mark Twain |