[R] Installing R on RedHat7.0

[Peter Dalgaard BSA]

Thomas Lumley <thomas at biostat.washington.edu> writes:

> As the value is 2^-63 it may be that the machine precision calculation is
> being done entirely internally to the chip, using the 80-bit extended
> double format, which has this level of precision.
> 
> This is certainly wrong and new; I don't know enough about the machine
> precision calculations to know if it is really a compiler bug or whether
> it is something the compiler was previously allowed to do but didn't.

Might well be that we need to be more careful about compiler switches
for machar.c. But the relevant computation seems to be

        for(;;) {
		temp = one + a;
		if (temp - one != zero)
			break;
		a = a * beta;
		*machep = *machep + 1;
	}

i.e. decrease a until adding it to one does not change it. However, a,
one, and temp are all declared volatile, and that would normally
ensure that they are stored off-chip as ordinary 64-bit doubles. I
don't see how the compiler can treat them as 80-bit extended doubles
while still respecting the volatility. Unless it stores them to memory
as long longs...

-- 
   O__  ---- Peter Dalgaard             Blegdamsvej 3  
  c/ /'_ --- Dept. of Biostatistics     2200 Cph. N   
 (*) \(*) -- University of Copenhagen   Denmark      Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)             FAX: (+45) 35327907
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._