On Friday, February 7, 2014 7:26:24 AM UTC-5, Felix wrote:
> true; they should, but I guess to comply with the IEEE 754 standard > http://en.wikipedia.org/wiki/IEEE_754-1985 > so you can go for 64 or 32, they could have build in Float to > detect the cpu arch type but there is a reason for not > doing so. > There is no reason for doing so. The size of addresses on your machine, or the widest hardware integer type, has nothing whatsoever to do with the width of floating-point types. It also has nothing to do with the IEEE-754 standard (which specifies the format and behavior of 32- and 64-bit floating-point types, but not whether they are implemented in hardware or what they are called in computer languages). This a common myth of 64-bit vs. 32-bit hardware. I can't count the number of times I've heard the false statement that "32-bit machines had no 64-bit registers [false], so 64-bit machines will be faster at double-precision arithmetic [false]." Almost as common as the old "64-bit machines process data twice as quickly as 32-bit machines" canard. But again, this is just general information about computer architectures, nothing to do with Julia specifically.
