Just poking around randomly, it seems as though I get the mostly the same answers with x: and 21312313123123123123123123123x in J6 32 and 64.
I'd assume there are tests that ensure identical results (?), and this applies to all versions and all supported platforms. With large enough number, a number turns into an integer x: 1234551231231.123123122 9876409849849r8 x: 12345512312312.123123122 12345512312312 Does anyone know the rule for losing the rational part? The above is about 1.2e14, which doesn't seem to match any 32bit or 64bit boundaries (around 2^47). Also is there an integer x that is too big? There is not a hard rule against rational numbers greater than 2^47 though. x: (2^64x) - x: 1.1 184467440737095516149r10 but not for this... Actuall a different result for J6-32: x: 12345512312312 + x: 0.123123122 12345512312312 and J6-64 x: 12345512312312 + x: 0.123123122 18227766461577215791144763r1476469019710 So is there a defined range where J32 and J64 provide identical results? ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
