Skip to main content
18 events
when toggle format what by license comment
Jun 30, 2017 at 9:54 answer added sathyarokz timeline score: 0
S Aug 28, 2016 at 12:28 history bounty ended RE60K
S Aug 28, 2016 at 12:28 history notice removed RE60K
Aug 28, 2016 at 12:28 vote accept RE60K
Aug 28, 2016 at 6:34 answer added FirefoxMetzger timeline score: 4
Aug 22, 2016 at 20:10 history tweeted twitter.com/StackCodeReview/status/767816236034064385
Aug 22, 2016 at 15:53 answer added DJHenjin timeline score: 2
S Aug 21, 2016 at 7:32 history bounty started RE60K
S Aug 21, 2016 at 7:32 history notice added RE60K Draw attention
Aug 19, 2016 at 6:02 comment added RE60K @Quuxplusone oh sorry.
Aug 18, 2016 at 19:32 comment added Random832 @ADG Most C/C++ implementations use IEEE754 32-bit single precision for float and IEEE754 64-bit double precision for double - implementations that do not use these representations are very rare, even among implementations that do not fully conform to the standard for operations. And *(float*)&x isn't a cast to a float, it is a "type-pun" via a pointer cast (generally in C++ you would use reinterpret_cast). You could also convert it to float or double in a portable way with ldexp[f].
Aug 18, 2016 at 17:19 comment added Random832 @ADG float, then.
Aug 18, 2016 at 14:26 history edited 200_success
edited tags
Aug 18, 2016 at 11:48 history edited ratchet freak CC BY-SA 3.0
fixed code markup
Aug 18, 2016 at 10:41 history edited RE60K CC BY-SA 3.0
added 2268 characters in body; edited title
Aug 18, 2016 at 6:46 comment added Quuxplusone You also leak memory like the dickens, and your use of macros is quite unusual for C++. Consider rewriting your macros as static inline functions instead.
Aug 18, 2016 at 6:46 comment added Quuxplusone You should provide more information about your constraints, because otherwise the clearest code would obviously be *(double*)&c = *(double*)&a + *(double*)&b; instead of all this stuff. Your test harness would also be more compelling if it compared the output of your bit-twiddling code to the naive floating-point implementation, instead of requiring the input file to contain the answer in hex.
Aug 18, 2016 at 6:23 history asked RE60K CC BY-SA 3.0