xbox: Implement nexttoward() #32
Conversation
| @@ -1,20 +1,108 @@ | |||
| #include <math.h> | |||
| #include <assert.h> | |||
|
|
|||
| // Adapted from public domain code by Danny Smith <[email protected]> | |||
There was a problem hiding this comment.
nit: I'd put this comment in the first line, so it's clear that this affects the entire file, and not just this function (even though the blank line hints at this).
| } | ||
|
|
||
| long double nexttowardl(long double x, long double y) | ||
| { | ||
| assert(0); // Not implemented | ||
| return 0.0; | ||
| static_assert(sizeof(double) == sizeof(long double), "This function assumes that long double and double are the same size."); |
There was a problem hiding this comment.
nit: I probably prefer a shorter message like "long double and double must be the same size".
There was a problem hiding this comment.
Wow, I managed to miss your comment earlier. It's changed now. I also made nexttowardl simply call nexttoward, it just was the same code anyway, so there's no reason to duplicate it.
|
Code looks good, builds without warnings. I haven't tested the functionality myself but I trust cygwin. |
|
Thanks @thrimbor I removed that work-around in Peanut-GB and was successfully able to compile and run with this change |
|
I've since tested with Peanut-gb too. Works as expected.👍 |
I saw that @Ryzee119's Peanut-GB port works around our lack of
nexttowardf, so I decided to quickly add an implementation. The code is taken from a public domain implementation by Danny Smith and slightly adapted (mostly indentation and curly braces changes).Even though the code comes from Cygwin and I expect it to be tested, I ran a few quick tests against glibc, using positive and negative numbers as well as combinations with
NANandINFINITY, and the results matched.