[9fans] c99 div and mod round toward zero

[9fans] c99 div and mod round toward zero

[erik quanstrom]

according to /sys/src/cmd/cc/c99:

	Not done (yet?):
	[...]
	25. Division and mod truncate toward zero.

on 8c this works as expected. i.e.

	; cat mod.c
	#include <u.h>
	#include <libc.h>

	void
	main(void)
	{
		fprint(2, "-22/7 = %d\n", -22/7);
		fprint(2, "-22%%7 = %d\n", -22%7);
	}
	; 8c mod.c && 8l mod.8 && 8.out
	-22  7 = -3
	-22 % 7 = -1

is this fixed, or an archiitechture-dependent thing?  i don't
have a way of checking anything else.

- erik

Re: [9fans] c99 div and mod round toward zero

[Russ Cox]

> is this fixed, or an archiitechture-dependent thing?  i don't
> have a way of checking anything else.

traditionally it is architecture+compiler dependent.
off the top of my head i don't know what the other
compilers do.

russ

Re: [9fans] c99 div and mod round toward zero

[geoff]

I took Erik's question to be whether 8c rounded to zero because the
x86 (or 8c) does it that way or because the common C compiler code had
been changed to force C99 behaviour.

Re: [9fans] c99 div and mod round toward zero

[Russ Cox]

> I took Erik's question to be whether 8c rounded to zero because the
> x86 (or 8c) does it that way or because the common C compiler code had
> been changed to force C99 behaviour.

the former.

Re: [9fans] c99 div and mod round toward zero

[dmr]

 > is this [rounding to 0] fixed, or an archiitechture-dependent thing?  i don't
 > have a way of checking anything else.

In C89 it was allowed to vary by architecture (subject to obvious consistency
constraints).  In C99 the result is well-determined, and does round to 0.
This is consistent with the usual math convention, but
in fact not necessarily best (e.g. for graphics, as Rob has pointed out).

	Dennis