Re: [PATCH v4 1/3] math.h: add DIV_ROUND_UP_NO_OVERFLOW

[Linus Torvalds]

[ Since you're cc'ing the s390 people, I assume that means that this
all ended up being a follow-up to the s390 issue discussion ]

On Tue, 24 Oct 2023 at 06:19, Sebastian Reichel
<sebastian.reichel@xxxxxxxxxxxxx> wrote:
>
> Add a new DIV_ROUND_UP helper, which cannot overflow when
> big numbers are being used.

This is really horrendously bad on some architectures (ie it might
require *two* divisions):

> +#define DIV_ROUND_UP_NO_OVERFLOW(n, d) (((n) / (d)) + !!((n) % (d)))

but David Laight at least had a suggestion for that: when allowing
multiple uses, you can just do

   #define DIV_ROUND_UP(n,d) ((n) ? ((n)-1)/(d)+1 : 0)

so now you're back to just one division and no horrible extra expense.

But we can't allow those multiple uses in general, sadly.

I would really prefer to just make our regular DIV_ROUND_UP() DTRT.  But:

 - people do use it with complex first arguments (ie function calls
etc) that we don't want to evaluate twice

 - we can't make it an inline function, because the types aren't fixed

 - we can't even use a statement expression and __auto_type, because
these things are used in type definitions etc and need to be constant
expressions

That last thing means that even the "__builtin_choose_expr()" doesn't
work, because the statement expression still needs to be _parsed_ as
such, and that's not something we can do in those contexts.

Very annoying. Let me think about this.

                     Linus