[PATCH 02/10] compiler.h: add is_const() as a replacement of __is_constexpr()

Linus Torvalds torvalds at linux-foundation.org
Sat Dec 7 18:19:34 UTC 2024


On Sat, 7 Dec 2024 at 04:24, Vincent Mailhol <vincent.mailhol at gmail.com> wrote:
>
> > No good - expands everything twice.
>
> And? __is_const_zero() does not evaluate its arguments, so no side effect:

No, the problem is literally the expansion.

Double expansion of these fundamental helpers gets exponential,
because they are used in various nested ways in other fundamental
helpers.

That's why we then spent so much effort on trying to clean up the
min/max macros, because a single line of code would expand to
literally tens of megabytes of horrific expansions.

And the problem with these things is that you can't make them inline
functions, so they have to be macros, and then you build up other
macros using them (like that "clamp()" macro), and it really gets
horrendous and affects the build time.

And yes, it is very sad. Particularly since a compiler would have a
really easy time with some nice helper builtins.

Of course, often the compiler *does* have helper builtins, but we
can't use them, because they aren't *quite* the right thing. Like that
"__builtin_constant_p()" not actually working for some situations
where we absolutely need not just a constant value, but a constant
_expression_ due to C parsing rules.

Quite a lot of the pain we tend to have with these things is directly
related to the fact that we often want to do these tests in contexts
like global array initializers etc.

If there is one feature of C I would have liked it is "allow inline
functions and statement expressions with constant arguments as
constant expressions". Other languages have done that, and it really
does help. And yes, it means that you have to basically have a
language interpreter in the compiler (you do want to allow loop
constructions etc), but it really is very useful.

Oh well. Even if compilers added that today, it would be years until
we could take advantage of it.

At one point I literally was thinking I'd do 'sparse' as a
pre-processor for kernel code, in order to have extended language
facilities like that.

            Linus


More information about the Intel-gfx mailing list