C++ Logo

std-proposals

Advanced search

Re: [std-proposals] Freestanding std modules

From: Marcin Jaczewski <marcinjaczewski86_at_[hidden]>
Date: Sun, 30 Jun 2024 20:21:29 +0200
niedz., 30 cze 2024 o 20:01 Ville Voutilainen
<ville.voutilainen_at_[hidden]> napisaƂ(a):
>
> On Sun, 30 Jun 2024 at 20:54, Marcin Jaczewski
> <marcinjaczewski86_at_[hidden]> wrote:
> > Probably the best mental model of how modules work is "object files"
> > as Ville said,
> > instead of classic `#include` files. This mean would be better to have
> > few big modules
> > than many small form usage perspective as the overall final binary
> > should be similar.
> > (this would not even be problem with recompiation as we have module partitions)
>
> The size of the module has nothing to do with the size of your
> program. And imports of one big
> module are splendidly fast, and in many cases faster than importing
> multiple smaller modules.
> Such results have been measured with prototype implementations of One
> Big Standard Module
> and multiple smaller standard modules, for instance, when Microsoft
> prototyped such options
> with their implementation.
>

Yes, exactly this I had in mind but probably
not spell out that clearly like you.

> > > > There is no issue with any compiler here, nor with the design of C++
> > > > modules. The sender of the email that started this whole discussion is
> > > > unhinged.
> > >
> > > To clarify, I mean the email sent to the gcc list. :)
> >
> > @Ville uff... I was start worrying that I was this person :D
>
> My apologies for the momentary potential for confusion; I was indeed
> referring to the person who
> posted the message on the gcc mailing list, that person has a
> multi-year history of posting garbage
> like that both on gcc mailing lists and gcc bugzilla, and many other
> forums, including this one.
> Due to that person's use of multiple email addresses to post such
> garbage, banning that garbage
> is a game of whack-a-mole.

I know, but many times it could be a kernel of truth in their grievances.
Like one of them calls GCC "bad" because in some specific cases
it does not generate the best possible assembly.
This means it is a real problem if you want to have 100% performance,
but how they react to it is not productive to fixing that.
Not mentioning that in engineering there are tradeoffs, if you fix
code gen in one place it could become worse in diffrent place.

Again thanks for answers that clarified a bit more how modules will work.

Received on 2024-06-30 18:21:48