Date: Fri, 3 Mar 2023 18:40:35 +0000
On Fri, 3 Mar 2023, 18:09 Andrey Semashev via Std-Proposals, <
std-proposals_at_[hidden]> wrote:
> On 3/3/23 19:52, Jason McKesson via Std-Proposals wrote:
> > On Fri, Mar 3, 2023 at 4:53 AM Andrey Semashev via Std-Proposals
> > <std-proposals_at_[hidden]> wrote:
> >> or if `import std` is more expensive than `#include <cstddef>` (or pick
> any
> >> other lightweight header)
> >
> > We already have an answer to this. There's evidence, from the compiler
> > that supports a standard library import, that it is close to an order
> > of magnitude faster to `import std;` than to include even one header.
> > The evidence is listed in P2412.
>
> While this is better than nothing (thanks for sharing the reference to
> the proposal), this is just one implementation on a platform that is
> infamous for its slow filesystem. Given the history with precompiled
> headers, which offered a significant speedup with MSVC on Windows and
> nearly no benefit on other platforms, I'll keep my reservations until I
> see numbers on other platforms and implementations.
>
It's true for gcc too, based on testing a local patch.
> > I know there are a lot of questions about modules out there. But
> > people keep reraising these questions as if there aren't already
> > answers to at least some of them.
>
> I would say, that's because finding this kind of information is
> virtually impossible unless you follow closely the particular proposals
> or on the discussions on the MLs and reflectors.
>
> --
> Std-Proposals mailing list
> Std-Proposals_at_[hidden]
> https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals
>
std-proposals_at_[hidden]> wrote:
> On 3/3/23 19:52, Jason McKesson via Std-Proposals wrote:
> > On Fri, Mar 3, 2023 at 4:53 AM Andrey Semashev via Std-Proposals
> > <std-proposals_at_[hidden]> wrote:
> >> or if `import std` is more expensive than `#include <cstddef>` (or pick
> any
> >> other lightweight header)
> >
> > We already have an answer to this. There's evidence, from the compiler
> > that supports a standard library import, that it is close to an order
> > of magnitude faster to `import std;` than to include even one header.
> > The evidence is listed in P2412.
>
> While this is better than nothing (thanks for sharing the reference to
> the proposal), this is just one implementation on a platform that is
> infamous for its slow filesystem. Given the history with precompiled
> headers, which offered a significant speedup with MSVC on Windows and
> nearly no benefit on other platforms, I'll keep my reservations until I
> see numbers on other platforms and implementations.
>
It's true for gcc too, based on testing a local patch.
> > I know there are a lot of questions about modules out there. But
> > people keep reraising these questions as if there aren't already
> > answers to at least some of them.
>
> I would say, that's because finding this kind of information is
> virtually impossible unless you follow closely the particular proposals
> or on the discussions on the MLs and reflectors.
>
> --
> Std-Proposals mailing list
> Std-Proposals_at_[hidden]
> https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals
>
Received on 2023-03-03 18:40:51