public inbox for libstdc++@gcc.gnu.org
 help / color / mirror / Atom feed
From: Patrick Palka <ppalka@redhat.com>
To: Patrick Palka <ppalka@redhat.com>
Cc: gcc-patches@gcc.gnu.org, libstdc++@gcc.gnu.org
Subject: Re: [PATCH 2/1] libstdc++: Non-triv-copyable extra args aren't simple [PR100940]
Date: Tue, 15 Jun 2021 15:28:23 -0400 (EDT)	[thread overview]
Message-ID: <571933d-e747-23e5-bd26-e141841aa682@idea> (raw)
In-Reply-To: <20210615185321.1082491-1-ppalka@redhat.com>

On Tue, 15 Jun 2021, Patrick Palka wrote:

> This force-enables perfect forwarding call wrapper semantics whenever
> the extra arguments of a partially applied range adaptor aren't all
> trivially copyable, so as to avoid incurring unnecessary copies of
> potentially expensive-to-copy objects (such as std::function objects)
> when invoking the adaptor.
> 
> Tested on x86_64-pc-linux-gnu, does this look OK for trunk/11?
> 
> 	PR libstdc++/100940
> 
> libstdc++-v3/ChangeLog:
> 
> 	* include/std/ranges (__adaptor::__adaptor_has_simple_extra_args): Also
> 	require that the extra arguments are trivially copyable.
> 	* testsuite/std/ranges/adaptors/100577.cc (test04): New test.
> ---
>  libstdc++-v3/include/std/ranges               |  6 ++++--
>  .../testsuite/std/ranges/adaptors/100577.cc   | 19 +++++++++++++++++++
>  2 files changed, 23 insertions(+), 2 deletions(-)
> 
> diff --git a/libstdc++-v3/include/std/ranges b/libstdc++-v3/include/std/ranges
> index 856975c6934..e858df88088 100644
> --- a/libstdc++-v3/include/std/ranges
> +++ b/libstdc++-v3/include/std/ranges
> @@ -818,8 +818,10 @@ namespace views::__adaptor
>    // True if the behavior of the range adaptor non-closure _Adaptor is
>    // independent of the value category of its extra arguments _Args.
>    template<typename _Adaptor, typename... _Args>
> -    concept __adaptor_has_simple_extra_args = _Adaptor::_S_has_simple_extra_args
> -      || _Adaptor::template _S_has_simple_extra_args<_Args...>;
> +    concept __adaptor_has_simple_extra_args
> +      = (_Adaptor::_S_has_simple_extra_args
> +	 || _Adaptor::template _S_has_simple_extra_args<_Args...>)
> +	&& (is_trivially_copyable_v<_Args> && ...);

On second thought, perhaps it'd be cleaner to leave this concept alone
and instead encode the trivial-copyability requirement as a separate
constraint on the relevant partial specializations of _Partial?
Something like:

-- >8 --


	PR libstdc++/100940

libstdc++-v3/ChangeLog:

	* include/std/ranges (__adaptor::_Partial): For the "simple"
	forwarding partial specializations, also require that
	the extra arguments are trivially copyable.
	* testsuite/std/ranges/adaptors/100577.cc (test04): New test.
---
 libstdc++-v3/include/std/ranges                    |  8 +++++---
 .../testsuite/std/ranges/adaptors/100577.cc        | 14 ++++++++++++++
 2 files changed, 19 insertions(+), 3 deletions(-)

diff --git a/libstdc++-v3/include/std/ranges b/libstdc++-v3/include/std/ranges
index 856975c6934..24411124580 100644
--- a/libstdc++-v3/include/std/ranges
+++ b/libstdc++-v3/include/std/ranges
@@ -892,11 +892,12 @@ namespace views::__adaptor
     };
 
   // Partial specialization of the primary template for the case where the extra
-  // arguments of the adaptor can always be safely forwarded by const reference.
-  // This lets us get away with a single operator() overload, which makes
-  // overload resolution failure diagnostics more concise.
+  // arguments of the adaptor can always be safely and efficiently forwarded by
+  // const reference.  This lets us get away with a single operator() overload,
+  // which makes overload resolution failure diagnostics more concise.
   template<typename _Adaptor, typename... _Args>
     requires __adaptor_has_simple_extra_args<_Adaptor, _Args...>
+      && (is_trivially_copyable_v<_Args> && ...)
     struct _Partial<_Adaptor, _Args...> : _RangeAdaptorClosure
     {
       tuple<_Args...> _M_args;
@@ -926,6 +927,7 @@ namespace views::__adaptor
   // where _Adaptor accepts a single extra argument.
   template<typename _Adaptor, typename _Arg>
     requires __adaptor_has_simple_extra_args<_Adaptor, _Arg>
+      && is_trivially_copyable_v<_Arg>
     struct _Partial<_Adaptor, _Arg> : _RangeAdaptorClosure
     {
       _Arg _M_arg;
diff --git a/libstdc++-v3/testsuite/std/ranges/adaptors/100577.cc b/libstdc++-v3/testsuite/std/ranges/adaptors/100577.cc
index 8ef084621f9..06be4980ddb 100644
--- a/libstdc++-v3/testsuite/std/ranges/adaptors/100577.cc
+++ b/libstdc++-v3/testsuite/std/ranges/adaptors/100577.cc
@@ -21,6 +21,7 @@
 // PR libstdc++/100577
 
 #include <ranges>
+#include <functional>
 
 namespace ranges = std::ranges;
 namespace views = std::ranges::views;
@@ -113,4 +114,17 @@ test03()
   x | std::views::drop(S{});
 }
 
+void
+test04()
+{
+  // Non-trivially-copyable extra arguments make a closure not simple.
+  using F = std::function<bool(bool)>;
+  static_assert(!std::is_trivially_copyable_v<F>);
+  using views::__adaptor::__closure_has_simple_call_op;
+  static_assert(!__closure_has_simple_call_op<decltype(views::take_while(std::declval<F>()))>);
+  static_assert(!__closure_has_simple_call_op<decltype(views::drop_while(std::declval<F>()))>);
+  static_assert(!__closure_has_simple_call_op<decltype(views::filter(std::declval<F>()))>);
+  static_assert(!__closure_has_simple_call_op<decltype(views::transform(std::declval<F>()))>);
+}
+
 // { dg-prune-output "in requirements" }
-- 
2.32.0.93.g670b81a890


  reply	other threads:[~2021-06-15 19:28 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-06-14 16:35 [PATCH] libstdc++: Refine range adaptors' "simple extra args" mechanism [PR100940] Patrick Palka
2021-06-15 18:53 ` [PATCH 2/1] libstdc++: Non-triv-copyable extra args aren't simple [PR100940] Patrick Palka
2021-06-15 19:28   ` Patrick Palka [this message]
2021-06-16 13:43     ` Jonathan Wakely
2021-06-17 10:25 ` [PATCH] libstdc++: Refine range adaptors' "simple extra args" mechanism [PR100940] Jonathan Wakely

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=571933d-e747-23e5-bd26-e141841aa682@idea \
    --to=ppalka@redhat.com \
    --cc=gcc-patches@gcc.gnu.org \
    --cc=libstdc++@gcc.gnu.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).