On 27/06/18 21:03 +0200, François Dumont wrote:
On 26/06/2018 18:03, Jonathan Wakely wrote:
On 18/06/18 23:01 +0200, François Dumont wrote:
Hi
I abandon the idea of providing Debug algos, it would be too
much code to add and maintain. However I haven't quit on reducing
Debug mode performance impact.
So this patch make use of the existing std::__niter_base to
get rid of the debug layer around __gnu_debug::vector<>::iterator
so that __builtin_memmove replacement can take place.
As _Safe_iterator<> do not have a constructor taking a pointer
I change algos implementation so that we do not try to instantiate
the iterator type ourselves but rather rely on its operators + or
-.
The small drawback is that for std::equal algo where we can't
use the __glibcxx_can_increment we need to keep the debug layer to
make sure we don't reach past-the-end iterator. So I had to remove
usage of __niter_base when in Debug mode, doing so it also disable
potential usage of __builtin_memcmp when calling std::equal on
std::__cxx1998::vector<> iterators. A rather rare situation I
think.
Note that I don't know how to test that __builtin_memmove has
been indeed used. So I've been through some debug sessions to
check that.
The attached patch (not fully tested) seems to be a much simpler way
to achieve the same thing. Instead of modifying all the helper
structs, just define a new function to re-wrap the result into the
desired iterator type.
Through my proposal I tried to limit the presence of Debug code in the
normal mode code. If you prefer to preserve existing code it is indeed
a simpler approach.
But to avoid any dependency on _GLIBCXX_DEBUG you had to make more
changes in more places, and complicate all the helper structs.
If we really want to avoid the #ifdef _GLIBCXX_DEBUG code in
<bits/stl_algobase.h> we could move the debug mode overloads into
<debug/stl_iterator.h>. I don't feel strongly about that, I think it's
OK in <bits/stl_algobase.h> and would be OK in the debug header too
(assuming it still works).
diff --git a/libstdc++-v3/include/debug/stl_iterator.h
b/libstdc++-v3/include/debug/stl_iterator.h
index a6a2a76..eca7203 100644
--- a/libstdc++-v3/include/debug/stl_iterator.h
+++ b/libstdc++-v3/include/debug/stl_iterator.h
@@ -120,4 +120,19 @@ namespace __gnu_debug
#endif
}
+#if __cplusplus >= 201103L
+namespace std
+{
+_GLIBCXX_BEGIN_NAMESPACE_VERSION
+
+template<typename _Iterator, typename _Container, typename _Sequence>
+ _Iterator
+ __niter_base(const __gnu_debug::_Safe_iterator<
+ __gnu_cxx::__normal_iterator<_Iterator, _Container>,
+ _Sequence>&);
+
+_GLIBCXX_END_NAMESPACE_VERSION
+}
+#endif
Why is this overload only defined for C++11 and later? I defined it
unconditionally in the attached patch.
I initially wrote something like:
template<typename _Iterator, typename _Sequence>
inline auto
__niter_base(const __gnu_debug::_Safe_iterator<_Iterator,
_Container>& __it)
-> decltype(__niter_base(__it.base()))
{ return __niter_base(__it.base()); }
which require C++11
I forgot to remove the __cplusplus check when I eventually use a more
specific overload.
Ah OK, I wondered if there was some reason I had missed.
What do you think?
Yes, that is a fine approach. Do you wat to apply it or do you want me
to work on it a little and re-submit the patch ?
I sent a follow-up with a complete patch that passes the testsuite.
I'm happy with that version of the patch, but do you have any
suggestions to improve it?
