Commit Graph

54 Commits

Author SHA1 Message Date
Howard Hinnant d77851e837 Implement [util.smartptr.shared.atomic]. This is the last unimplemented
section in libc++.  This requires a recompiled dylib.  Failure to rebuild
the dylib will result in a link-time error if and only if the functions from
[util.smartptr.shared.atomic] are used.

The implementation is not lock free.  After considerable thought, I know of no
way to make the implementation lock free.  Ideas welcome along that front.  But
changing the ABI of shared_ptr is not on the table at this point.

The mutex used to lock these function is encapsulated by std::__sp_mut.  The
only thing the client knows about std::__sp_mut is that it has a void* data
member, can't be constructed, and has lock and unlock members.  Within the
binary __sp_mut is currently implemented as a pointer to a std::mutex.  That can
change in the future without disturbing the ABI (as long as sizeof(__sp_mut)
remains constant.

I specifically did not make __sp_mut a spin lock as I have a pathological
distrust of spin locks.  Testing on OS X reveals that the use of std::mutex in
this role is not a large performance penalty as long as the contention for the
mutex is low (more likely to get the lock than to have to wait).  In the future
we can still make __sp_mut a spin lock if that is what is desired (without ABI
damage).

The dylib contains 16 __sp_mut's to be chosen based on the hash of the address
of the shared_ptr.  The constant 16 is a ball-park reasonable space/time
tradeoff.

std::hash<T*> was changed to call __murmur2_or_cityhash, instead of the identity
function.  I had thought we had already done this, but I was mistaken.

All of this is under #if __has_feature(cxx_atomic) even though the
implementation is not lock free, because the signatures require access to
std::memory_order, which is currently available only under
__has_feature(cxx_atomic).

llvm-svn: 160940
2012-07-30 01:40:57 +00:00
Howard Hinnant c0937e8add Appy constexpr to <memory>. Picked up a few missing noexcepts as well.
llvm-svn: 159902
2012-07-07 20:56:04 +00:00
Howard Hinnant ed81e910bd Revert fix to http://llvm.org/bugs/show_bug.cgi?id=12867 for the reason now included in the code comment.
llvm-svn: 157128
2012-05-19 20:20:49 +00:00
Howard Hinnant 9042d623bf Protect __shared_weak_count::__get_deleter declaration with _LIBCPP_NO_RTTI. Fixes http://llvm.org/bugs/show_bug.cgi?id=12867
llvm-svn: 157049
2012-05-18 13:06:21 +00:00
Howard Hinnant bff1bfc6be Greatly scale back ambitions of emulating move semantics in C++03 mode. It was causing more problems than it solved. This fixes http://llvm.org/bugs/show_bug.cgi?id=12704.
llvm-svn: 155918
2012-05-01 15:37:54 +00:00
Howard Hinnant f2f2d8b98d Modernize conversion to bool to the explicit bool conversion operator (library wide). This fixes http://llvm.org/bugs/show_bug.cgi?id=12058.
llvm-svn: 151088
2012-02-21 21:46:43 +00:00
Howard Hinnant 67f3964766 Modernize relational operators for shared_ptr and unique_ptr. This includes adding support for nullptr, and using less<T*>. Fixes http://llvm.org/bugs/show_bug.cgi?id=12056.
llvm-svn: 151084
2012-02-21 21:02:58 +00:00
Howard Hinnant 9741d6c96e Implement a few optimizations for vector push_back and insert. Fixes r10828365.
llvm-svn: 150542
2012-02-15 00:41:34 +00:00
Howard Hinnant 21fa1887e0 Enable full functionality of shared_ptr<const void> by adding allocator<const void>. Credit to John Hurley for discovering this bug.
llvm-svn: 148508
2012-01-19 23:15:22 +00:00
Howard Hinnant 8e25104b10 1. Fix make_shared<const T>. 2. Allow allocator<const T> as an extension. 3. Refactor work which fixed unique_ptr<const T[]>. 4. Remove no-longer-needed private declarations from unique_ptr. 5. Add constraints to some shared_ptr and weak_ptr constructors and assignment operators so that is_constructible/is_assignable give the correct answers for shared_ptr and weak_ptr. 6. Make defensive preparations in the shared_ptr free functions for the introduction of shared_ptr<T[]> in the future. 7. As an optimization, add move constructor and move assignment to weak_ptr.
llvm-svn: 147437
2012-01-02 17:56:02 +00:00
Howard Hinnant b34b48196c The exception recovery mechanism for the uninitialized_* algorithms did not work for iterators into discontiguous memory.
llvm-svn: 147343
2011-12-29 17:45:35 +00:00
Howard Hinnant a87b5e3446 Fix http://llvm.org/bugs/show_bug.cgi?id=11616
llvm-svn: 146881
2011-12-19 17:58:44 +00:00
Howard Hinnant e4097ad7e6 Allow unique_ptr<T const []> to be constructed and assigned from a unique_ptr<T[]>
llvm-svn: 146853
2011-12-18 21:19:44 +00:00
Howard Hinnant d6f44b6601 Allow unique_ptr<T const []> to be constructed with a T* (in addition to a const T*)
llvm-svn: 146736
2011-12-16 15:37:23 +00:00
Howard Hinnant 42b8bb5033 Fix http://llvm.org/bugs/show_bug.cgi?id=11461. Credit Alberto Ganesh Barbati.
llvm-svn: 146345
2011-12-11 20:31:33 +00:00
Howard Hinnant 53d2fb0aa2 Installation of CityHash by Craig Silverstein
llvm-svn: 146329
2011-12-10 20:28:56 +00:00
Howard Hinnant f3d14a65ca Starting using murmur2 when combining multiple size_t's into a single hash, and also for basic_string. Also made hash<thread::id> ever so slighly more portable. I had to tweak one test which is questionable (definitely not portable) anyway.
llvm-svn: 145795
2011-12-05 00:08:45 +00:00
Howard Hinnant 9b0cd149aa Version #next on the hash functions for scalars. This builds on Dave's work, extends it to T*, and changes the way double and long double are handled (no longer convert to float on 32 bit). I also picked up a minor bug with uninitialized bits on the upper end of size_t when sizeof(size_t) > sizeof(T), e.g. in hash<float>. Most of the functionality has been put in one place: __scalar_hash in <memory>. Unfortunately I could not reuse __scalar_hash for hash<long double> on x86 because of the padding bits which need to be zeroed. I didn't want to add this zeroing step to the more general __scalar_hash when it isn't needed (in the absence of padding bits). I'm not ignoring the hash<string> issue (possibly changing that to a better hash). I just haven't gotten there yet.
llvm-svn: 145778
2011-12-03 21:11:36 +00:00
Howard Hinnant c206366fd7 Quash a whole bunch of warnings
llvm-svn: 145624
2011-12-01 20:21:04 +00:00
Howard Hinnant c003db1fca Further macro protection by replacing _[A-Z] with _[A-Z]p
llvm-svn: 145410
2011-11-29 18:15:50 +00:00
Howard Hinnant ab4f438239 Add protection from min/max macros
llvm-svn: 145407
2011-11-29 16:45:27 +00:00
Howard Hinnant 073458b1ab Windows support by Ruben Van Boxem.
llvm-svn: 142235
2011-10-17 20:05:10 +00:00
Howard Hinnant 2c0a65ee78 Fix <rdar://problem/10217868>.
llvm-svn: 140907
2011-10-01 00:26:36 +00:00
Howard Hinnant 89bdcd7ef5 Configure to get along with 2.9 clang
llvm-svn: 136526
2011-07-29 21:35:53 +00:00
Howard Hinnant e3163f5ae3 http://llvm.org/bugs/show_bug.cgi?id=10390
llvm-svn: 135393
2011-07-18 15:51:59 +00:00
Howard Hinnant 5a33687da0 Correct for new rules regarding implicitly deleted special members. http://llvm.org/bugs/show_bug.cgi?id=10191
llvm-svn: 134248
2011-07-01 19:24:36 +00:00
Howard Hinnant ce48a1137d _STD -> _VSTD to avoid macro clash on windows
llvm-svn: 134190
2011-06-30 21:18:19 +00:00
Douglas Gregor 64ec101eb6 Teach libc++ about the addressof() overloads it needs to work with
Objective-C Automatic Reference Counting, where Objective-C object
pointers can have several different qualifiers (__strong, __weak,
__autoreleasing, __unsafe_unretained). These addressof() overloads are
only provided in ARC mode, and the __weak variant is conditionalized
on having weak-reference support in the ARC runtime.

For historical reasons, Clang provides these definitions itself, and
defines the macro _LIBCPP_PREDEFINED_OBJC_ARC_ADDRESSOF to note when
it as done so. The code belongs here, and this redundancy will be
eliminated in the future.

Addresses <rdar://problem/9658274>.

llvm-svn: 133656
2011-06-22 22:17:44 +00:00
Howard Hinnant ce53420e37 Provide names for template and function parameters in forward declarations. The purpose is to aid automated documentation tools.
llvm-svn: 133008
2011-06-14 19:58:17 +00:00
Howard Hinnant b58f59cdb3 Second try at getting noexcept on move and swap for deque. I changed std::alloctor to propagate_on_container_move_assignment so as to make deque<T> move assignment noexcept. What we really need is a compile-time switch that says an allocator always compares equal.
llvm-svn: 132490
2011-06-02 21:38:57 +00:00
Howard Hinnant 3739fe79e5 noexcept for <memory>. I've added a few extension noexcept to: allocator_traits<A>::deallocate, allocaate<T>::deallocate, return_temporary_buffer, and default_delete<T>::operator()(T*) const. My rationale was: If a std-dicated noexcept function needs to call another std-defined function, that called function must be noexcept. We're all a little new to noexcept, so things like this are to be expected. Also included fix for broken __is_swappable trait pointed out by Marc Glisse, thanks Marc|. And fixed a test case for is_nothrow_destructible. Destructors are now noexcept by default|
llvm-svn: 132261
2011-05-28 14:41:13 +00:00
Howard Hinnant 5eb6bdfa1d __invokable and __invoke_of now check for incomplete types and issue a compile-time diagnostic if they are used with incomplete types for anything except a return type. Note that both arguments *and* parameters are checked for completeness.
llvm-svn: 131818
2011-05-22 00:09:02 +00:00
Howard Hinnant f738497500 Corrected some bugs in both memory and the tests. Preparing for being able to turn on support for alias templates.
llvm-svn: 131199
2011-05-11 20:21:19 +00:00
Howard Hinnant 72c5e14203 Qualify calls to addressof with std::. Bug 9106
llvm-svn: 124726
2011-02-02 17:36:20 +00:00
Howard Hinnant 119cd0ba12 fix guard
llvm-svn: 123269
2011-01-11 20:02:45 +00:00
Michael J. Spencer f5799be4a8 Add CMake build and fix major Linux blockers.
llvm-svn: 121510
2010-12-10 19:47:54 +00:00
Howard Hinnant ca74048398 N3142. Many of these traits are just placeholders with medium quality emulation; waiting on compiler intrinsics to do it right.
llvm-svn: 119854
2010-11-19 22:17:28 +00:00
Howard Hinnant 48d05bd26c LWG 1339
llvm-svn: 119699
2010-11-18 16:13:03 +00:00
Howard Hinnant a4a1ef1fc2 LWG 1404
llvm-svn: 119609
2010-11-18 01:40:00 +00:00
Howard Hinnant 412dbebe1b license change
llvm-svn: 119395
2010-11-16 22:09:02 +00:00
Howard Hinnant 9b35c8275c Dave Zarzycki showed how the efficiency of shared_ptr could be significantly
increased.  The following program is running 49% faster:

#include <iostream>
#include <memory>
#include <chrono>
#include <vector>
#include "chrono_io"

int main()
{
    typedef std::chrono::high_resolution_clock Clock;
    Clock::time_point t0 = Clock::now();
    {
        std::shared_ptr<int> p(new int (1));
        std::vector<std::shared_ptr<int> > v(1000000, p);
        v.insert(v.begin(), p);
        v.insert(v.begin(), p);
        v.insert(v.begin(), p);
        v.insert(v.begin(), p);
    }
    Clock::time_point t1 = Clock::now();
    std::cout << (t1-t0) << '\n';
}

llvm-svn: 119388
2010-11-16 21:33:17 +00:00
Howard Hinnant c4931c4746 I have reverted all contributions made by Jesse Towner in revision 110724
llvm-svn: 119383
2010-11-16 21:10:23 +00:00
Howard Hinnant 848a5374d0 visibility-decoration.
llvm-svn: 114551
2010-09-22 16:48:34 +00:00
Howard Hinnant f9cca3b7d0 I am experimenting with putting visibility-default attributes on all struct/classes in libc++. This checkin decorates only basic_string and vector as an experiment, and for review by those in this audience that might know more about visibilty than I do. If I get no negative feedback on this procedure I will begin to decorate the entire library in this way.
llvm-svn: 113590
2010-09-10 16:42:26 +00:00
Howard Hinnant 7609c9b665 Changed __config to react to all of clang's currently documented has_feature flags, and renamed _LIBCPP_MOVE to _LIBCPP_HAS_NO_RVALUE_REFERENCES to be more consistent with the rest of the libc++'s flags, and with clang's nomenclature.
llvm-svn: 113086
2010-09-04 23:28:19 +00:00
Howard Hinnant b3371f6f49 Fixing whitespace problems
llvm-svn: 111750
2010-08-22 00:02:43 +00:00
Howard Hinnant 9f1102b7fe US 108, N3109
llvm-svn: 111747
2010-08-21 21:14:53 +00:00
Howard Hinnant 20cc2a42b8 US 107
llvm-svn: 111538
2010-08-19 18:39:17 +00:00
Howard Hinnant 54b409fdb9 now works with -fno-exceptions and -fno-rtti
llvm-svn: 110828
2010-08-11 17:04:31 +00:00
Howard Hinnant eb2692571f patch by Jesse Towner, and bug fix by Sebastian Redl
llvm-svn: 110724
2010-08-10 20:48:29 +00:00