- Boost - lists.boost.org

Re: [boost] [Block Pointer] Benchmarks
by Glen Fernandes 15 Mar '16

15 Mar '16

On Sun, Mar 13, 2016 at 1:02 PM, Phil Bouchard wrote: > Thanks for your help. > > So block_ptr<> wouldn't be a drop-in replacement for shared_ptr<> but an > add-on for: > - complex containers, > - neural networks, > - C# / Java / Javascript engines, > - etc. > No problem. Also renaming the thread to reflect the current benchmarks (since it doesn't appear that block_ptr is faster than shared_ptr where creation is involved). 1. Gathering these benchmarks is always useful, even if block_ptr was not designed to be a drop-in replacement for share_ptr. So you had the right idea, even if the execution needed some work. It's important to get the execution correct, though, otherwise the results are not meaningful - which is where I (and I believe Rob, also) was trying to steer you. 2. Looking at the numbers you obtained after modifying my benchmark example, they are close enough that at least you could say "Using block_ptr over shared_ptr won't be a significant performance loss": > unique_ptr (new): 47.7686 > unique_ptr (make_unique): 46.8545 > shared_ptr (new): 77.8261 > shared_ptr (make_shared): 50.8072 > shared_ptr (allocate_shared_noinit): 33.021 > block_ptr (new): 69.6554 3. There are more benchmarks beyond just creation (though creation is likely the most meaningfully expensive one). Copying overhead might be one: I haven't looked at what is involved in copying a block_ptr, but some work happens when you copy a shared_ptr. 4. shared_ptr by way of allocate_shared allows creation to not involve 'new' expressions, or even a call to '::operator new(std::size_t)' at all. i.e. For those C++ projects that have a requirement for all dynamic allocation in their project to involve some stateful custom allocator instances of some stateful custom allocator type, they can still use shared_ptr (with allocate_shared). Is this possible with block_ptr? I hope to take a look at the block_ptr motivation, design, and implementation when I have some time next week. The title of the thread caught my attention; "X is 600% faster than Y" always has a high excitement potential. Glen

8 16

[Block Pointer] Update
by Phil Bouchard 14 Mar '16

14 Mar '16

Hi, - So I've restructured all the code in the correct namespaces boost::smart_ptr & boost::smart_ptr::detail: https://github.com/philippeb8/block_ptr/tree/master/include/boost/smart_ptr - I am now using another smart pointer to refer to proxies because it is safer but the code is 15% slower than without it (now ~= shared_ptr<>): make: auto_ptr: 26211292 ns shared_ptr: 27291607 ns block_ptr: 61097868 ns new: auto_ptr: 23240569 ns shared_ptr: 48016207 ns block_ptr: 55554121 ns - Because of the removal of the proxy optimization there is a new caveat; if you assign an rvalue (temporary pointer on the stack) to a pointer on the heap (opposite of normal heap object assigned to stack pointer) then there the behavior will be undefined. I am looking to fix it or throw an exception but for now it's better not to use rvalues when assigning to pointer living on the heap as such: https://github.com/philippeb8/block_ptr/blob/master/example/t100_test1.cpp#… Regards, -Phil

2 13

[Fit] formal review ends today, 13th March.
by Vicente J. Botet Escriba 14 Mar '16

14 Mar '16

Hi, The formal review of Paul Fultz II's Fit library ends today, 13th March. If you had not time to do the review and you plan to do it, please let us know so that we see if the review can be extended. Up to now we had 3 formal review (2 for conditional acceptance and 1 against, not now). I have not done a review myself, but I have created a lot of issues on Github. Best regards, Vicente Fit is a header-only C++11/C++14 library that provides utilities for functions and function objects. Fit is: - Modern: Fit takes advantages of modern C++11/C++14 features. It support both `constexpr` initialization and `constexpr` evaluation of functions. It takes advantage of type deduction, varidiac templates, and perfect forwarding to provide a simple and modern interface. - Relevant: Fit provides utilities for functions and does not try to implement a functional language in C++. As such, Fit solves many problems relevant to C++ programmers, including initialization of function objects and lambdas, overloading with ordering, improved return type deduction, and much more. - Lightweight: Fit builds simple lightweight abstraction on top of function objects. It does not require subscribing to an entire framework. Just use the parts you need. Fit is divided into three components: * Function Adaptors and Decorators: These enhance functions with additional capability. * Functions: These return functions that achieve a specific purpose. * Utilities: These are general utilities that are useful when defining or using functions Fit has been tested on gcc 4.6-4.9, clang 3.4-3.7, and Visual Studio 2015. For more information see: Github:https://github.com/pfultz2/Fit/tree/boost Documentation:http://pfultz2.github.io/Fit/doc/html/ We encourage your participation in this review. At a minimum, kindly state: - Whether you believe the library should be accepted into Boost * Conditions for acceptance - Your name - Your knowledge of the problem domain. You are strongly encouraged to also provide additional information: - What is your evaluation of the library's: * Design * Implementation * Documentation * Tests * Usefulness - Did you attempt to use the library? If so: * Which compiler(s) * What was the experience? Any problems? - How much effort did you put into your evaluation of the review? More information about the review process can be found here:http://www.boost.org/community/reviews.html We await your feedback!

4 6

[Fit] Review
by Zach Laine 13 Mar '16

13 Mar '16

- Whether you believe the library should be accepted into Boost Yes, conditionally. * Conditions for acceptance I think the documentation and naming issues I raise below, and those filed on GitHub by me and others should be addressed before acceptance. Also, the ADL issues that Steven Watanabe alluded to previously need to be fixed. After that, I think it should be re-submitted. - Your name Zach Laine - Your knowledge of the problem domain. I do generic programming regularly, and have done for years. You are strongly encouraged to also provide additional information: - What is your evaluation of the library's: * Design I think it's good. I particularly like the fact that it attempts to keep everything constexpr- and SFINAE-friendly wherever possible. I *really* like reveal() and failure(). Too few TMP-heavy libraries pay attention to providing good error messages. I have no problem with the constness requirements on Callables used throughout Fit, as was discussed previously. I even consider this a feature, allowing me to detect non-const Callables, which are almost always an error on my part. This is especially true since std::ref() and mutable_() let me explicitly choose how to handle mutable Callables, in a way that is also explicit at the call site. There are a number of adaptors that I find have obscure names: by() is a projection; why not call it project()? compress() is defined in terms of "fold", with a link to the Wikipedia page for same; why not call it foldl()? Similarly, reverse_compress() should be foldr(). flow() is a super obscure name! Why is this not reverse_compose() or similar? indirect() could more easily be understood if it were called deref() or dereference(). I don't feel as strongly about partial(), but I think it might be clearer if it were called partial_apply() or curry(). conditional() should be called invoke_first(), call_first_of(), or similar. I find it too easy to confuse with if_(). * Implementation I only briefly looked at the implementation, and only a few adaptors at that. What I saw did not leave me any concerns in particular. However, I *am* concerned by Steven Watanabe's claim that the library is ADL-unsafe. He did not state exactly why that I could tell, but this needs to be addressed if so. * Documentation The Quick Start section is good at showing a compelling motivating case for using the library. The result at the end of the section looks like a very good start on a "Scrap Your Boilerplate" solution for simple, dump-style printing. It's relatively easy to read and write. The documentation is what needs the most work, from what I've seen. The Quick Start is quite nice, but then there are some things that are underexplained in the subsequent Overview and Basic Concepts sections. For instance, why are these macros used everywhere and what do they do? They should be used for the first time after showing what the expanded code looks like. This will justify their use for those users that want them, and show those that don't what they can write instead. Others have written during the review that they don't understand why this library is useful. I think this is simply a documentation failure. I find the printing example from the Quick Start compelling, but perhaps Paul could show how it would have to be done if hand-written (even if only partially). Also, Paul has told me offline that there are a small number of primitive operations on which everything else is built (pack, unpack, and apply, IIRC). Perhaps show how those can be used to implement the other adapters, and how hard they are to write yourself. Showing how at least some of the library can be used as fundamental building blocks would go a long way. Some specifics. I have read the online docs before, and I put several bugs into the GitHub page for what I consider insufficient examples or explanations. Some new points, though: I think alias() is in bad need of a motivating example. I don't understand the distinction between lazy() and protect() just from the descriptions. In particular, this is what protect()'s docs say: '' The protect function adaptor can be used to make a bind expression be treated as a normal function instead. Both bind and lazy eargerly evaluates nested bind expressions. The protect adaptor masks the type so bind or lazy no longer recognizes the function as bind expression and evaluates it. '' I don't know what "masks the type" means here. This really needs an example. * Tests I did not run the tests, and just sort of spot-checked them. They seem a bit thin. In particular, one of the ones I looked at was conditional(). This one seems like it is needs more coverage, including thorny ones that must be kept working when maintenance occurs. There are probably others adapters that need similar treatment. Also, consider this -- Steven pointed out this as a failure case: boost::fit::conditional(identity(), identity())(0); and Paul says that it is supposed to fail, and that this failure is a feature, not a bug. If so, a minimal use like this that fails to compile needs to be added as a fail-compile test (and of course this intention should be documented). * Usefulness I think the library is very useful. There seem to be some things in there that use of which might be deemed a bit clever (not the good kind), like infix(), but then there are things like conditional() that I can see being *extremely* useful. - Did you attempt to use the library? If so: Yes. * Which compiler(s) Clang 3.6 * What was the experience? Any problems? I found I could gin up quick uses of things like conditional() (the most interesting adaptor for my purposes) in a short period of time. It was quite straightforward. - How much effort did you put into your evaluation of the review? About 5 or 6 hours. Zach

9 23

Alternative names to Boost.Fit
by paul Fultz 13 Mar '16

13 Mar '16

Hello All, Boost.Fit is a function utility library, however, the name currently doesn't have anything with this. I originally chose the name because it started with F(for function) and it was short and sounded nice. However, there could be some names(or abbreviations) that more adequately describe the library. Here are some alternatives that are short: - Boost.FP - Boost.FPL - Boost.Fun - Boost.FN Of course, I prefer the Boost.FN name. Names like Boost.FP or Boost.FPL could imply that it provides full functional constructs whereas Boost.Fit does not and is mainly focused on functions. Either way, this is something reviewers could consider as they review the library as I am not strongly tied to the name Boost.Fit. Thanks, Paul

13 27

[Block Pointer] Up to 600% faster than Shared Pointer
by Phil Bouchard 13 Mar '16

13 Mar '16

- I made the proxy explicit. Now it can be used 3 different ways: 1) Clothesline style: block_proxy x; block_ptr<int> v = block_ptr<int>(x, new block<int>(11)); block_ptr<int> w = block_ptr<int>(x, new block<int>(12)); 2) Root style: proxy_ptr<char[9]> u = proxy_ptr<char[9]>(new block<char[9]>()); 3) Container style (very easy to write containers now): struct list : block_proxy { public: list() : root(*this) {} [...] private: block_ptr<node> root; }; - I removed the capacity to unify sets. Why? Because it's faster & cleaner without it. - If we use fastblock<>() then the speedup can go up to 600% compared to shared_ptr<>: new: auto_ptr: 23174753 ns shared_ptr: 49726615 ns block_ptr: 7812659 ns - Once again all the examples can be found here: https://github.com/philippeb8/block_ptr/tree/master/example - And the pointer itself here: https://github.com/philippeb8/block_ptr/blob/master/include/boost/smart_ptr… Regards, -Phil

3 18

32/64 library name conflict under Windows?
by Peter Dimov 12 Mar '16

12 Mar '16

I remember that we discussed this problem before: C:\Projects\boost-git\boost>b2 --with-system toolset=msvc-10.0 address-model=32,64 [...] error: Name clash for '<pstage\lib>libboost_system-vc100-mt-gd-1_61.lib' error: error: Tried to build the target twice, with property sets having error: these incompatible properties: error: error: - <address-model>32 error: - <address-model>64 error: error: Please make sure to have consistent requirements for these error: properties everywhere in your project, especially for install error: targets. but don't remember what we concluded. Will we support building both 32 and 64 bit libraries under Windows? At present if I want to have an IDE project that has both Win32 and x64 configurations I need to build the 32 bit libraries, move them out of stage/lib somewhere, then build the 64 bit libraries, then move them out of stage/lib to somewhere else, and set my library directories in the IDE accordingly. It would be much more convenient if 32/64 were encoded into the name, so that I can leave my library path to point at stage/lib and rely on autolink to choose the right .lib file automatically, like it works today for the other options (debug/release, static/dynamic runtime).

14 25

[Fit] Louis Dionne's formal review
by Louis Dionne 12 Mar '16

12 Mar '16

First of all, sorry for submitting my review so late. > - Whether you believe the library should be accepted into Boost Yes, conditionally. > - Your name Louis Dionne > - Your knowledge of the problem domain. Extensive. In fact, I have implemented something very similar to Fit in parallel to Paul, but in the Hana library. Hence, I understand what Fit can be useful for, and I know many of the challenges Paul must have faced in designing the library. > You are strongly encouraged to also provide additional information: > - What is your evaluation of the library's: > * Design Good. It is simple and quite lean, which is nice. However, - The issue about `capture`'s semantics (and `pack`'s semantics too) ought to be addressed. I created an issue for this: https://github.com/pfultz2/Fit/issues/141. - It would be nice to provide a master header: https://github.com/pfultz2/Fit/issues/143 - I am not sure that `alias` has its place as part of the public interface of the library. I have only looked at it quickly, but it seems to diverge pretty heavily from the purpose of the rest of the library. I know it's used to implement `fit::pack`, but is it necessary to make it part of the public interface? Zach Laine already opened a similar issue here: https://github.com/pfultz2/Fit/issues/95 - I'm not sure about `compress` and `reverse_compress`. I think `fold` and `reverse_fold` (or my favorite `fold_left` and `fold_right`) would be better suited. An issue was already filed by Vicente here: https://github.com/pfultz2/Fit/issues/62 > * Implementation I have not looked at the implementation intensively, but I have to say that the profusion of macros is a bit off-putting. That being said, it's probably difficult to do better while supporting archaic compilers such as GCC 4.6. Hats off to Paul on this one. Also, I think the `fit/detail/ref_tuple.hpp` is unused and should be removed. Finally, why do you use your own `move` function? Is it because it is not marked constexpr in pre-C++14? > * Documentation I submitted pull requests to fix typos, but these are not worth mentioning here. Comments that might be of interest for the purpose of this review were made in the form of GitHub issues: - https://github.com/pfultz2/Fit/issues/131 - https://github.com/pfultz2/Fit/issues/132 - https://github.com/pfultz2/Fit/issues/133 - https://github.com/pfultz2/Fit/issues/135 - https://github.com/pfultz2/Fit/issues/136 - https://github.com/pfultz2/Fit/issues/138 - https://github.com/pfultz2/Fit/issues/139 - https://github.com/pfultz2/Fit/issues/140 - https://github.com/pfultz2/Fit/issues/142 - https://github.com/pfultz2/Fit/issues/144 > * Tests I have looked quickly, and basically all the individual components seem to be tested. It's hard to evaluate the depth in which each component is tested, but it seems OK. Also, the fact that the library is tested on Travis CI on each commit is a blessing, and it makes me much more confident about its quality. > * Usefulness This, I think, is the biggest problem of the library. I personally know very well why the library is useful, but I can also understand that someone with a different background would not feel the same. In my view, the library is mostly useful as a replacement to shortcomings of lambdas in C++11/14. Not only this, but in large part yes. Indeed, let's say I have a function object `f`, and I'd like to get a function object that swaps the arguments and calls `f`. The obvious solution (simplified) is auto g = [f](auto x, auto y) { return f(y, x); }; In other words, the simplest way to express my intent is to use a lambda and to do exactly what I mean. Unfortunately, this has several limitations - The lambda above can't be constexpr - The lambda can't appear in an unevaluated context So I end up using `fit::flip`. In a different situation, I might want to capture many variables in a lambda, but moving them into the capture (not copying). So I write auto f = [x = std::move(x)...](auto const& param) { ... }; but unfortunately there's no way to capture a parameter pack by moving each of its elements in a lambda. So I resort to using `fit::capture`. Examples like this abound, and IMO these limitations are the main (but not the only) reason for using Fit. Actually, this is how I implemented a library similar to Fit inside Hana; I started by using lambdas everywhere and then I wanted to make everything constexpr, capturing by move and things like that. All of this is very useful, but I feel like the documentation completely misses the goal of explaining the problem solved by Fit (the problem I would use it for, at least). This failure of motivating its own existence is the main reason why I think Fit should be accepted conditionally. > - Did you attempt to use the library? If so: > * Which compiler(s) AppleClang 6.1.0.6020053 (shipped with Xcode 6.4) Clang 3.5, 3.6, 3.7 I used both the Makefile and the Ninja generators for CMake, and both worked. > * What was the experience? Any problems? Classic CMake setup, everything works well and the tests compile and run fast. No problem whatsoever, as expected. > - How much effort did you put into your evaluation of the review? A couple of hours for the formal review itself, but I have been looking at Fit's development for a while. Summary ------- Fit should be accepted conditionally and re-submitted for a mini-review before it can be accepted. My conditions are 1. Correctly justify the purpose of the library 2. Resolve the issues on `capture` and `pack`'s semantics I think all the other points I raised are important too, but they shouldn't refrain Fit from being accepted. I also trust that Paul would be willing to continue working on the issues that were raised during this review if Fit were to be accepted. Regards, Louis

4 5

Boost.ASIO SSL examples
by Umit Sami 11 Mar '16

11 Mar '16

Hi All, SSL examples in boost.asio is very limited and not that clear. How can I improve my knowledge of this great network/security library ? what are some of well known online websites and books I can read/buy to better write secure/encrypted client-server applications ? Thank you, U

2 1

[release] Need help to test new release archives.
by Rene Rivera 09 Mar '16

09 Mar '16

As part of my attempt to make the Boost release management less painful, for everyone, I've implemented building of the release archives in Travis-CI. As of yesterday archives of the "boost" super-project master branch are getting uploaded to the bintray.com service at < https://bintray.com/boostorg/snapshots/master/view>. There are still some kinks in how this works but it's automated enough at this point that it might serve the upcoming release needs. I would appreciate it if people download the archives available at the above link and test them out. In particular I'm looking to see if there's anything missing or generally incorrect. For reference the archives are named with the commit SHA that they correspond to. Thank you. -- -- Rene Rivera -- Grafik - Don't Assume Anything -- Robot Dreams - http://robot-dreams.net -- rrivera/acm.org (msn) - grafikrobot/aim,yahoo,skype,efnet,gmail

6 16