Towards improved C++ support for games, graphics, real-time, embedded systems [pdf] | Hacker News Reader

Towards improved C++ support for games, graphics, real-time, embedded systems [pdf]

open-std.org

70 pointsJSnake38 comments

The allocator support in C++11 and, from what I've seen, the proposed allocator support in C++1y/z/whatever still leaves a massive amount to be desired.

Polymorphic allocators solve what's basically a non-problem for games (ABI boundary problems) while introducing huge other issues. Alignment is a PITA but not really a pressing IMO. The real issue is stuff like rebind and the requirements to call destructors (which destroys many practical implementations of pools and arenas, respectively). This probably falls under "STL puts an emphasis on correctness before practicality and performance".

...

Honestly (and I am aware this isn't a popular opinion, although it's much more popular inside the game industry than outside of it), IMO the STL-style of C++ programming is fairly wrong for games in general, largely because of memory usage concerns. This style of programming emphasizes worrying about lifetimes of objects by way of RAII, in order to allow a uniform way of treating all types of resources, both memory and otherwise.

This is convenient, but games need extremely tight control over non-memory resource management (need to reuse anything from the system, need to load lots of stuff in the background, etc). This means tying it to object and scope lifetime tends to be inflexible and cause future pain when you need to change the resource management scheme.

And for memory, games tend not to care a lot about cleaning up each objects memory individually, preferring to do it in bulk (e.g. resetting an arena or pool). This means that using RAII (or using something that uses RAII) for memory management tends to fit poorly for games.

So, IDK. I'm not holding my breath on the C++ standards committee making a decision that I care for, but it looks like they're making steps in the right direction nonetheless.

There are no gamdevs on the c++ committee, probably because games companies are cheap bastards and don't want to expend resources on something that will affect them 10 years down the line (or even 2). Join the committee.

Haha, even if that were something I was interested in at all, I'm honestly not sure what you could do to fix it at this point. I think that ship has sailed. Most of the issues are library issues (and are far deeper than my complaints about allocators). The language issues largely come from having too many features, and its not like any of them can be taken out.

FWIW, I'm currently pinning my hopes on JAI, even though it's a long shot.

What's "JAI"? I've never heard of it, and Google/Wikipedia aren't showing any plausible candidates.

Jonathan Blow's language/compiler/live performance.

https://www.youtube.com/user/jblow888/playlists

See https://sites.google.com/site/jailanguageprimer/ for some fan-made notes from the videos

I'm not a game developer, but I experienced something similar to that when playing around with SFML. Their texture objects adhere to RIAA, so copy constructors will copy the texture from GPU memory to main memory, copy it to another location in main memory, re-upload it to GPU memory, then destroy the original texture and the memory it took up when copying to main memory.

I asked about it and the response I got was basically "It's RIAA, that's how you do C++". The worst part is they have a caching system underneath that tracks OGL handles across different objects, and it's even safe for access by threads.

But whoever designed their texture class chose not to use it.

So when I read your description I understand where you're coming from a bit. I'm not sure I understand why you couldn't just avoid using destructors, but I've honestly never tried to use a C++ allocator for anything as I've never needed control quite that tight for anything I do.

Well, you can avoid using desctructors, but then you can't use the majority of the STL, so what's the point? (This is basically what I do anyway)

You don't have to tie resource lifetime to object lifetime, that's the mistake the SFML folks made.

Right, my point is that I typically don't really want to tie memory lifetime to object lifetime either. So the whole of RAII is not a great fit, outside of a couple edge cases.

RAII is about resources, not memory. If you don't want to tie memory lifetime to object lifetime then don't. That's the entire point of things like smart pointers.

That's exactly the issue SFML had. This idea that RAII implies tying resources directly to an objects lifetime. It doesn't.

Still no support for static compile and runtime introspection. That could be easily and cheaply implemented, at least for POD and aggregates thereof.

If I have to build one more system that requires either a custom parser or repeating the names of all members, I think I will cry. The compiler ALREADY KNOWS!

The latest mailing (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2015/#mai...) includes several new reflection proposals. N4447 probably covers what you want (although I'm not a huge fan of it).

There's actually an official "study group" looking at compile time refection proposals, so this is definitely something in the works. However, the proposals have ranged from mediocre to downright horrible so it's probably not going to be in the standard any time soon.

Some light reading

  [1] (static if) http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3613.pdf
  [2] (call for compile-time reflection) http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3814.html
  [3] (enum properties queries) http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3815.html
  [4] http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n3996.pdf

I am very much in support of this. Not for games but serialization. Pretty surprised that there don't seem to be any compiled languages supporting static compile time reflection. I think D does, but that is it.

One (longish) shortcut is getting the info from the debugger which already knows. That is, in the build system, compiling a file #including what you want, firing a debugger, printing out the necessary type information, generating code, building that code with the rest of your system.

Static introspection is available via third party libraries such as Boost.Fusion.

"We started by looking at a paper from Paul Pedriana on N2771 “EASTL -- Electronic Arts Standard Template Library”. While this paper was submitted in 2007, neither chairs of LWG recall reviewing this paper and no record of its review exists."

I give you the C++ standardization committee. Someone submits a paper to them with their issues and, moreover with solutions fully implemented in a library and also details the whys and wherefores. I have nothing to say whether this was good, bad or middling work just that the C++ standardization committee "has no record" of actually having read and considered it 8 years later and the chairs don't recall doing so. But I'm sure they all ran around as "anointed smart people" claiming all the others on the committee are anointed smart people instead. Sorry if I'm sounding a bit bitter but they really need to be called out, this is really gross. It's contemptuous of people making the effort to make the language better is the best you can say of it. Arrogant, stupid, unkind, vicious and bitter might describe me and I am sorry for it if you feel that way (please note it and I will try to do better), but I think it's also a fair description of this act by the committee.

Also on the front page today was Bruce Eckel's "thoughts on scala" which were utterly ridiculous for the first 10 minutes I watched. -He's a C++ stanardization committee alumni.

"We feel that a holistic "here's a new STL and our problems" approach to any change seems a bit doomed. "

WTF? You can't learn anything about what changes are desirable in the language from the implementation of how they solved their problems and assess whether they are more generally applicable? Pieces of them aren't worth assessing to be just adopted (or rejected)? You can't criticise what they tried and note the implementation and why the STL can do better?

Seriously if you can't learn anything from this on the committee what the hell are you doing? Well NOW they are looking at it, hurrah! While justifying why it wasn't done before, boo!

Mea culpa and a rocket up the committee members of the time seems vastly more appropriate if you actually care about this language. (Obviously if you don't, resign from the committee, immediately, anything else is immoral) If the members didn't care then, highlight the fact and state it is unacceptable going forward.

If it's common or garden (hopefully isolated) incompetence like all of us display at times. "We're very sorry this happened." Would seem to be about the minimum standard here. It's the arrogance that goes along with the committee when they act like this that really gets me.

I also remember Bloomberg submitted in 2005 some very nice template library changes that avoid some very real pain points of STL:

www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1850.pdf

"Using an existing implementation of the C++ standard library as a base, we made backward-compatible modifications to enable such per-instance allocators."

I still don't know how I'd achieve per-instance allocator in modern "part-of-standard C++ STL." Is it possible? If yes, how? Does anybody know?

> I still don't know how I'd achieve per-instance allocator in modern "part-of-standard C++ STL." Is it possible? If yes, how? Does anybody know?

Well, there are two things you can do.

1. (Accepted in C++11) Allow stateful allocators. This is especially useful for things like pool allocators. Your container would keep a pointer to its allocator and perform allocator operations through it.

2. (Proposed but not accepted yet) Allow polymorphic allocators. This would allow different data structures to use different allocators but still have the same type. Basically it would look like 1 above, but instead of keeping an allocator pointer, you would keep a pointer to an allocator interface and perform allocator operations through virtual method calls.

C++ has left the horrors of the '90s largely behind it, and I'm rather liking it these days... But every time I work on a C++ project, I soon get "array angst" -- a feeling of profound incompetence because I feel uncertain about basic array operations.

Seems like every project has its own take on array classes and their ownership semantics. Std::vector appears to be largely useless (or at least it's not used anywhere in practice, so the effect is the same). Boost is just scary. Qt is almost like a different language by now.

I wish the C++ standard had something like Cocoa's NSArray/CFArrayRef -- a general-purpose ordered container that is consistently used everywhere in APIs and does the job everywhere, even though it may not be quite the optimal data structure. But I guess it's 15 years too late for that.

Could you explain why you think std::vector is useless?

Like I wrote in my comment, it doesn't seem to be used in the real-world projects and SDKs that I deal with.

Not if those projects/SDK's have to interface with C, but that's just the reality.

> I wish the C++ standard had something like Cocoa's NSArray/CFArrayRef

I think the proposed array_view is a much better choice in C++ land. The actual storage format of data (the data structure) should be an implementation detail, but there should be standard ways to share that data across various unrelated pieces of code.

http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n417...

It's actually kind of fascinating that the HN shortened title omits "games," which is really what this document focuses on. Specifically the document investigates "EASTL", i.e. the Electronic Arts STL, and how its learnings can be applied to the common STL.

The actual title: "Towards improved support for games, graphics, real-time, low latency, embedded systems."

These little pieces of censorship—that's not a sign that for some reason, we really don't like to talk about games. There's no pervasive second-class-citizenry attitude towards gaming. I don't really know how to interpret the omission, other than it's a little piece of little consequence.

Gaming keeps surfacing in places big and small as a motivation behind some technical or business innovation. Here we get very real improvements to the STL. This certainly matters: we celebrate Quake's square-root estimation trick as a cultural artifact, evidence of the beauty of programming. Elsewhere, in recent news, we hear Stewart Butterfield's narrative, which was as much about Flickr and Slack as it is about his twice-failed game concept. Lots of ideas and characters in the tech world have touched gaming in one way or another.

In my personal background at my startup, I keep running into folks pushed out of gaming. They start making "apps." Surrounded by app people, I might miss that a third of all new apps in the App Store are actually games[0]. Even then, games are disproportionately represented in Apple Design Awards. It's tough to get an accurate figure we'd all agree on, but I believe games are nonetheless also overrepresented in revenue (definitely from the looks of top grossing, if we're strictly speaking about mobile software).

If we really want to talk about software, I'd expect 1/3 to 1/2 of the content on the front page to actually be about games. It's certainly what most software people are working on, and it is probably where the plurality of the money is made. I'm not going to speculate as to why (I'd leave that for certain ex-Google developers, who possibly wanted to make a card game all his life). Here, we're actually seeing gaming omitted from the title, as though it's verboten.

[0] http://www.pocketgamer.biz/metrics/app-store/

I certainly didn't truncate "games" from the title myself. Did HN do that on its own?

> It's certainly what most software people are working on, and it is probably where the plurality of the money is made.

While gaming does drive innovation and it's a very interesting field, I doubt that's what most software people are working on and surely it's not the biggest sector of software industry in general, it's actually quite small in comparison with many others. That's not to undermine its value, just pointing out the broader perspective.

>It's certainly what most software people are working on, and it is probably where the plurality of the money is made.

Not sure where you got those ideas, but they're not even remotely true. Enterprise, while boring as all hell, is _huge_. The largest software vendors are (consistently) Microsoft, Oracle, IBM, and SAP[1]. According to Gartner, software was a $407.3B industry in 2013, and enterprise driven companies accounted for nearly half[2]. The video game industry pulled in $887M in 2014[3]. Not even close at .21%.

[1]: http://en.wikipedia.org/wiki/Software_industry#Size_of_the_i...

[2]: http://www.gartner.com/newsroom/id/2696317

[3]: http://www.statista.com/statistics/201093/revenue-of-the-us-...

> software was a $407.3B industry in 2013

And that's not counting embedded software, which is often sold as part of a hardware package.

You misread that chart; The video game industry pulled in $887M in the US in One Month (February) of 2014. For scale, bear in mind that GTA5 made worldwide revenue of $1billion in it's first three days.

If we go by Gartner[1], the games industry was predicted at $93billion in 2013. That's over 21%, so a couple of orders of magnitude more than 0.21%.

[1]:http://www.gartner.com/newsroom/id/2614915

You're right; I can't tell where exactly that 407B number comers from. It's irrelevant though in the context of what I was refuting, i.e., the claim that "most" software devs are in the game industry, which is ridiculous.

I think a moderator restored the original title as we usually do, but thought there wasn't enough room for the word "games" (there actually was, because we don't count " [pdf]" against the 80 char limit, but that isn't obvious). Invoking "censorship" here seems a bit over the top. Besides, games and games development stories are all over HN!

> censorship

It is censorship in the sense that the original authors' intention is modified to exclude the 'game' factor, but then .. why is this happening in the first place? Why do you think the editing-OP decided to omit this aspect?

> "pushed out of gaming"

There are a couple of schools of thought about 'games programmers' and 'game systems' which can be plotted with wide variance, yet have some interesting points of intersection. For example, the 'factoid' that games developers have a high turn-over/rockstar-phase/kidzone aspect, unless your game is making millions, in which case its serious business and only big boys need apply. So 'game ideology' as an industrious, conscious body of knowledge, has the punk-rock factor, wearing a 2000 dollar suit.

>> we really don't like to talk about games.

Game development as a software ethos, either from consumer or developer side in my opinion, is an interesting quandry.

Many times the notion that you are not making a game if you're making, instead, a business app, or a web scraper, or an OS kernel, and so on .. But actually the notion of game is key to all computing. From a technological perspective, we're all just playing games. Even Oracle has its event and frame "inner loop".

So I think there's an ugly truth about why we "don't like o talk about games", and it is that there is a zen to the subject. The zen is profitable, and it is also deadly.

Keep playing games - no matter where you go and what you do there. Keep making games, too.

>Why do you think the editing-OP decided to omit this aspect?

I didn't.

You're getting it backwards. Games are prominent because it's in their modus operandi. Their relationship with tech remains symbiotic only as long as they can act as a flashy demo. This is a thing that occurs only some of the time, and more often than not, one of the two is compromised.

The other side of it, that a lot of games are made, is mostly reflective of the wishful thinking everyone has about games as a business. There are far more modified WordPress blogs out there than games, and they're probably more successful from a value standpoint.

Still, we discuss games often enough because they have that tendency to keep floating into the consumer space of tech.

Games have been a huge driving force behind tech from the beginning. For example, Unix only exists because Ken Thompson wanted to port his game Space Travel to the PDP-7.

Is this true? That's a great piece of trivia if so.

Yeah, see http://en.wikipedia.org/wiki/Space_Travel_(video_game)

Reply on news.ycombinator.com