Possible quick win in GC?
Perhaps I've too had much caffeine today but I've had an idea which might give a fairly quick win on the GC speed situation. It's a simple idea at heart so it's very possible/likely that this is a well known idea that has already been discarded but anyway here goes. I got the idea after thinking that it should be fairly simple for the compiler to detect straightforward cases of when a variable can be declared as going on the stack - i.e. no references to it are retained after its enclosing function returns. At the moment AIUI it is necessary for a class instance to be declared by the programmer as 'scoped' for this to take place. Further, I was considering the type of ownership and boundary considerations that could be used to improve memory management - e.g. using the notion of an owner instance which, upon destruction, destroys all owned objects. Afer some consideration it seems to me that by using only static analysis a tree of references could be constructed of references from a root 'scoped' object to all referred to objects that are allocated after the allocation of the root object. When the root object goes out of scope it is destroyed and all the descendent objects from the root object (as identified by the static analysis) could also be destroyed in one simple shot. The static analysis of course constructs the tree by analysing the capturing of references from one object to another. It could be the case that even a simple static analysis at first (e.g. discard the technique in difficult situations) could cover a lot of use cases (statistically). Of course, if one of the descendent objects is referred to by an object which is not in the object tree, then this technique cannot be used. However, I envisage that there are many situations where upon the destruction of a root object all related post-allocated objects can also be destroyed. In terms of implementation I see this being done by what I am calling 'bands' within the GC. With the allocation of any identified root object, a new band (heap) is created in the GC. Child objects of the root object (i.e. only referred to by the root object and other child objects in its tree) are placed in the same band. When the root object goes out of scope the entire band is freed. This by definition is safe because the static analysis has ensured that there are no 'out-of-tree' references between child objects in the tree and out-of-tree (out-of-band) objects. This property also means that normal GC runs do not need to add the scoped root object as a GC root object - this memory will normally only be freed when the scoped root object at the top of the tree goes out of scope. If memory becomes constrained then the bands can be added as root objects to the GC and memory incrementally freed just as with regularly allocated objects. Sorry if this idea is daft and I've wasted your time!
Re: Possible quick win in GC?
Here's a code snippet which mopefully makes things a bit clearer:
/**
* In this example the variable foo can be statically analysed as
safe to go on the stack.
* The new instance of Bar allocated in funcLevelB is only
referred to by foo. foo can
* be considered a root 'scoped' variable and the GC can delete
both foo and the new Bar()
* when foo goes out of scope. There is no need (except when under
memory pressure) for
* the GC to scan the band created for foo and it's related child
allocations.
*/
import std.stdio;
class Bar {
public:
int x;
this(int x) {
this.x = x;
}
}
class Foo {
public:
Bar bar;
}
void funcLevelA() {
Foo foo = new Foo(); // static analysis could detect this as
able to go on the stack
funcLevelB(foo);
writeln(foo.bar.x);
}
void funcLevelB(Foo foo) {
foo.bar = new Bar(12); // this allocated memory is only referred
to by foo, which
// static analysis has
established can go on the stack
}
void main() {
funcLevelA();
}
Re: GCs in the news
On Thursday, 17 July 2014 at 16:56:56 UTC, Vic wrote: On Thursday, 17 July 2014 at 13:29:18 UTC, John wrote: snip If D came without GC, it would have replaced C++ a long time ago! Agree +1000. If GC is so good, why not make it an option, have a base lib w/o GC. If I want GC, I got me JRE. It seems that some in D want to write a better JRE, and that just won't happen ever. Cheers, Vic I can't think of anyone posting here, to be honest, who wants to write a better JRE. The JRE is a virtual machine, and java compiles to bytecode that is run on the JVM. On the contrary, and in accordance with the core principle that D is a systems programming language, D compiles to native and (hopefully) highly optimised native machine code. There does exist something of a 'culture clash' where, by the very nature of GCs, there can be not-insignificant pauses in the running of the program that would be inimicable to real-time software such as high res complex games, operating systems, drivers etc. The response to this in the forums is either to improve the GC so that it doesn't ever pause for more than a certain amount of time (e.g. concurrent GCs, remove the global lock so other threads can continue to run), or to offer alternative memory management approaches such as ARC, which can also have pauses, but at other inflections as the program runs. Personally I'm a bit disappointed that the good work that has been done on GCs so far doesn't seem to be being picked up and run with, and nor do I see any reasons given as to why that is the case. Adnrei was threatening to start another GC an one point but unfortunately I haven't seen any more of that and we all know how short of time every one seems to be these days. Also on a personal note, I see some slightly snarky comments about D targeting C# and Java. Well from my perspective I'm extremely happy with the fact that D is a better C# and a better Java. I just wish it had Qt (I must finish my bindings for Qt) and/or ran on Android! The GC issues are irrelevant for me.
Re: Need Feedback for a new book - D Cookbook
Hi Paushali I'm interested in reviewing the book, you can contact me at alynch4...@gmail.com. I'd put the review on Amazon.
Re: A Perspective on D from game industry
On Sunday, 15 June 2014 at 11:28:12 UTC, Peter Alexander wrote: http://c0de517e.blogspot.ca/2014/06/where-is-my-c-replacement.html?m=1 The arguments against D are pretty weak if I'm honest, but I think it's important we understand what people think of D. I can confirm this sentiment is fairly common in the industry. Watch out for the little jab at Andrei :-P Reading his summary of the alternatives I felt D came out clearly on top, it's just that he didn't have the motivation to switch. Towards the end he mentions the web, for me (as an application developer rather than systems level guy) Android/iOS is the fly in the ointment - I'm torn as to whether to invest my energies in following D through its explorations or knuckling down and learning the Android API - after all, JDK8 + tooling is bearable now.
Re: A Perspective on D from game industry
On Sunday, 15 June 2014 at 13:19:12 UTC, Russel Winder via Digitalmars-d wrote: On Sun, 2014-06-15 at 12:30 +, Abdulhaq via Digitalmars-d wrote: […] learning the Android API - after all, JDK8 + tooling is bearable now. On the other hand Android API is Apache Harmony which is Java 6. Yes I keep forgetting that - wishful thinking maybe. Of note: Groovy finally works on Android, so you can use what Java 8 brings, on Java 6 and Java 7 using Groovy. And note Groovy may be a dynamic language, but it is also a static language. I'll look into it. Perhaps this question is just too broad, but if you wanted to develop an application on the Android platform right now, what approach would you take? Java, Groovy, web-based?
Re: A Perspective on D from game industry
On Sunday, 15 June 2014 at 20:10:34 UTC, Walter Bright wrote: On 6/15/2014 9:20 AM, Xinok wrote: Given that he lives in Italy, it's safe to assume that English is not his first language. But rather than consider what he has to say or dispute his arguments, you completely dismissed his point of view because his level of writing doesn't meet your standards. Xinok does have a point that we all should be aware of. I've found a very strong correlation between poor writing skills and disorganized thinking. (Your point about non-native English speakers is well taken, one must not confuse unfamiliarity with English with disorganized thinking.) I'm hardly the only one. If one wants their views to be taken seriously, pay attention to spelling, grammar, paragraphs, organized writing, etc. There's an awful lot of stuff to read on the internet, and poor writing often elicits a meh, I'll skip this one and move on reaction. True but if I'm going to judge a comment by the way it's written I'll take a second language piece over a foul and insulting rant any day of the week.
SurveyMonkey for D users OS - Results
There's been 100 votes and the results are: Linux 64 bits: 53 Linux 32 bits: 4 Windows 64 bits:27 Windows 32 bits: 3 Mac: 7 Other: 6: ArchLinux Android Centos 6 MAC OSX, LINUX 64, Windows 64, FreeBSD 64 bsd64 One 'other' vote was spoiled. It turns out that the free SurveyMonkey account only allows 100 votes max, but the profile has been much the same since 50 votes so I think the ratios are clear. If anyone has an OS other than the ones mentioned above then perhaps they could mention it in this thread.
Re: SurveyMonkey for D users OS - Results
On Saturday, 31 May 2014 at 13:37:26 UTC, Abdulhaq wrote: There's been 100 votes and the results are: Linux 64 bits: 53 Linux 32 bits: 4 Windows 64 bits:27 Windows 32 bits: 3 Mac: 7 Other: 6: ArchLinux Android Centos 6 MAC OSX, LINUX 64, Windows 64, FreeBSD 64 bsd64 One 'other' vote was spoiled. It turns out that the free SurveyMonkey account only allows 100 votes max, but the profile has been much the same since 50 votes so I think the ratios are clear. If anyone has an OS other than the ones mentioned above then perhaps they could mention it in this thread. See the graph at https://www.surveymonkey.com/results/SM-5GGGJV5/
Re: SurveyMonkey for D users OS - Results
On Saturday, 31 May 2014 at 13:52:46 UTC, Rikki Cattermole wrote: On 1/06/2014 1:45 a.m., Abdulhaq wrote: On Saturday, 31 May 2014 at 13:37:26 UTC, Abdulhaq wrote: There's been 100 votes and the results are: Linux 64 bits: 53 Linux 32 bits: 4 Windows 64 bits:27 Windows 32 bits: 3 Mac: 7 Other: 6: ArchLinux Android Centos 6 MAC OSX, LINUX 64, Windows 64, FreeBSD 64 bsd64 One 'other' vote was spoiled. It turns out that the free SurveyMonkey account only allows 100 votes max, but the profile has been much the same since 50 votes so I think the ratios are clear. If anyone has an OS other than the ones mentioned above then perhaps they could mention it in this thread. See the graph at https://www.surveymonkey.com/results/SM-5GGGJV5/ I'm personally not surprised by these results. But they will be skewed because of time zones and the limited number of participants. Which is a shame. Not to mention all those who use D plus don't read the NG. Shame it didn't make 24 hrs as all time zones would have been covered, still I think it's probably a pretty fair picture of the whole thing. I'm wondering what's the Linux 32 bit usages - embedded I guess. 64 bits seems to dominate in general. A couple of linux users seem not to know if they are 32 or 64 bit?
SurveyMonkey for D users OS
I've created a SurveyMonkey survey to gather stats on D users OS usage. It only takes a few seconds to answer: https://www.surveymonkey.com/s/3BJRWP8 We can leave it running for a week or so, I'll keep you updated on results. Abdulhaq
Re: SurveyMonkey for D users OS
On Friday, 30 May 2014 at 17:24:51 UTC, Abdulhaq wrote: I've created a SurveyMonkey survey to gather stats on D users OS usage. It only takes a few seconds to answer: https://www.surveymonkey.com/s/3BJRWP8 We can leave it running for a week or so, I'll keep you updated on results. Abdulhaq 12 responses in 12 minutes, keep voting folks it's interesting :-) I won't skew the results by spilling the beans just yet.
Re: SurveyMonkey for D users OS
On Friday, 30 May 2014 at 18:20:21 UTC, Dejan Lekic wrote: Abdulhaq wrote: I've created a SurveyMonkey survey to gather stats on D users OS usage. It only takes a few seconds to answer: https://www.surveymonkey.com/s/3BJRWP8 We can leave it running for a week or so, I'll keep you updated on results. Abdulhaq No offense, but for one question survey you could easily ask here instead of SurveyMonkey. No offense taken. IMO threads tend to go off topic too easily to make them a good tool for surveys. This way makes it very easy to keep focus and gather results. It only took me a few minutes to set it up.
Re: SurveyMonkey for D users OS
On Friday, 30 May 2014 at 18:56:49 UTC, Kagamin wrote: On Friday, 30 May 2014 at 17:39:00 UTC, Abdulhaq wrote: I won't skew the results by spilling the beans just yet. Do you think, users will migrate to another OS just for this survey if they see the results? No I don't think that. What can happen is that people see that 'their team is falling behind' and therefore vote when they otherwise would not be inclined. It can lead to more evenly balanced (and incorrect) results than a blind survey would see (it seems to me).
