www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - Freelists and clear/emplace

reply d coder <dlang.coder gmail.com> writes:
Greetings

Is emplace/clear  mechanism mature enough to be used to create freelists?

I looked but found very scanty documentation on emplace/clear on dlang.org.

Regards
- Puneet
Jun 15 2012
next sibling parent reply Timon Gehr <timon.gehr gmx.ch> writes:
On 06/16/2012 03:07 AM, d coder wrote:

 Greetings

 Is emplace/clear  mechanism mature enough to be used to create freelists?

 I looked but found very scanty documentation on emplace/clear on
 dlang.org <http://dlang.org>.

 Regards
 - Puneet


I create most of my objects with emplace and have not experienced any 
trouble.

Make sure to keep your allocations aligned, though (16 bytes are sufficient)

http://d.puremagic.com/issues/show_bug.cgi?id=6635
Jun 16 2012
parent Dmitry Olshansky <dmitry.olsh gmail.com> writes:
On 16.06.2012 15:51, Timon Gehr wrote:

 On 06/16/2012 03:07 AM, d coder wrote:

 Greetings

 Is emplace/clear mechanism mature enough to be used to create freelists?

 I looked but found very scanty documentation on emplace/clear on
 dlang.org <http://dlang.org>.

 Regards
 - Puneet





 I create most of my objects with emplace and have not experienced any
 trouble.


+1. And I used freelists with raw cast T* and had no trouble either.
(these Ts  were PODs, of course)


 Make sure to keep your allocations aligned, though (16 bytes are
 sufficient)

 http://d.puremagic.com/issues/show_bug.cgi?id=6635



-- 
Dmitry Olshansky
Jun 16 2012
prev sibling parent reply "F i L" <witte2008 gmail.com> writes:
On Saturday, 16 June 2012 at 01:14:06 UTC, d coder wrote:

 Greetings

 Is emplace/clear  mechanism mature enough to be used to create 
 freelists?

 I looked but found very scanty documentation on emplace/clear 
 on dlang.org.

 Regards
 - Puneet


I did performance tests awhile ago and emplace() was virtually 
identical to raw assignment. I'm not sure how familiar you are 
with D, but personally I prefer D's mixin templates for freelists:

   mixin template Pool(T) { // FreeList
     static T poolHead;
     public T poolNext;

     static auto create() { ... }
     static void delete() { ... }
   }


   class MyClass {
     mixin Pool!MyClass;
   }

   struct MyStruct {
     mixin Pool!MyStruct;
   }


   void main() {
     auto mc = MyClass.create();
     auto ms = MyStruct.create();

     ...

     mc.delete();
     ms.delete();
   }
Jun 16 2012
next sibling parent "F i L" <witte2008 gmail.com> writes:
^^ Typo, should be:

   struct MyStruct {
     mixin Pool!MyStruct*;
   }
Jun 16 2012
prev sibling parent reply Timon Gehr <timon.gehr gmx.ch> writes:
On 06/16/2012 09:34 PM, F i L wrote:

 On Saturday, 16 June 2012 at 01:14:06 UTC, d coder wrote:

 Greetings

 Is emplace/clear  mechanism mature enough to be used to create freelists?

 I looked but found very scanty documentation on emplace/clear on
 dlang.org.

 Regards
 - Puneet


 I did performance tests awhile ago and emplace() was virtually identical
 to raw assignment. I'm not sure how familiar you are with D, but
 personally I prefer D's mixin templates for freelists:

    mixin template Pool(T) { // FreeList
      static T poolHead;
      public T poolNext;

      static auto create() { ... }
      static void delete() { ... }
    }


    class MyClass {
      mixin Pool!MyClass;
    }

    struct MyStruct {
      mixin Pool!MyStruct;
    }


    void main() {
      auto mc = MyClass.create();
      auto ms = MyStruct.create();

      ...

      mc.delete();
      ms.delete();
    }


'typeof(this)' can be used to avoid stuttering the type at the mixin 
location.

mixin template Pool() { mixin Pool!(typeof(this)); }

class MyOtherClass {
     mixin Pool;
}
Jun 16 2012
parent reply "F i L" <witte2008 gmail.com> writes:
Timon Gehr wrote:

 'typeof(this)' can be used to avoid stuttering the type at the 
 mixin location.

 mixin template Pool() { mixin Pool!(typeof(this)); }

 class MyOtherClass {
     mixin Pool;
 }


For some reason I was under the impression that typeof(this) 
couldn't be resolved in this situation (no 'this' value), but I'm 
sure you're correct here. Which of course means you'd need to 
"static if(isValueType!typeof(this)) { ... } else { ... }" to 
make it work with structs.
Jun 16 2012
parent reply Timon Gehr <timon.gehr gmx.ch> writes:
On 06/16/2012 09:55 PM, F i L wrote:

 Timon Gehr wrote:

 'typeof(this)' can be used to avoid stuttering the type at the mixin
 location.

 mixin template Pool() { mixin Pool!(typeof(this)); }

 class MyOtherClass {
     mixin Pool;
 }


 For some reason I was under the impression that typeof(this) couldn't be
 resolved in this situation (no 'this' value),


It is indeed special cased.

class S{
     pragma(msg, typeof(this)); // ok
     pragma(msg, this.stringof);
  // Error: 'this' is only defined in non-static member functions, not S
}


 but I'm sure you're
 correct here. Which of course means you'd need to "static
 if(isValueType!typeof(this)) { ... } else { ... }" to make it work with
 structs.
Jun 16 2012
parent Timon Gehr <timon.gehr gmx.ch> writes:
On 06/16/2012 10:02 PM, Timon Gehr wrote:

 On 06/16/2012 09:55 PM, F i L wrote:

 Timon Gehr wrote:

 'typeof(this)' can be used to avoid stuttering the type at the mixin
 location.

 mixin template Pool() { mixin Pool!(typeof(this)); }

 class MyOtherClass {
     mixin Pool;
 }


 For some reason I was under the impression that typeof(this) couldn't be
 resolved in this situation (no 'this' value),


 It is indeed special cased.

 class S{
      pragma(msg, typeof(this)); // ok
      pragma(msg, this.stringof);
   // Error: 'this' is only defined in non-static member functions, not S
 }


That example actually compiles, but I'd consider that a bug. Removing 
".stringof" makes it fail as expected.
Jun 16 2012