Mid-life crisis

Avoiding Expensive Garbage Collections in Asynchrony

Ken Brubaker — Thu, 04 Dec 2003 23:55:00 GMT

re: Mid-life crisis

Ian Griffiths — Fri, 05 Dec 2003 18:28:00 GMT

I'm surprised to learn that setting variables to null helps so much. My experiments seem to suggest that the garbage collector is doing liveness analysis (or is taking advantage of the liveness analysis performed by the JIT). So with something like this:

MyObj o = new MyObj();
o.DoSomethingWithO();
BlockForAges();
return;

the GC usually seems to work out that 'o' isn't used after that second line, and will happily collect the object (or at least it'll run its Finalize if I put a finalizer in there as an experiment to see what the GC is doing) if a GC occurrs during BlockForAges. Setting o to null doesn't seem to make any difference, since the liveness analysis has already worked out that o is effectively out of scope.

Are there situations in which this liveness analysis is defeated? Or is this not the scenario you're thinking of when you say to set things to null?

re: Mid-life crisis

Rico Mariani — Fri, 05 Dec 2003 18:55:00 GMT

You're absolutely right, if the JIT can statically determine that the variable is dead when the code is generated then there's no need null things out. However, what often happens in servers is that there are certain transaction state variables that are, for instance, on the "this" pointer which are still reachable. Those are the ones to null if you can.

For instance, suppose you got your input in XML format and you had a series of functions to extract what you need to do the query out of the XML, building up a SQL query string as you go. Just before you make the SQL query, it would be good to release all the stuff related to the XML that you can so that it can be collected. Objects like that are often fields accessable via your "this" pointer rather than local variables, so they are reachable until the object holding them goes away.

Other times there are helper collection classes that assist in the parsing and validation of the inputs. These objects also need to survive across several function calls (they are often accumulating results as process goes along), again these would seem live to the collector but perhaps they can be nulled or emptied.

re: Mid-life crisis

Rico Mariani — Fri, 05 Dec 2003 19:10:00 GMT

Let me alter your example just a tad, this is a stupid example using intermediate strings just to illustrate

class MyObj
{
private String s1;
private String s2;
private String s3;

public void DoSomethingFromInputs(String s)
{
...
s1 = AnElegantOperation(s);
...
}

public void ComputeInterimResults(String options)
{
...
s2 = SomethingEvenMoreElegant(s1, options);
...
}

public void ComputerFinalQuery(String database)
{
...
s3 = SQLFormatting(s2, database);
...
}

public String GetResults()
{
...
s1 = null; // this is what I'm talking about
s2 = null; // this is what I'm talking about
// this blocks a long time
String r = GetDataFromDatabase(s3)
...
return r;
}
}

MyObj o = new MyObj();
o.DoSomethingFromInputs(s);
o.ComputeInterimResults(options);
o.ComputeFinalQuery(databasename);
return o.GetResults();

(please forgive my syntax, I hope you can get the jist)

re: Mid-life crisis

Jerry Dennany — Sat, 06 Dec 2003 04:49:00 GMT

Rico, thanks for clearing that up - I also shared Ken's concerns when reading your original post.

GC Performance

Code and Stuff — Wed, 11 Feb 2004 17:36:00 GMT

GC Performance

Code and Stuff — Wed, 11 Feb 2004 17:36:00 GMT

GC Performance

Code 'n' Stuff — Wed, 25 Feb 2004 03:10:00 GMT

re: XmlSerializer and Stale UI's

Code 'n' Stuff — Sat, 28 Feb 2004 10:58:00 GMT

GC Performance

Extreme RAD from a Trading Desk — Thu, 11 Mar 2004 11:34:00 GMT

A Few Good GC Links

Chris Lyon's WebLog — Tue, 14 Sep 2004 20:23:00 GMT

A Few Good GC Links

Chris Lyon's WebLog — Tue, 14 Sep 2004 21:40:00 GMT

When to call GC.Collect()

Rico Mariani's Performance Tidbits — Mon, 29 Nov 2004 23:41:00 GMT

re: When to call GC.Collect()

Rico Mariani's Performance Tidbits — Tue, 30 Nov 2004 21:04:00 GMT

Top Ten reasons you should subscribe to my blog

Rico Mariani's Performance Tidbits — Mon, 13 Dec 2004 23:00:00 GMT

.net Garbage Collector Tips and Tricks

Tony's Blog — Fri, 17 Dec 2004 06:40:00 GMT

.net Garbage Collector Tips and Tricks

Tony's Blog — Tue, 28 Dec 2004 05:18:00 GMT

Performance Quiz #5: The performance cost of the garbage collector : Solution

Rico Mariani's Performance Tidbits — Fri, 07 Jan 2005 22:56:00 GMT

.NET and

Mike Taulty's Weblog — Tue, 01 Feb 2005 11:54:00 GMT

re: Common Sources of Processor Performance Penalties: Five Issues

Rico Mariani's Performance Tidbits — Thu, 03 Mar 2005 20:11:00 GMT

The Tale of two garbage collectors.

Josh Carlisle - .Net Brain Freeze — Thu, 10 Mar 2005 19:00:00 GMT

The Tale of Citrix Development and Two Garbage Collectors.

Josh Carlisle - .Net Brain Freeze — Thu, 10 Mar 2005 19:57:00 GMT

The Tale of Citrix Development and Two Garbage Collectors.

Josh Carlisle - .Net Brain Freeze — Thu, 10 Mar 2005 19:57:00 GMT

Three techniques for tracking down leaks due to undisposed objects

Rico Mariani's Performance Tidbits — Fri, 15 Apr 2005 20:51:53 GMT

People often ask me for tips/tricks on how to find out which objects are not being properly disposed.&nbsp;&nbsp;But...

Tame the CLR: Expect game-quality real-time performance from managed code

Robert Burke's Weblog — Thu, 24 Aug 2006 19:26:24 GMT

Game development is one of those dark arts where the usual laws of scalability don't always apply.&nbsp;...

On CLR Memory Management: Problem Solved…? « The Compulsive Coder

On CLR Memory Management: Problem Solved…? « The Compulsive Coder — Thu, 14 Dec 2006 18:23:36 GMT

PingBack from http://compulsivecoder.com/caffeine/?p=15

.NET Garbage Collector PopQuiz - Followup at Sanal Kiler

.NET Garbage Collector PopQuiz - Followup at Sanal Kiler — Sun, 15 Apr 2007 23:07:07 GMT

PingBack from http://sanal.org/?p=309

.NET Garbage Collector PopQuiz - Followup

If broken it is, fix it you should — Tue, 17 Apr 2007 12:14:16 GMT

It was really exciting to see that so many people answered the .NET GC PopQuiz , especially seeing that

Top 20 .NET Garbage Collection (GC) Articles

roy ashbrook — Fri, 22 Jun 2007 07:41:13 GMT

Ah. Garbage Collection... how I love and hate thee. =P I think one sad thing about programming in .net

Top 20 .NET Garbage Collection (GC) Articles

roy ashbrook — Fri, 22 Jun 2007 07:43:21 GMT

Ah. Garbage Collection... how I love and hate thee. =P I think one sad thing about programming in .net

Top 20 .NET Garbage Collection (GC) Articles

roy ashbrook — Fri, 22 Jun 2007 08:36:43 GMT

Ah. Garbage Collection... how I love and hate thee. =P I think one sad thing about programming in .net

.Net Performance pointers

Dotmad (on .Net) — Fri, 14 Mar 2008 15:33:54 GMT

After going this week to the Microsoft performance open house , here are few things to consider: Create

The CLR garbage collector is in need of help

C# Nuggets — Thu, 12 Jun 2008 14:48:41 GMT

Back in 2000 when the CLR was first shown it's generational garbage collector was fairly cutting

[??????]??????Top 20 .NET Garbage Collection (GC) ?????? : ????????????????????? TAHETA.ORG

Tue, 23 Dec 2008 15:35:21 GMT

PingBack from http://www.taheta.org/?p=94

James Kovacs' Weblog - Debug Leaky Apps: Identify and Prevent Memory Leaks in Managed Code

Fri, 26 Dec 2008 09:09:15 GMT

PingBack from http://www.jameskovacs.com/blog/DebugLeakyAppsIdentifyAndPreventMemoryLeaksInManagedCode.aspx