Databases
StackOverflow2013
Postsdbo
POSimple caching makes GC go crazy?
 

Note that there are some explanatory texts on larger screens.

plurals
  1. POSimple caching makes GC go crazy?
    primarykey
    data
    text
    <p>I'm using C# 3.5 and VS 2010 ultimate.</p> <p>I'm optimizing (for speed) a machine learning algorithm that has four nested for-loops. I've found that a simple cache (a tensor of objects) could potentially greatly improve the performance, because there's a lot of re-allocating of the same objects.</p> <p>Here's the before and after implementations.</p> <p>Before:</p> <pre><code>four nested for-loops: var object = new object(3 parameters); Calculate(object, other params) </code></pre> <p>After:</p> <pre><code>var cache = new object[count1,count2,count3]; three nested for-loops: cache[param1, param2, param3] = new object(3 params); four nested for-loops: var object = cache[3 parameters]; Calculate(object, other params) </code></pre> <p>I've profiled both methods and the "before" version was quite faster with ~18% time spent in GC, while the "after" version spend ~88% in GC. It's obvious that the addition of this cache made the GC activity rise, but I'm failing to see how this is possible. </p> <p>I am using many long-lived object in the application. My machine wasn't under heavy load at the time of profiling. The tensor was implemented using multi-dimensional array (not jagged arrays). The inner-most loop in both methods above is implemented using the <code>Parallel.For</code> construct, where right before the loop I am allocating a small <code>double</code> array.</p> <p>How can I reduce the time spent in GC? </p> <p>EDIT #1: The results are indeed from Release mode.</p> <p>EDIT #2: Real code of the after method's four for-loop:</p> <pre><code>List&lt;int&gt; labels = // count = 80 List&lt;int&gt; tokens = // count = 35 var table = new double[tokens.Count, labels.Count, labels.Count]; var cachedObjects = new CachedObject[tokens.Count, labels.Count, labels.Count]; for (int k = 0; k &lt; tokens.Count; k++) { foreach (var tagCurrent in labels) { foreach (var labelBack in labels) { double[] value = new double[labels.Count]; Parallel.For(0, labels.Count, (i) =&gt; { CachedObject CachedObject = cachedObjects[k, labelsBackFurther[i], labelBack]; var me = ModelEstimate(vOptimal, CachedObject, tagCurrent, labels); value[i] = table[k - 1, labels[i], labelBack] * me; }); var maxValue = 0; var maxTagIdx = 0; for (int j = 0; j &lt; value.Length; j++) { var item = value[j]; if (item &gt; maxValue) { maxValue = item; maxTagIdx = j; } } table[k, labelBack, tagCurrent] = maxValue; } } } </code></pre>
    singulars
    1. PO
      singulars
      1. PTAnswer
    1. This table or related slice is empty.
    1. PTQuestion
    1. USHowie
    1. USHowie
    plurals
    1. PL
      singulars
      1. LTLinked
    1. This table or related slice is empty.
    1. This table or related slice is empty.
    1. PO
      singulars
      1. PTAnswer
    2. PO
      singulars
      1. PTAnswer
    3. PO
      singulars
      1. PTAnswer
    1. VO
      singulars
      1. POSimple caching makes GC go crazy?
      1. This table or related slice is empty.
      1. VTUpMod
    2. VO
      singulars
      1. POSimple caching makes GC go crazy?
      1. This table or related slice is empty.
      1. VTUpMod
    3. VO
      singulars
      1. POSimple caching makes GC go crazy?
      1. This table or related slice is empty.
      1. VTDownMod
    1. COhttp://stackoverflow.com/questions/1132033/reasons-for-seeing-high-time-in-gc-in-perf-mon
      singulars
      1. POSimple caching makes GC go crazy?
      1. USIt'sNotALie.
    2. COIt would be easier to help you if you'd show real code - simplified to remove extraneous things, but real in terms of everything important. The pseudo-code is a bit ambiguous.
      singulars
      1. POSimple caching makes GC go crazy?
      1. USJon Skeet
    3. CO@JonSkeet is certainly right here. One other thing to note is that if you profiled this in `Debug` configuration all numbers are irrelevant. Please make sure that the numbers you provided were in fact snapshots of a `Release` build with optimizations included.
      singulars
      1. POSimple caching makes GC go crazy?
      1. USMike Perrenoud
 

Querying!

 
Guidance
A row detail

Detail views are divided into sections. All the information in the data section comes from columns in the selected row. The other sections display data from other, related rows.

Related data can be related in a to-one or a to-many fashion. Captions of data related in a to-many fashion link to a list view showing a filtered view of the table.

Try moving around until you find a non-empty to-many entry and click on the label to get to one. You can move back to the root by clicking on the database name in the header.

SQuiL has stopped working due to an internal error.

If you are curious you may find further information in the browser console, which is accessible through the devtools (F12).

Reload