<?xml version="1.0" encoding="utf-8"?>
<feed xml:lang="en" xmlns="http://www.w3.org/2005/Atom"><title>Recent changes to feature-requests</title><link href="https://sourceforge.net/p/wincachegrind/feature-requests/" rel="alternate"/><link href="https://sourceforge.net/p/wincachegrind/feature-requests/feed.atom" rel="self"/><id>https://sourceforge.net/p/wincachegrind/feature-requests/</id><updated>2011-12-11T19:42:18Z</updated><subtitle>Recent changes to feature-requests</subtitle><entry><title>New sort option: Time relative to heaviest child.</title><link href="https://sourceforge.net/p/wincachegrind/feature-requests/4/" rel="alternate"/><published>2011-12-11T19:42:18Z</published><updated>2011-12-11T19:42:18Z</updated><author><name>Anonymous</name><uri>https://sourceforge.net/u/userid-None/</uri></author><id>https://sourceforge.net8c6ace37fcb368708d7a2f3ea271bc53a942274e</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;I would like to see a new sort option, that really helps to find those function calls that are the most relevant to look at for optimization.&lt;/p&gt;
&lt;p&gt;Example 1:&lt;br /&gt;
foo() needs 510ms to run.&lt;br /&gt;
foo() calls bar(), which needs 500ms to run.&lt;br /&gt;
-&amp;gt; foo() is not interesting to look at, because the interesting stuff happens in bar().&lt;/p&gt;
&lt;p&gt;Example 2:&lt;br /&gt;
foo() needs 510 ms to run.&lt;br /&gt;
foo() calls bar_a(), which needs 200 ms.&lt;br /&gt;
foo() also calls bar_b(), which needs 300 ms.&lt;br /&gt;
-&amp;gt; foo() _is_ interesting to look at, because a big part of the duration is "branched" between two child function calls.&lt;/p&gt;
&lt;p&gt;Example 3:&lt;br /&gt;
foo() needs 510 ms to run.&lt;br /&gt;
foo() calls bar() 50 times, where each call needs 10 ms.&lt;br /&gt;
-&amp;gt; foo() _is_ interesting to look at. We can ask ourselves why bar() has to run so often.&lt;/p&gt;
&lt;p&gt;Example 4:&lt;br /&gt;
foo() needs 510 ms to run.&lt;br /&gt;
foo() calls bar(), but bar() only needs 100 ms.&lt;br /&gt;
the other 410 ms are in foo() itself.&lt;br /&gt;
-&amp;gt; foo() is interesting to look at, because it probably contains an insane loop.&lt;/p&gt;
&lt;p&gt;Example 5:&lt;br /&gt;
foo() needs 510 ms to run.&lt;br /&gt;
foo() calls itself 2x, and each call needs 250 ms. What follows is a big recursion.&lt;br /&gt;
-&amp;gt; foo() is interesting to look at.&lt;/p&gt;
&lt;p&gt;Example 6:&lt;br /&gt;
foo() needs 510 ms to run.&lt;br /&gt;
foo() calls itself 1x, and this call needs 500 ms. What follows is a big recursion.&lt;br /&gt;
-&amp;gt; Duh. We want to spot the recursion, but which instance of foo() should we look at? Anyway, this is somewhat interesting to look at.&lt;/p&gt;
&lt;p&gt;So, what is a good sort measure, that will push foo to the top of the list, for Examples 2-5, but will put it to the bottom of the list for Example 1.&lt;br /&gt;
Existing sort options fail, imo.&lt;/p&gt;
&lt;p&gt;--------&lt;/p&gt;
&lt;p&gt;Proposed measure:&lt;br /&gt;
Take the duration of the function call to foo(), and subtract the duration of the longest-enduring child call.&lt;br /&gt;
Accumulate this value, for all calls to foo().&lt;/p&gt;
&lt;p&gt;Example results:&lt;br /&gt;
Ex. 1) 10 ms for single call&lt;br /&gt;
Ex. 2) 200 ms for single call&lt;br /&gt;
Ex. 3) 500 ms for single call&lt;br /&gt;
Ex. 4) 410 ms for single call&lt;br /&gt;
Ex. 5) 260 ms for single call, + more for recursive calls.&lt;br /&gt;
Ex. 6) 10 ms for single call, + a lot more for recursive calls.&lt;/p&gt;
&lt;p&gt;One could even invent something new for the recursion stuff, but then we'd also need to consider indirect recursion etc.&lt;br /&gt;
So we should rather stick with this simple measure, and see where it leads us.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>(s)ftp file opening</title><link href="https://sourceforge.net/p/wincachegrind/feature-requests/3/" rel="alternate"/><published>2008-03-19T10:39:01Z</published><updated>2008-03-19T10:39:01Z</updated><author><name>Anonymous</name><uri>https://sourceforge.net/u/userid-None/</uri></author><id>https://sourceforge.netf0c6b594d9fd182eb402ce69bb0780f330a7e817</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Is it possible to use (s)FTP connections to open up profile data?&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>read aggregate profile output</title><link href="https://sourceforge.net/p/wincachegrind/feature-requests/2/" rel="alternate"/><published>2007-08-05T20:30:34Z</published><updated>2007-08-05T20:30:34Z</updated><author><name>Anonymous</name><uri>https://sourceforge.net/u/userid-None/</uri></author><id>https://sourceforge.net9c851f1b051e54e2ae12fa385fa1efdbf1ade7fb</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;XDebug 2.0.0 allows appending multiple 'profile sessions' to one output file. This allows me to profile over a longer period of time. Unfortunately, wincachegrind (and also kcachegrind) is unable to read these files (only the first profile is parsed).&lt;/p&gt;
&lt;p&gt;The output looks like:&lt;/p&gt;
&lt;p&gt;==== NEW PROFILING FILE ==============================================&lt;br /&gt;
version: 0.9.6&lt;br /&gt;
cmd: /home/kvdb/vodw/index.php&lt;br /&gt;
part: 1&lt;/p&gt;
&lt;p&gt;events: Time&lt;/p&gt;
&lt;p&gt;.&lt;br /&gt;
.&lt;br /&gt;
.&lt;/p&gt;
&lt;p&gt;==== NEW PROFILING FILE ==============================================&lt;br /&gt;
version: 0.9.6&lt;br /&gt;
cmd: /home/kvdb/vodw/getfile.php&lt;br /&gt;
part: 1&lt;/p&gt;
&lt;p&gt;.&lt;br /&gt;
.&lt;br /&gt;
.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>can WinCacheGrind generate call-graph like KCacheGrind?</title><link href="https://sourceforge.net/p/wincachegrind/feature-requests/1/" rel="alternate"/><published>2007-07-13T03:54:59Z</published><updated>2007-07-13T03:54:59Z</updated><author><name>Gemate King</name><uri>https://sourceforge.net/u/userid-1613575/</uri></author><id>https://sourceforge.netdbf7a9a08b52f4a0c142adb988a30bc073643e37</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;It will be very helpful if WinCacheGrind can generate call-graph like KCacheGrind.&lt;br /&gt;
Maybe we only need a bridge between cachegrind.xxx =&amp;gt; callgrind.xxx =&amp;gt; .dot script?&lt;/p&gt;&lt;/div&gt;</summary></entry></feed>