<div dir="ltr"><div><div>I'm having problems with the statistics...<br></div><div>I tried Andres' version of allSubclasses and to my surprise it is slower than my version in the Stack VM, that is the one I use for Live Typing, but it is slightly faster with the JIT VM (only a 0.01%)</div><div>I don't see the 20% improvement that you mentioned Andres, what are the number you got?</div><div>These are the numbers I get doing this, three times:</div><div><br></div><div>Smalltalk garbageCollect.<br>Time millisecondsToRun: [ 1000 timesRepeat: [ Behavior allSubclasses ]]. <br></div><div><br></div><div>JIT VM: </div><div>Breath first version: 169 170 171 </div><div>Depth first version: 181 179 180 </div></div><div><br></div><div>Stack VM:</div><div><div><div>Breath first version: 776 780 788</div><div>Depth first version: 749 749 751 </div></div><div><br></div></div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Sat, Aug 24, 2019 at 11:22 AM Juan Vuletich via Cuis-dev <<a href="mailto:cuis-dev@lists.cuis.st">cuis-dev@lists.cuis.st</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Uh, I hadn't seen you had already attached code, and only answered to <br> the comments. Apologies for that!<br> <br> This new version is much faster. Thanks! I integrated it at GitHub, with <br> a short comment about the iterative vs. recursive choice.<br> <br> Cheers,<br> <br> -- <br> Juan Vuletich<br> <a href="http://www.cuis-smalltalk.org" rel="noreferrer" target="_blank">www.cuis-smalltalk.org</a><br> <a href="https://github.com/Cuis-Smalltalk/Cuis-Smalltalk-Dev" rel="noreferrer" target="_blank">https://github.com/Cuis-Smalltalk/Cuis-Smalltalk-Dev</a><br> <a href="https://github.com/jvuletich" rel="noreferrer" target="_blank">https://github.com/jvuletich</a><br> <a href="https://www.linkedin.com/in/juan-vuletich-75611b3" rel="noreferrer" target="_blank">https://www.linkedin.com/in/juan-vuletich-75611b3</a><br> @JuanVuletich<br> <br> <br> <br> On 8/23/2019 7:34 PM, Andres Valloud via Cuis-dev wrote:<br> > See attached for your reference.<br> ><br> > On 8/22/19 05:36, Juan Vuletich wrote:<br> >> On 8/21/2019 4:29 PM, Andres Valloud via Cuis-dev wrote:<br> >>> What do you see as "ugly"?<br> >><br> >> It is way more complicated than needed to solve the problem. The <br> >> problem is trivial, this code is not. And there is no reason for <br> >> this: class hierarchy depths shouldn't be so deep to be a challenge <br> >> for the call stack.<br> >><br> >> Hernán's code is only as complex as the problem, without accidental <br> >> complexity (in the Fred Brooks sense). It is a clear explanation of <br> >> the problem and the solution.<br> >><br> >> To me, code aesthetics is not only about code being clear, and a good <br> >> explanation of the problem and the solution. It is also about being <br> >> the best fit for the problem (including human readers here: the <br> >> system is the curricula). Sometimes, performance requirements or <br> >> space constraints require additional complexity. Then, I see that <br> >> extra complexity as pretty. This is of course subjective, like any <br> >> aesthetic judgment.<br> >><br> >>> One detail I don't see mentioned here: the old code filled the <br> >>> ordered collection breadth first (before the set conversion), and so <br> >>> it has limited execution stack depth. The current code fills the <br> >>> collection depth first, so its execution stack depth is as deep as <br> >>> the hierarchy tree.<br> >>><br> >>> Maybe there was a reason why the users of allSubclasses were better <br> >>> off with breadth first, before someone put in asSet. Or perhaps <br> >>> asSet was always there, I don't see why that was beneficial.<br> >>><br> >>> In any case, I tried both the recursive and (properly implemented) <br> >>> iterative methods: breadth first is >20% faster.<br> >>><br> >>> On 8/21/19 05:44, Juan Vuletich via Cuis-dev wrote:<br> >>>> Wow, thanks. The old code was slow and ugly!<br> >>>><br> >>>> Integrating it.<br> >>>><br> >>>> Cheers,<br> >>>><br> >>>> -- <br> >>>> Juan Vuletich<br> >>>> <a href="http://www.cuis-smalltalk.org" rel="noreferrer" target="_blank">www.cuis-smalltalk.org</a><br> >>>> <a href="https://github.com/Cuis-Smalltalk/Cuis-Smalltalk-Dev" rel="noreferrer" target="_blank">https://github.com/Cuis-Smalltalk/Cuis-Smalltalk-Dev</a><br> >>>> <a href="https://github.com/jvuletich" rel="noreferrer" target="_blank">https://github.com/jvuletich</a><br> >>>> <a href="https://www.linkedin.com/in/juan-vuletich-75611b3" rel="noreferrer" target="_blank">https://www.linkedin.com/in/juan-vuletich-75611b3</a><br> >>>> @JuanVuletich<br> >>>><br> >>>><br> >>>> On 8/21/2019 9:26 AM, Hernan Wilkinson via Cuis-dev wrote:<br> >>>>> Hi,<br> >>>>> attached is a cs that changes the implementation of <br> >>>>> #allSubclasses and #withAllSubclasses and generates performance <br> >>>>> improvement drastically.<br> >>>>> Just as a quick example, with the current implementation:<br> >>>>> Time millisecondsToRun: [ 100 timesRepeat: [Behavior <br> >>>>> withAllSubclasses ]] 1140 .<br> >>>>> With the proposed change:<br> >>>>> Time millisecondsToRun: [ 100 timesRepeat: [Behavior <br> >>>>> withAllSubclasses2 ]] . 73<br> >>>>><br> >>>>> In the type checker of LiveTyping, checking Behavior goes from <br> >>>>> 9.5 seconds to 5 seconds only with that change.<br> >>>>> One difference between the current and proposed version, is that <br> >>>>> the current returns a Set while the proposed version returns an <br> >>>>> OrderedCollection.<br> >>>>> I looked carefully if that could break something and it does not. <br> >>>>> I've been using the image with this new implementation and <br> >>>>> everything works fine.<br> >>>>><br> >>>>> Juan, please take a look at it and integrate it if you think it <br> >>>>> is useful.<br> >>>>> Hernan.<br> >>>>><br> >>>>> -- <br> >>>>> *Hernán Wilkinson<br> >>>>> Agile Software Development, Teaching & Coaching*<br> >>>>> *Phone: +54-011*-4893-2057<br> >>>>> *Twitter: @HernanWilkinson*<br> >>>>> *site: <a href="http://www.10Pines.com" rel="noreferrer" target="_blank">http://www.10Pines.com</a> <<a href="http://www.10pines.com/" rel="noreferrer" target="_blank">http://www.10pines.com/</a>>*<br> >>>>> Address: Alem 896, Floor 6, Buenos Aires, Argentina<br> >>>><br> >><br> >><br> <br> <br> -- <br> Cuis-dev mailing list<br> <a href="mailto:Cuis-dev@lists.cuis.st" target="_blank">Cuis-dev@lists.cuis.st</a><br> <a href="https://lists.cuis.st/mailman/listinfo/cuis-dev" rel="noreferrer" target="_blank">https://lists.cuis.st/mailman/listinfo/cuis-dev</a><br> </blockquote></div><br clear="all"><div><br></div>-- <br><div dir="ltr" class="gmail_signature"><div dir="ltr"><div><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div><span style="font-family:tahoma,sans-serif;font-size:xx-small;border-collapse:collapse"><strong><span style="font-size:8pt"><span><span style="font-size:small"><font size="2"><span style="font-weight:normal"><span style="font-weight:bold">Hernán Wilkinson</span><br>Agile Software Development, Teaching & Coaching</span></font></span></span></span></strong></span></div><div><span style="font-family:tahoma,sans-serif;font-size:xx-small;border-collapse:collapse"><strong><span style="font-size:8pt"><span><span style="font-size:small"><font size="2"><span style="font-weight:normal">Phone: +54-011</span></font></span></span></span></strong></span><font face="tahoma, sans-serif" size="2">-4893-2057</font></div><div><strong style="font-family:tahoma,sans-serif;font-size:xx-small"><span style="font-size:8pt"><span style="font-size:small"><font size="2"><span style="font-weight:normal">Twitter: @HernanWilkinson</span></font></span></span></strong></div><div><span style="font-family:tahoma,sans-serif;font-size:xx-small;border-collapse:collapse"><strong><span style="font-size:8pt"><span><span style="font-size:small"><font size="2"><span style="font-weight:normal">site: <a href="http://www.10pines.com/" style="color:rgb(17,65,112)" target="_blank">http://www.10Pines.com</a></span></font></span></span></span></strong></span></div><div><font face="tahoma, sans-serif"><span style="border-collapse:collapse">Address: Alem 896</span></font>, Floor 6, Buenos Aires, Argentina</div></div></div></div></div></div></div></div></div></div></div></div>