December 2017 - Pharo-dev - lists.pharo.org

OpalCompiler evaluate speed
by Nicolas Cellier Feb. 15, 2018

Feb. 15, 2018

I have an ArbitraryPrecisionFloatTests doing an exhaustive test for printing and reevaluating all positve half precision float. That's about 2^15 or approximately 32k loop which evaluate snippets like (ArbitraryPrecisionFloat readFrom: '1.123' readStream numBits: 10) The test was naively written with Compiler evaluate: and was using the legacy Compiler. If I rewrite self class compiler evaluate: the test times out. Let's see what increase is necessary: [ ArbitraryPrecisionFloatTest new testPrintAndEvaluate ] timeToRun. -> 3s with legacy Compiler -> 14s with OpalCompiler It's not unexpected that intermediate representation (IR) reification has a cost, but here the 4.5x is a bit too much... This test did account for 1/4 of total test duration already (3s out of 12s). With Opal, the total test duration doubles... (14s out of 23s) So let's analyze the hot spot with: MessageTally spyOn: [ ArbitraryPrecisionFloatTest new testPrintAndEvaluate ]. (I didn't use AndreasSystemProfiler becuase outputs seems a bit garbbled, no matter since the primitives do not account that much, a MessageTally will do the job) I first see a hot spot which does not seem that necessary: | |24.6% {3447ms} RBMethodNode(RBProgramNode)>>formattedCode >From the comments I understand that AST-based stuff requires a pattern (DoIt) and an explicit return (^), but this expensive formatting seems too much for just evaluating. i think that we should change that. Then comes: | |20.7% {2902ms} RBMethodNode>>generate: which is split in two halves, ATS->IR and IR->bytecode | | |9.3% {1299ms} RBMethodNode>>generateIR | | | |11.4% {1596ms} IRMethod>>generate: But then I see this cost a 2nd time which also leave room for progress: | |10.9% {1529ms} RBMethodNode>>generateIR | | |12.9% {1814ms} IRMethod>>generate: The first is in RBMethodNode>>generateWithSource, the second in OpalCompiler>>compile Last comes the parse time (sourceCode -> AST) | 13.2% {1858ms} OpalCompiler>>parse Along with semantic analysis | 6.0% {837ms} OpalCompiler>>doSemanticAnalysis ----------------------------------- For comparison, the legacy Compiler decomposes into: | |61.5% {2223ms} Parser>>parse:class:category:noPattern:context:notifying:ifFail: which more or less covers parse time + semantic analysis time. That means that Opal does a fair work for this stage. Then, the direct AST->byteCode phase is: | 16.9% {609ms} MethodNode>>generate IR costs almost a 5x on this phase, but we know it's the price to pay for the additional features that it potentially offers. If only we would do it once... And that's all for the legacy one... -------------------------------------- This little exercize shows that a 2x acceleration of OpalCompiler evaluate seems achievable: - simplify the uselessely expensive formatted code - generate bytecodes once, not twice Then it will be a bit more 2x slower than legacy, which is a better trade for yet to come superior features potentially brought by Opal. It would be interesting to carry same analysis on method compilation

6 12

Weird ZnClient benchmarking result
by Sven Van Caekenberghe Feb. 6, 2018

Feb. 6, 2018

Hi, While benchmarking ZnClient I encountered something weird and counter intuitive that I cannot explain. This was on macOS running Pharo 7 and 6. First you need to setup a fast local HTTP server, can be anything, for example: $ brew install nginx $ brew services start nginx $ curl http://localhost:8080 In Pharo, the benchmark code looks as follows: ZnClient new in: [ :client | [ client get: 'http://localhost:8080' ] bench ]. On my machine this gives about 2500 req/s. ZnClient uses object logging: it emits special event objects that contain information regarding what happens (full details and timing information) using an Announcer. Subscribers can then decide what to do with this detailed information. This mechanism is always active, but there is an option to turn it totally off for maximum performance: ZnClient new in: [ :client | client loggingOff. [ client get: 'http://localhost:8080' ] bench ]. This gives about 3000 req/s, a non-trivial + 500 req/s improvement. Now comes the weird thing. GT contains a custom view for Announcers, inspect: ZnLogEvent announcer. And select the Announcements tab. Leave this window open. Now, the first expression (with full logging enabled and an active subscriber with a UI) is way FASTER than first, it yields about 7000 to 8000 req/s. Since the benchmark runs in the main UI thread, no actual drawing happens until the expression finishes. Running both expression inside a Time profiler show how in the first case significant time (actually *most* time ;-) is spent in the logging: In the second case, with logging disabled, that changes to: One has to take into account that the request has to be processed by the server, so from the client's perspective, most time will always be spent waiting for a reply, #readResponse. With the logging enabled, probably less time is spent waiting, since the client does a bit more work. Still what should be slower is more than 2 times as fast ! I have a feeling that this has to do with scheduling (probably interacting with network IO), as going deep into the last #readResponse above, I see: But I still cannot understand why doing apparently more work results in faster throughput. Can anyone explain this behaviour ? Thx, Sven

6 8

nil suspend
by Aliaksei Syrel Feb. 5, 2018

Feb. 5, 2018

Hi, I was looking at nil (already funny) and found a very interesting method: suspend "Kills off processes that didn't terminate properly" "Display reverse; reverse." "<-- So we can catch the suspend bug" Processor terminateActive It means that the following code is probably the shortest way to kill your image (save first!) nil suspend It terminates a UI process if evaluated from anywhere using CMD+d :D Cheers, Alex

5 4

[ANN] Fog - Ethereum driver
by Santiago Bragagnolo Jan. 30, 2018

Jan. 30, 2018

Hi all. Im happy to announce a pre release of the Fog ethereum driver that we develop in the space of an Inria project. It still not complete but is already usable for some experiments and simple projects. You can downloadit from https://github.com/sbragagnolo/Fog/ (https://github.com/sbragagnolo/Fog/releases/tag/v0.1.1.1) Dependencies <https://github.com/sbragagnolo/Fog#rhash>RHash sudo apt-get install rhash <https://github.com/sbragagnolo/Fog#solidity>Solidity npm install solc <https://github.com/sbragagnolo/Fog#download-code>Download code <https://github.com/sbragagnolo/Fog#iceberg--baseline>Iceberg / Baseline Metacello new baseline: 'Fog'; repository: 'github://sbragagnolo/Fog:v0.1.1.1/src'; load. <https://github.com/sbragagnolo/Fog#by-hand>By hand You may want to use this version for having access to some scripts and contracts samples. git checkout git@github.com:sbragagnolo/Fog.git git checkout v0.1.1.1 Metacello new baseline: 'Fog'; repository: 'filetree:///path/to/git-repository/Fog/src';; load. It's based on the standar API for javascript ( https://github.com/ethereum/wiki/wiki/JavaScript-API) . It provides interaction with remote contracts, it do as well provides a way for navigating the architecture objects: blocks, transactions, accounts and contracts. I hope you find it useful. Feel free to fill the github issue tracker with anything you find :). I will come to you with some new exciting news about ethereum soon :) Santiago.

5 9

How to get rid of empty XML nodes?
by Stephane Ducasse Jan. 29, 2018

Jan. 29, 2018

)Hi we are manipulating an XML document and I would like to get rid of the spurious empty string. We saw that the gt panes are doing it. (aNodeWithElements isStringNode and: [aNodeWithElements isEmpty or: [aNodeWithElements isWhitespace]] Is there a way not to produce empty nodes? Is there a simple way not to have to handle them Now each time we are dealing with a node with have to check. Stef

3 10

Moving from mc to tonel?
by Martin Dias Jan. 25, 2018

Jan. 25, 2018

Hi all, I'm considering to move Epicea (+Ombu+Hiedra) to github. I read that tonel is the new format and there is Peter's git-migration, which could be the tool to do it. Is it good idea to move now? Is it the right tool? Bests, MartÃn

9 20

Re: [Pharo-dev] [Pharo-users] ZnClient in Pharo 6.1 not working for Https on Windows
by Cyril Ferlicot D. Jan. 21, 2018

Jan. 21, 2018

Le 06/10/2017 Ã 13:02, kmo a Ã©critÂ : > I was trying to Soup on Windows 7 and found I could not access https sites > > On windows 7 and 10 the following code fails: > > ZnEasy get:'https://genius.com/Alice-nuvole-lyrics'. > > This works fine in Pharo 5 on Windows. Also works fine with Pharo 6.1 (32 > bit) on Linux. > > The problem is Pharo 6.1 on Windows. > > The error is: > > "SSL/TLS plugin initialization failed (VM plugin missing ? OS libraries > missing ?)" > > I know that his has been seen on Linux in the past - see > http://forum.world.st/SSL-TLS-plugin-initailization-failed-VM-plugin-missin… > > I'm running the latest vm and 6.1 image. > > > > -- > Sent from: http://forum.world.st/Pharo-Smalltalk-Users-f1310670.html > I just saw that Windows Defender consider this plugin as a Trojan and delete it. Name of the menace: Trojan:Win32/Azden.A!cl Should this be reported on Pharo's issue tracker or on opensmalltalk one? -- Cyril Ferlicot https://ferlicot.fr http://www.synectique.eu 2 rue Jacques PrÃ©vert 01, 59650 Villeneuve d'ascq France

5 5

[IMPORTANT] PHARO CONTRIBUTION PROCESS UPDATED! PLEASE READ!
by Esteban Lorenzano Jan. 19, 2018

Jan. 19, 2018

Hi! Iâm working on simplifying the contribution process, after collecting opinions/experiences last couple of months. As you know, Pharo contribution process is still WIP and we aim to have it as smooth as possible for Pharo 7.0 release. Now, after observe the idea of the âsystem repositoriesâ was a bad idea because it introduced extra and non standard âpathâ to contribution, I managed to remove that to reestablish âthe regular wayâ: you will now need to add pharo repository just as any other repository you add, by cloning or adding local repository. I took Guilleâs doc and moved it to pharo project (it does not has sense to have it living in a contributorâs repository when is so important). You can find it here: https://github.com/pharo-project/pharo/wiki/Contribute-a-fix-to-Pharo <https://github.com/pharo-project/pharo/wiki/Contribute-a-fix-to-Pharo> This document is also updated to reveal this new process, please read it. How to update your startup scripts? Some people has added startup scripts to easy the first part of contribution. Instead enabling system repositories, etc. you now need to replace that with this: (IceRepositoryCreator new location: '/path/to/pharo-project/pharo' asFileReference; subdirectory: 'src'; createRepository) register PLEASE, PLEASE, PLEASEâ¦ take a moment to read and try the document. Is very important that document reflects new process and works reliable in different scenarios (I validated it on macOS and Windows, and assumed it worked fine on linux but you knowâ¦ bad assumptions is the base of failure ;) ) Iâm eager to hear your feedback and continue enhancing the process. (yes, Stef, I know UI is still cumbersomeâ¦ Iâm working on that :) ) cheers! Esteban

13 57

New Year Wishlist (2018) ?
by Esteban Lorenzano Jan. 6, 2018

Jan. 6, 2018

Hi everybody, This looks like a good moment of the year to ask all of you what would you want to see in Pharo next year. Features, improvements, radical changes, etcâ¦. whatever you want. Of course, this list will not be a roadmap and it does not means we will implement all of it (as always, time drives our possibilities), but is a good moment for us a a community to check where we are and where we wan to go next :) So, letâs those wishes come! cheers, Esteban

19 50

gtInspector+FastTable enhancement requests
by Ben Coman Jan. 5, 2018

Jan. 5, 2018

Feedback for GT'ers... I'm probably ab-using FastTable support in gtInspector, but as a quick hack during initial prototyping it was *very* nice to be able review retrieved data as a table of instances with a column per instance variable. Some enhancements that would make this even better... 1. Horizontal scroll bar - There are columns past the right margin that I can't get to for lack of one. 2. Column resizing to shift subsequent columns, which could provide another path to seeing the hidden columns. 3. Individual column width auto-fit - like in Excel where you double-click the edge https://www.youtube.com/watch?v=Yeb8emsjtf4 4. All columns auto-fit 5. Auto-fit columns prior to first opening. 6. Being able to split column labels over two lines - to allow for thinner columns For a demonstration, could you please load the attached ST fileout (its small), and also... Gofer it smalltalkhubUser: 'SvenVanCaekenberghe' project: 'Neo'; configurationOf: 'NeoJSON'; loadStable. Then inspect.... BittrexMarket getAll. cheers -ben

3 4