Tachyon-list Mars 2011

tachyon-list@iro.umontreal.ca

4 participants
45 discussions

A Proposal...
by chevalma＠iro.umontreal.ca 01 Avr '11

01 Avr '11

I just had a discussion with Erick, and he told me that he would rather implement the GC in C, as a first version. Since he's the one writing it, and he seems interested in the idea of having a JS GC eventually, I will respect his preference and stop insisting on this issue. I would, however, like to propose something else. I would like to volunteer to write the image writer/loader. I would also like to implement the writer in JavaScript. I think that this could help me expose tricky issues that will potentially come up while the GC is being written. I also think that it could be beneficial if I do this around the same time Erick is writing the GC, as we would be able to discuss technical issues that affect both. I would also need the same back-end support the GC will need. I'm enthusiastic about doing this, I know the system well, and I think I've shown I can be pretty productive. I'm confident it could realistically be done fairly quickly, no slower than the GC will take to be completed. I will also argue that doing this in JS is less bug-prone than the GC, as it will not need to destroy the current heap, we can trigger serialization whenever we want (as opposed to the GC) and the output file can be examined easily using external tools. What do you think? - Maxime

3 4

Object Model
by chevalma＠iro.umontreal.ca 30 Mar '11

30 Mar '11

I'm subscribed to the PyPy blog and it seems they released a series of articles about their tracing JIT recently. One of the parts talks about maps and object models, and the kinds of things you can do to optimize access to those objects: http://morepypy.blogspot.com/2011/03/controlling-tracing-of-interpreter-wit… - Maxime

1 0

TraceMonkey
by Marc Feeley 30 Mar '11

30 Mar '11

Here is an interesting overview of TraceMonkey by David Mandelin which we met at Mozilla: http://hacks.mozilla.org/2009/07/tracemonkey-overview/ It is funny to see that he uses a very similar example to the one we presented (sum of numbers). Marc

2 1

Fwd: [Ref#2011032610000013] Swap sur baro
by Marc Feeley 27 Mar '11

27 Mar '11

Begin forwarded message: > From: Gestion des requêtes <radar(a)iro.umontreal.ca> > Date: March 27, 2011 7:57:00 AM PDT > To: Marc Feeley <feeley(a)iro.umontreal.ca> > Subject: Re: [Ref#2011032610000013] Swap sur baro > > Bonjour > > Marc Feeley <feeley(a)iro.umontreal.ca> écrit: > >> J'aimerais augmenter le swap de baro à 200 GB >> (pour rouler des programmes avec >> de très gros espaces mémoire virtuel). > > J'ai repartitionne les trois disques de 2 T > pour y ajouter un partition SWAP de 100g. > J'ai ensuite reconstruit la parttion de data > sur chacun d'eux. > > Il y a maintenant sur baro, 3 partitions swap > de 100g reparties sur 3 disques de donnees > pour un total de pres de 300g (pertes pour > espace de gestion et confusion 1024/1000). > -- > Claude Goutier > Soutien technique DIRO >

1 0

Tachyon compiling Tachyon
by Marc Feeley 26 Mar '11

26 Mar '11

On baro I've managed to get statistics on the bootstrapped Tachyon. I measured 2 versions of Tachyon: TV -- Tachyon compiled by V8 (64 bit version) TTV -- Tachyon compiled by TV (also in 64 bit version) The execution time and space used for compiling the Tachyon source code by these compilers were as follows. TV compiling the Tachyon source code: Compiling Tachyon: 403.425 s, 21104.470703125 MB allocated time=82% alloc=71% -- Machine code generation time=43% alloc=9% -- Linear scan time=17% alloc=3% -- Resolution time=12% alloc=36% -- IR to ASM translation time=2% alloc=7% -- Computing live intervals time=1% alloc=2% -- Operand assignment time=1% alloc=3% -- Assembly time=0% alloc=2% -- Number instructions time=0% alloc=1% -- Computing fixed intervals time=0% alloc=0% -- Order blocks time=14% alloc=20% -- IR lowering time=7% alloc=10% -- comm elim time=2% alloc=2% -- opt patterns pass 1 time=1% alloc=1% -- const prop time=1% alloc=1% -- opt patterns pass 2 time=1% alloc=4% -- inlining/transform time=0% alloc=0% -- r/w analysis time=1% alloc=5% -- IR generation time=0% alloc=1% -- Machine code linking time=0% alloc=0% -- Parsing TTV Tachyon compiling the Tachyon source code: Compiling Tachyon: 4737 s, 47687 MB allocated time=57% alloc=41% -- IR lowering time=46% alloc=23% -- comm elim time=3% alloc=5% -- opt patterns pass 1 time=2% alloc=2% -- const prop time=1% alloc=2% -- opt patterns pass 2 time=1% alloc=6% -- inlining/transform time=0% alloc=0% -- r/w analysis time=37% alloc=48% -- Machine code generation time=25% alloc=26% -- IR to ASM translation time=5% alloc=4% -- Linear scan time=4% alloc=2% -- Resolution time=0% alloc=1% -- Computing live intervals time=0% alloc=2% -- Assembly time=0% alloc=0% -- Operand assignment time=0% alloc=1% -- Computing fixed intervals time=0% alloc=0% -- Number instructions time=0% alloc=0% -- Order blocks time=3% alloc=7% -- IR generation time=0% alloc=1% -- Machine code linking time=0% alloc=1% -- Parsing So on this "benchmark", Tachyon compiles the source code to x86_64 code which runs 11.7 times slower than the one produced by V8 and which takes 2.6 times the space that the V8 generated code takes. Moreover, it takes 79 minutes to compile Tachyon with Tachyon. It is hard to say how much time V8 takes to compile Tachyon, given that probably not all parts of Tachyon actually get compiled to machine code. But my guess is that for the parts that do get compiled to machine code the time is probably a few seconds. Does someone know how to measure the V8 compilation time? Marc

1 0

ES5 compatibility table
by Marc Feeley 24 Mar '11

24 Mar '11

An interesting table of supported ES5 features. http://kangax.github.com/es5-compat-table/ Marc

3 3

Bignum Divisors
by chevalma＠iro.umontreal.ca 24 Mar '11

24 Mar '11

Hello, I tested the use of 12 padding bits for 64bit boxed integers. Everything seemed to work fine, for the most part, except that a few unit tests failed with the exception: we don't yet support multi-digit divisors I'm guessing some of the values produced during constant propagation are outside of what is representable with a single big digit. Marc, would you have time to implement support for multiple big digit divisors? - Maxime

2 6

Shell working under boostrap
by chevalma＠iro.umontreal.ca 23 Mar '11

23 Mar '11

I tried to run the shell boostrapped and ran into a "callee is not a function" error... So I had the idea of adding source code locations to those error messages (why didn't I think of that before!). I found out we were missing the String.prototype.substr library function (not to be confused with its cousin String.prototype.substring). The said function is now implemented, and the shell works under the bootstrap :) - Maxime

2 2

Twisted lambda
by Marc Feeley 23 Mar '11

23 Mar '11

Let's call this icon the "twisted lambda"... quite fitting for JavaScript. Here in .png format of various resolutions. Marc

2 1

64 bit bootstrap and bt-bt-parser succeed ;-)
by Erick Lavoie 23 Mar '11

23 Mar '11

Just to let you know guys that it takes around 32 minutes to bootstrap tachyon and have the bootstrapped tachyon compiler compile the parser, then run it on fib and experience an immense success and a pleasurable after life. Erick

2 1

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

Tachyon-list Mars 2011