8 ms·

Brief History of JIT in MoarVM for Perl 6

83 points by Ultimatt 8y ago

3rdAccount 8y ago
I'm always impressed with the size of the problem that the Perl6 core team is trying to solve relative to the number of volunteers and funding. It is a really nice language based off of all the tutorials and presentations I've watched. It just needs some better performance before I'll really start using it and I think there are many in the community that are in the same boat. It's a pity, because it's a chicken or the egg kind of problem. I would try to help, but when it comes to computer science I am a consumer and not a producer (I have my own large projects in my domain :)).
- tyil 8y ago
  > It is a really nice language based off of all the tutorials and presentations I've watched. It just needs some better performance before I'll really start using it I'm also very impressed in how the language is turning out. It's a pleasure to write and read. I know most people really don't like Unicode operators, but I find them to be very useful. They're short and concise, and allow those writing mathematical equations to stick to their regular symbols. And for those that don't know how to do Unicode on their OS/editor, there's ASCII equivalents. Performance-wise, it seems to be fast enough for all my personal projects, and it's getting faster every release. Is there a certain piece of code that you want to have running within a certain amount of time? That would allow you (and Perl 6 devs) to have a goal to look at, and work towards.
  - 3rdAccount 8y ago
    Running grammars past very large datasets. You hear enough people talk about the performance and you decide to stay away a bit until things improve. I also have some numeric code and that is generally probably not a great P6 fit (more of a Python + Numpy thing). As a general scripting language, P6 looks nice as a lot of features such as concurrency that are aren't super straightforward (Ex: python) seem to be dead simple in P6.
    - b2gills 8y ago
      Perl 6 was designed in part to allow PDL (Perl Data Language) type features to be added easily. In fact some things that you would do in PDL have already been added, they just currently aren't as fast as they could be. (1,2,3; 4,5,6; 7,8,9) »**» 2 (1,4,9; 16,25,36; 49,64,81) The above is specifically allowed to do all of the operations in parallel, but doesn't currently. (Perhaps even on a GPU)
totalperspectiv 8y ago
I appreciate the time taken to fill in the history!
aidenn0 8y ago
Why did I think the Perl 6 VM was called parrot?
- ssutch3 8y ago
  Many years ago, it was. Parrot is a separate team, and was designed to be the "mother of all VMs" with Perl 6 as its sort of halo language implementation. It split off as the Perl 6 project matured and needed to own the vm/compiler.
  - chrisseaton 8y ago
    What was it that Parrot couldn't do that Perl 6 needed? We often have efforts to make one VM that can support everything and they have usually failed in the past. I wonder why it failed this time?
    - lizmat 8y ago
      The history in short from my perspective: * Parrot was intended as the runtime for Perl 6. * Perl 6 was not ready to indicate what it needed when development started on Parrot * Scope/Goal of Parrot was extended to other scripting languages * Parrot started to become more and more unsuitable as the VM for Perl 6 once Perl 6 figured out what it needed * Parrot failed to attract other clients for its VM * The Great List Refactor in Perl 6 the year before its official release, forced a decision on whether to also refactor for the Parrot backend or not (on top of the refactoring needed for the JVM and MoarVM backend) * The decision was no: but really, the GLR was the straw that broke the camel's back, so to say. Particularly, NFG support and async/event driven support were deemed to be impossible to implement with the state Parrot was in at the time. * With no more clients for its VM, developers left the Parrot project
      - acqq 8y ago
        So is “nomen” now “omen?” Is the Parrot dead? (1) (2) I can’t figure that out. On http://parrot.org http://parrot.org It seems it maybe still moves? 1) https://en.wikipedia.org/wiki/Dead_Parrot_sketch https://en.wikipedia.org/wiki/Dead_Parrot_sketch 2) The name's origin is really the sketch see: https://web.archive.org/web/20040203055250/http://www.oreilly.com/news/parrotstory_0401.html https://web.archive.org/web/20040203055250/http://www.oreill... and https://en.wikipedia.org/wiki/Parrot_virtual_machine#cite_ref-4 https://en.wikipedia.org/wiki/Parrot_virtual_machine#cite_re... "The name Parrot came from an April Fool's joke which announced a hypothetical language, named Parrot, that would unify Python and Perl.[4][5]"
sunseb 8y ago
Is Perl 6 fast now?
- lizmat 8y ago
  Since H. Merijn Brandt started working on the Perl 6 version of Text::CSV, he has kept a log of how long a specific test (referred to as test-t) too to run. https://tux.nl/Talks/CSV6/speed.log https://tux.nl/Talks/CSV6/speed.log Since then, Rakudo Perl 6 has become more than 100x as fast, at least for this test. And more than 250x as fast using a concurrent version of the test (which was as simple as putting a `.race` in the code). This makes it still slower than the Perl 5 version, which uses XS (all parsing logic written in C). But the gap is closing with another round of optimizations to be merged later this week after an extensive refactor that took about 2 months. Of which you will most certainly see a report on https://6guts.wordpress.com https://6guts.wordpress.com , the blog of Jonathan Worthington.
- b2gills 8y ago
  MoarVM has recently gotten the ability to write optimization plugins in higher level code. This should make it easier to create certain optimizations as they will have access to the syntax tree. So it may be that the rate at which it gets faster has gotten faster.
- raiph 8y ago
  Quoting the blog post this HN article is supposed to be about[1]: PS: You might read this and be reasonably surprised that Rakudo Perl 6 is not, after all this, very fast yet. I have a - not entirely serious - explanation for that: All problems in computer science can be solved with a layer of indirection. Many layers of indirection make programs slow. Perl 6 solves many computer science problems for you ;-) In the future, we'll continue to solve those problems, just faster. [1] http://brrt-to-the-future.blogspot.com/2018/07/perl-6-on-moarvm-has-had-jit-for-few.html http://brrt-to-the-future.blogspot.com/2018/07/perl-6-on-moa...
- jdoege 8y ago
  It really depends. The question really should be, "Is Perl 6 fast enough?" For my purposes, the answer is almost always yes. I can imagine some applications where the answer would be no, but then I would probably be looking at C/C++ for a solution, anyway. Language bench-racing has really begun to look counter-productive to me.

83 points by Ultimatt 8y ago

I'm always impressed with the size of the problem that the Perl6 core team is trying to solve relative to the number of volunteers and funding. It is a really nice language based off of all the tutorials and presentations I've watched. It just needs some better performance before I'll really start using it and I think there are many in the community that are in the same boat. It's a pity, because it's a chicken or the egg kind of problem. I would try to help, but when it comes to computer science I am a consumer and not a producer (I have my own large projects in my domain :)).

tyil 8y ago
> It is a really nice language based off of all the tutorials and presentations I've watched. It just needs some better performance before I'll really start using it I'm also very impressed in how the language is turning out. It's a pleasure to write and read. I know most people really don't like Unicode operators, but I find them to be very useful. They're short and concise, and allow those writing mathematical equations to stick to their regular symbols. And for those that don't know how to do Unicode on their OS/editor, there's ASCII equivalents. Performance-wise, it seems to be fast enough for all my personal projects, and it's getting faster every release. Is there a certain piece of code that you want to have running within a certain amount of time? That would allow you (and Perl 6 devs) to have a goal to look at, and work towards.
- 3rdAccount 8y ago
  Running grammars past very large datasets. You hear enough people talk about the performance and you decide to stay away a bit until things improve. I also have some numeric code and that is generally probably not a great P6 fit (more of a Python + Numpy thing). As a general scripting language, P6 looks nice as a lot of features such as concurrency that are aren't super straightforward (Ex: python) seem to be dead simple in P6.
  - b2gills 8y ago
    Perl 6 was designed in part to allow PDL (Perl Data Language) type features to be added easily. In fact some things that you would do in PDL have already been added, they just currently aren't as fast as they could be. (1,2,3; 4,5,6; 7,8,9) »**» 2 (1,4,9; 16,25,36; 49,64,81) The above is specifically allowed to do all of the operations in parallel, but doesn't currently. (Perhaps even on a GPU)

totalperspectiv 8y ago

I appreciate the time taken to fill in the history!

aidenn0 8y ago

Why did I think the Perl 6 VM was called parrot?

ssutch3 8y ago
Many years ago, it was. Parrot is a separate team, and was designed to be the "mother of all VMs" with Perl 6 as its sort of halo language implementation. It split off as the Perl 6 project matured and needed to own the vm/compiler.
- chrisseaton 8y ago
  What was it that Parrot couldn't do that Perl 6 needed? We often have efforts to make one VM that can support everything and they have usually failed in the past. I wonder why it failed this time?
  - lizmat 8y ago
    The history in short from my perspective: * Parrot was intended as the runtime for Perl 6. * Perl 6 was not ready to indicate what it needed when development started on Parrot * Scope/Goal of Parrot was extended to other scripting languages * Parrot started to become more and more unsuitable as the VM for Perl 6 once Perl 6 figured out what it needed * Parrot failed to attract other clients for its VM * The Great List Refactor in Perl 6 the year before its official release, forced a decision on whether to also refactor for the Parrot backend or not (on top of the refactoring needed for the JVM and MoarVM backend) * The decision was no: but really, the GLR was the straw that broke the camel's back, so to say. Particularly, NFG support and async/event driven support were deemed to be impossible to implement with the state Parrot was in at the time. * With no more clients for its VM, developers left the Parrot project
    - acqq 8y ago
      So is “nomen” now “omen?” Is the Parrot dead? (1) (2) I can’t figure that out. On http://parrot.org http://parrot.org It seems it maybe still moves? 1) https://en.wikipedia.org/wiki/Dead_Parrot_sketch https://en.wikipedia.org/wiki/Dead_Parrot_sketch 2) The name's origin is really the sketch see: https://web.archive.org/web/20040203055250/http://www.oreilly.com/news/parrotstory_0401.html https://web.archive.org/web/20040203055250/http://www.oreill... and https://en.wikipedia.org/wiki/Parrot_virtual_machine#cite_ref-4 https://en.wikipedia.org/wiki/Parrot_virtual_machine#cite_re... "The name Parrot came from an April Fool's joke which announced a hypothetical language, named Parrot, that would unify Python and Perl.[4][5]"

sunseb 8y ago

Is Perl 6 fast now?

lizmat 8y ago
Since H. Merijn Brandt started working on the Perl 6 version of Text::CSV, he has kept a log of how long a specific test (referred to as test-t) too to run. https://tux.nl/Talks/CSV6/speed.log https://tux.nl/Talks/CSV6/speed.log Since then, Rakudo Perl 6 has become more than 100x as fast, at least for this test. And more than 250x as fast using a concurrent version of the test (which was as simple as putting a `.race` in the code). This makes it still slower than the Perl 5 version, which uses XS (all parsing logic written in C). But the gap is closing with another round of optimizations to be merged later this week after an extensive refactor that took about 2 months. Of which you will most certainly see a report on https://6guts.wordpress.com https://6guts.wordpress.com , the blog of Jonathan Worthington.
b2gills 8y ago
MoarVM has recently gotten the ability to write optimization plugins in higher level code. This should make it easier to create certain optimizations as they will have access to the syntax tree. So it may be that the rate at which it gets faster has gotten faster.
raiph 8y ago
Quoting the blog post this HN article is supposed to be about[1]: PS: You might read this and be reasonably surprised that Rakudo Perl 6 is not, after all this, very fast yet. I have a - not entirely serious - explanation for that: All problems in computer science can be solved with a layer of indirection. Many layers of indirection make programs slow. Perl 6 solves many computer science problems for you ;-) In the future, we'll continue to solve those problems, just faster. [1] http://brrt-to-the-future.blogspot.com/2018/07/perl-6-on-moarvm-has-had-jit-for-few.html http://brrt-to-the-future.blogspot.com/2018/07/perl-6-on-moa...
jdoege 8y ago
It really depends. The question really should be, "Is Perl 6 fast enough?" For my purposes, the answer is almost always yes. I can imagine some applications where the answer would be no, but then I would probably be looking at C/C++ for a solution, anyway. Language bench-racing has really begun to look counter-productive to me.