F# News

interesting..I was using clojure and now I'm l...

2012-05-14T14:28:20.000Z

interesting..I was using clojure and now I'm learning f#, It's very common listen about the clojure-jvm power for concurrence and parallel programming, but now than I'm using the clr and mono is really impressing how prehistoric feels the jvm, actually when .Net was moving fast creatng a better environment for parallel and concurrent programming, the jvm discuss if lambdas was a good idea for the java language....

@Art: Good question. I don't know what the per...

2012-05-14T14:11:11.983Z

@Art: Good question. I don't know what the performance difference between QR and Jacobi is but this article covered QR and the article Visualizing linear algebra: Singular Value Decomposition used Jacobi. I'll benchmark them when I get the chance.

@Anonymous "Perhaps I'm stupid but doesn&...

2012-05-14T10:53:32.579Z

@Anonymous "Perhaps I'm stupid but doesn't lazy evaluation mean your test is worthless?". There is no lazy evaluation anywhere in either solution.

2012-05-07T08:17:10.191Z

This comment has been removed by the author.

@coco: "functional code is much more readable...

2012-05-06T23:08:34.984Z

@coco: "functional code is much more readable and expressive". Not always. Graph algorithms are one place where functional programming falls down, not least because there are no purely functional weak dictionaries. The implementations of many graph algorithms in languages like Haskell are several times longer than their C equivalents.

"Is really a lack than f# inmutable estructure be so slow?, can be it improve??...o the real solution for multicore and hight performance is a language with functional tools but where you program in an wmpiric way??" Improvements to the .NET garbage collector could dramatically improve the performance of purely functional data structures. Some significant advances were made in .NET 4. But it is not possible for purely functional data structures to be competitively performant in the general case. They come with too much baggage.

Interesting, I've read your post about "F...

2012-05-06T14:19:22.852Z

Interesting, I've read your post about "F# Map vs .NET Dictionary performance" and I was searching any benchmark between clojure maps and native java...I guess clojure is slower but I'm not sure if is so dramatic...clojure is much more functional than f# (although they are similar, the clojure community only use mutable data when it is really necesary) and the clojure performance is not so bad (you must considered it's a dynamic language too...

For me (I come from clojure and CLips) functional code is much more readable and expressive, f# code seems olny a better c# where the programmers only encapsulate mutable code between functions, and problem solve!!..

Is really a lack than f# inmutable estructure be so slow?, can be it improve??...o the real solution for multicore and hight performance is a language with functional tools but where you program in an wmpiric way??...

@David: I think Rx will be squeezed out between di...

2012-05-06T10:35:45.889Z

@David: I think Rx will be squeezed out between different axes, of which performance is one. Rx might be faster now for what it provides but its functionality is limited and threads+locks are probably substantially faster still. In the other direction, async is slightly slower but much easier to reason about and debug. So I'm not sure Rx has anywhere to thrive.

Also, something like the Disruptor might give you the best of both worlds for certain (very common!) applications.

Would you have any opinions to share about the lat...

2012-05-04T15:30:39.675Z

Would you have any opinions to share about the latest version of Rx which seems to have enjoyed more aggressive optimization.

I seriously doubt now that F# async would have comparable performance profile anymore. Benchmarks I ran a few months ago showed Rx to be faster than both CCR and F# async yet with worst scaling.

Just a quick comment. I have learnt F# without kno...

2012-04-08T06:42:22.544Z

Just a quick comment. I have learnt F# without knowing C#. Now I learn C# a little bit to use it just in case :)

QH - "... this section is devoted to the comp...

2012-03-24T00:09:05.498Z

QH - "... this section is devoted to the computational aspect ... use the WPF vector type ... to simplfy our solution." GR8

Vector Graphics Editor -
"Although WPF provides a 2D Vector type we shall define our own that supports structural equality and comparison: ..." GR8

How to combine? Override WPF Vector? (Use XNA/DXn?) ... F# for Numerics ... other math.libs?

A very interesting read. I will be applying some o...

2012-03-20T19:53:35.952Z

A very interesting read. I will be applying some of the ideas before long.

Nigel Wilson

Perhaps I'm stupid but doesn't lazy evalua...

2012-03-08T08:29:07.552Z

Perhaps I'm stupid but doesn't lazy evaluation mean your test is worthless ?

Are you sure F# actually does the work.

Parallel programming? I think you meant concurrent...

2012-03-07T10:19:42.632Z

Parallel programming? I think you meant concurrent programming, using multithreading does not make program
parallel my Dr. Cliff , because every parallel program is concurrent but every concurrent is not parallel, the thing is OS is scheduling processes on cores not threads so in your test it does not matter if you use 8 cores processes one thread at a time will be executing at a core while the other threads will be waiting although you have 7 core free to use. 2 threads of one process CAN'T
executing on two cores at the same time. only a few OS do that and i think you didn't use one that schedule threads. so if the result are not reliable.

As much as I think that @Jan is a senseless Java f...

2012-03-07T06:37:59.653Z

As much as I think that @Jan is a senseless Java fanboy, for LOC count you should really compare Java to C#. The speed difference would probably be even wider as the C# compiler is optimised better than the F# one.

Congratulations!

2012-03-04T15:31:50.468Z

Congratulations!

And by coincidence, some hours ago I was working o...

2012-02-28T08:54:13.484Z

And by coincidence, some hours ago I was working on a Reactive Extensions version.
:):)
Excerpt at https://gist.github.com/1931336

Jon, I think you should run your popular filter ag...

2012-01-12T22:13:18.380Z

Jon, I think you should run your popular filter again and remove Anon ...

Jon, QH uses "update" in several places....

2012-01-10T17:41:41.288Z

Jon, QH uses "update" in several places. How do the uses (signatures) differ?

> let move canvas update (e: Input.MouseEventArgs) = ...
update i;;

let update (polygon: Shapes.Polygon) (ellipses: _ []) i =
...

[]
do
...
let update = update polygon ellipses
...
canvas.MouseMove.Add(move canvas update)

Thanks

@Art: Yes, ResizeArray (an alias for the .NET List...

2012-01-09T10:17:08.707Z

@Art: Yes, ResizeArray (an alias for the .NET List collection) is surprisingly inefficient as an extensible array. Filling a raw array is much faster.

I take it Resize.Array slows things down?

2012-01-09T02:03:40.097Z

I take it Resize.Array slows things down?

Looking for insight into ResizeArray. Thanks.

2012-01-05T22:49:50.910Z

Looking for insight into ResizeArray. Thanks.

"...Key aspect of the GUI ... This design fol...

2011-12-12T23:41:51.797Z

"...Key aspect of the GUI ... This design follows the mantra
'make illegal states unrepresentable'.
... Using data structures such as the option type ... revisit ... interactive manipulation"

Crystalline!, as always.

Hi Jon, any chance of a future article on conjugat...

2011-12-04T00:03:39.242Z

Hi Jon, any chance of a future article on conjugate-gradient?
From IOD's RPP Discrete Comput Geom (2009) 41: 444–460
Sorry for my poor html math notation.

"JJTl = JD +αe, (7)
KKTl = KD +αf (9)
(7) and (9) can be solved efficiently using the conjugate-gradient method"

Hi Jon. Whilst reading Golub and Van Loan Matrix C...

2011-11-21T23:45:12.741Z

Hi Jon.
Whilst reading Golub and Van Loan Matrix Computations, Thrid Edition, ... 8.4.6 Parallel Jacobi, pg. 431, "... Perhaps the most interesting distinction between the QR and the Jacobi approaches to the symmetric eigenvalue problem is the rich inherent parallelism of the latter. ... rotations within each ... "non-conflicting." ... a parallel ordering of the set ... .

Question -- what is the performance difference? And will many-core architectures capture the benefit? CPU? or GPU?

@neilsb: This exact server was done under contract...

2011-11-08T15:52:58.592Z

@neilsb: This exact server was done under contract for a client but the basic ideas have already been covered in several F#.NET Journal articles:
Concurrent web crawling using asynchronous workflows
An introduction to asynchronous workflows
Distributed message passing
Concurrent programming: TCP relay server
Generic server architecture
We also have a forthcoming article about servers with efficient compressed IO.