Math.NET Project News - All Comments

Tune Up Your PC » Post Topic » dnAnalytics + Iridium –> Math.NET Numerics

Tune Up Your PC » Post Topic » dnAnalytics + Iridium –> Math.NET Numerics — Sat, 24 Oct 2009 17:20:24 GMT

Pingback from Tune Up Your PC » Post Topic » dnAnalytics + Iridium –> Math.NET Numerics

re: dnAnalytics + Iridium = Math.NET Numerics

Alexey Zakharov — Fri, 23 Oct 2009 18:15:26 GMT

Good news! C# really needs such library in stable version.

re: dnAnalytics + Iridium = Math.NET Numerics

Harry — Wed, 26 Aug 2009 14:41:00 GMT

Note the array[rowBeginIndex + colIndex] will perhaps be slower since this must two addition each time an element is accessed. The JIT might compile this away though. The unsafe version eliminates this as the current element pointer can be used and incremented within the loop i..e

fixed (double* arrayPtr = array[0,0])

{

double* rowEndPtr = arrayPtr + width * height (OR stride);

for (double* rowPtr = arrayPtr; rowPtr != rowEndPtr; rowPtr += stride)

{

double* colEndPtr = rowPtr + width;

for (double* colPtr = rowPtr; colPtr != colEndPtr; ++colPtr)

{

double value = *colPtr;

}

This is similar to how STL in C++ works.

re: dnAnalytics + Iridium = Math.NET Numerics

Harry — Wed, 26 Aug 2009 08:55:17 GMT

With regard to double[,] being slower than double[][] I find that very hard to believe then this is due to bad indexing into the memory.

That is, when you are using double[][] you are using the trick that double[0] gives you the entire row which you can then index into. But if you constantly index array[row, col] in a double for loop then this will of course be a lot slower.

The same is, however, possible with double[,]/double[] and will be better due to cache since this is a continous memory block, by using unsafe code or strided as you mention. I.e. for double[,]

fixed (double* rowPtr = array[row,0])

OR double[]

int rowBeginIndex = row * stride;

array[rowBeginIndex + colIndex]

The main point is that continuous memory will be faster; to allocate and to process due to cache. Additionally, it will allow for direct interop with IPP/MKL.

re: dnAnalytics + Iridium = Math.NET Numerics

Christoph Rüegg — Tue, 11 Aug 2009 08:49:11 GMT

It's very unlikely that we'll move back to double[,] since it's actually the worst choice in terms of performance (Iridium got a speedup of almost 2 when we switched from double[,] to double[][]).

However, we do consider double[] with stride instead (or in addition to) double[][].

re: dnAnalytics + Iridium = Math.NET Numerics

Harry — Mon, 10 Aug 2009 07:51:58 GMT

Will that mean that you finally stop having matrices as multiple arrays i.e. you will use float[,] instead of float[][]?

Latter is slow and bad for native interop.

re: dnAnalytics + Iridium = Math.NET Numerics

Joannes Vermorel — Mon, 03 Aug 2009 19:22:18 GMT

Congratulations! Sparse linear algebra is really a nice move (I am sorry I had not been able to push it forward at the time).

re: Complete Iridium Feature List

Christian — Wed, 05 Nov 2008 13:07:20 GMT

Hello Christoph,

it's great that you will be spending more time on Math.NET. I think your library has great potential!

Keep up the good work!

Christian

re: Complete Iridium Feature List

Joannes Vermorel — Wed, 05 Nov 2008 07:18:20 GMT

Hi Christoph,

Good to know that you're back on Math.NET because we are using it more than ever. We just started to use FFT for time-series analysis.

Best regards,

Joannes

re: Iridium 2008 August Release (2008.8.16.470)

Christoph Rüegg — Thu, 14 Aug 2008 20:40:28 GMT

Thanks. Yes, indeed, the SourceForge mirror infrastructure is unbeatable :)

re: Iridium 2008 August Release (2008.8.16.470)

Joannes Vermorel — Thu, 14 Aug 2008 13:19:00 GMT

Thanks for this good work. I have also noticed that you've posted the release on Sourceforge.Net. This is really a nice thing to do (so far, I have always found Sourceforge to be a top notch backup provider :-)

re: Iridium 2008 March Release (v2008.3.12.405)

Christoph Rüegg — Sun, 06 Apr 2008 11:49:42 GMT

Thanks, I'm looking forward to it! :)

re: Iridium 2008 March Release (v2008.3.12.405)

Joannes Vermorel — Mon, 31 Mar 2008 12:15:14 GMT

Nice to see that new people are joining Math.NET. If I get the time, I will try to submit some Erlang-related formulas and distribution for the next release :-)

re: Math.NET Tightened Security - working fine for you?

Christoph Rüegg — Mon, 31 Mar 2008 09:38:48 GMT

The certificate validation causes network access to download the certificate revocation lists from the certificate authority webserver. Because of this, we have decided to drop signing with certificates for now.

Math.NET Security Aspects - working fine for you?

Math.NET Project News — Sat, 09 Feb 2008 23:04:26 GMT

Since the February 2008 Release , Math.NET Iridium and Neodym both adapted a new security model: Strong