Category Archives: Usage

N-dim Array Broadcasting Efficiency in ILNumerics 4.10

Next to other great improvements in version 4.10 of ILNumerics Ultimate VS, it is especially one new feature which requires some attention: general broadcasting for n-dimensional ILArrays.

Broadcasting as a concept today is found in many popular mathematical prototyping systems. The most direct correspondence probably exists in the numpy package. Matlab and Octave offer similar functionality by means of the bsxfun function.

The term ‘broadcasting’ refers to a binary operator which is able to apply an elementwise operation to the elements of two n-dimensional arrays. The ‘operation’ is often as simple as a straight addition (ILMath.add, ILMath.divide, a.s.o.). However, what is special about broadcasting is that it allows the operation even for the case where both arrays involved do not have the same number of elements.

Broadcasting in ILNumerics prior Version 4.10

In ILNumerics, broadcasting is available for long already. But prior version 4.10 it was limited to scalars operating on n-dim arrays and vectors operating on matrices. Therefore, we had used the term ‘vector expansion’ instead of broadcasting. Obviously, broadcasting can be seen as a generalization of vector expansion.

Let’s visualize the concept by considering the following matrix A:

A

1  5   9  13  17
2  6  10  14  18
3  7  11  15  19
4  8  12  16  20

Matrix A might represent 5 data points of a 4 dimensional dataset as columns. One common requirement is to apply a certain operation to all datapoints in a similar way. In order to, let’s say, scale/weight the components of each dimension by a certain factor, one would multiply each datapoint with a vector of length 4.


ILArray<double> V = new[] { 0.5, 3.0, 0.5, 1.0 };

0.5
3.0
0.5
1.0

The traditional way of implementing this operation would be to expand the scaling vector by replicating it from a single column to a matrix matching the size of A.

VExp = repmat(V, 1, 5); 

0.5  0.5  0.5  0.5  0.5
3.0  3.0  3.0  3.0  3.0
0.5  0.5  0.5  0.5  0.5
1.0  1.0  1.0  1.0  1.0

Afterwards, the result can be operated with A elementwise in the common way.

ILArray<double> Result = VExp * A;

0.5   2.5   4.5   6.5   8.5
6.0  18.0  30.0  42.0  54.0
1.5   3.5   5.5   7.5   9.5
4.0   8.0  12.0  16.0  20.0

The problem with the above approach is that the vector data need to be expanded first. There is little advantage in doing so: a lot of new memory is being used up in order to store completely redundant data. We all know that memory is the biggest bottleneck today. We should prevent from lots of memory allocations whenever possible. This is where vector expansion comes into play. In ILNumerics, for long, one can prevent from the additional replication step and operate the vector on the matrix directly. Internally, the operation is implemented in a very efficient way, without replicating any data, without allocating new memory.

ILArray<double> Result = V * A;

0.5   2.5   4.5   6.5   8.5
6.0  18.0  30.0  42.0  54.0
1.5   3.5   5.5   7.5   9.5
4.0   8.0  12.0  16.0  20.0

Generalizing for n-Dimensions

Representing data as matrices is very popular in scientific computing. However, if the data are stored into arrays of other shapes, having more than two dimensions, one had to fall back to repmatting in order for the binary operation to succeed. This nuissance has been removed in version 4.10.

Now it is possible to apply broadcasting to two arrays of any matching shape – without the need for using repmat. In order for two arrays to ‘match‘ in the binary operation, the following rules must be fullfilled:

  1. All corresponding dimensions of both arrays must match.
  2. In order for two  corresponding dimensions to match,
    • both dimensions must be of the same length, or
    • one of the dimensions must be of length 1.

An example of two matching arrays would be a vector running along the 3rd dimension and a 3 dimensional array:

3D Cubes 4 Broadcasting2In the above image the vector (green) has the same length as the corresponding dimension of the 3D array (gray). The size of the vector is [1 x 1 x 6]. The size of the 3D array is [4 x 5 x 6]. Hence, any dimension of both, the vector and the 3D array ‘match’ in terms of broadcasting. A broadcasting operation for both, the vector and the array would give the same result as if the vector would be replicated along the 1st and the 2nd dimensions. The first element will serve all elements in the first 4 x 5 slice in the 1-2 plane. This slice is marked red in the next image: 3D Cubes 4 Broadcasting_sliceNote that all red elements here derive from the same value – from the first element of the green vector.  The same is true for all other vector elements: they fill corresponding slices on the 3D array along the 3rd dimension.

Slowly, a huge performance advantage of broadcasting becomes clear: the amount of memory saved explodes when more, longer dimensions are involved.

 Special Case: Broadcasting on Vectors

In the most general case and if broadcasting is blindly applied, the following special case potentially causes issues. Consider two vectors, one row vector and one column vector being provided as input parameters to a binary operation. In ILNumerics, every array carries at least two dimensions. A column vector of length 4 is considered an array of size [4 x 1]. A row vector of length 5 is considered an array of size [1 x 5]. In fact, any two vectors match according to the general  broadcasting rules.

As a consequence operating a row vector [1 x 5] with a column vector [4 x 1] results in a matrix [4 x 5]. The row vector is getting ‘replicated’ (again, without really executing the replication) four times along the 1st dimension, and the column vector 5 times along the rows.

array(new[] {1.0,2.0,3.0,4.0,5.0}, 1, 5) +array(new[] {1.0,2.0,3.0,4.0}, 4, 1)

<Double> [4,5]
[0]:          2          3          4          5          6
[1]:          3          4          5          6          7
[2]:          4          5          6          7          8
[3]:          5          6          7          8          9

Note, in order for the above code example to work, one needs to apply a certain switch:

Settings.BroadcastCompatibilityMode = false;

The reason is that in the traditional version of ILNumerics (just like in Matlab and Octave) the above code would simply not execute but throw an exception instead. Originally, binary operations on vectors would ignore the fact that vectors are matrices and only take the length of the vectors into account, operating on the corresponding elements if the length of both vectors do match. Now, in order to keep compatibility for existing applications, we kept the former behavior.

The new switch ‘Settings.BroadcastCompatibilityMode’ by default is set to ‘true’. This will cause the Computing Engine to throw an exception when two vectors of inequal length are provided to binary operators. Applying vectors of the same length (regardless of their orientation) will result in a vector of the same length.

If the ‘Settings.BroadcastCompatibilityMode’ switch is set to ‘false’ then general broadcasting is applied in all cases according to the above rules – even on vectors. For the earlier vector example this leads to the resulting matrix as shown above: operating a row on a column vector expands both vectors and gives a matrix of corresponding size.

Further reading: binary operators, online documentation

Installing ILNumerics – Unexpected behavior

At ILNumerics we get a lot of support requests every day. During the last couple of months some questions were related to installing ILNumerics. In some cases an unexpected error message appears. The easy solution is to manually uninstall our extension from all Visual Studio instances and then reinstall ILNumerics. This issue will be resolved once we release our new installer.

Continue reading Installing ILNumerics – Unexpected behavior

Directions to the ILNumerics Optimization Toolbox

As of yesterday the ILNumerics Optimization Toolbox is out and online! It’s been quite a challenge to bring everything together: some of the best algorithms, the convenience you as a user of ILNumerics expect and deserve, and the high performance requirements ILNumerics sets the scale on for. We believe that all these goals could be achieved quite greatly.

Continue reading Directions to the ILNumerics Optimization Toolbox

Getting to know your Scene Graph

Did you ever miss a certain feature in your ILNumerics scene graph? You probably did. But did you know, that most of the missing “features” mean nothing more than a missing “property”? Often enough, there is only a convenient access to a certain scene graph object needed in order to finalize a required configuration.

Recently, a user asked how to turn the background of a legend object in ILNumerics plots transparent. There doesn’t seem to be a straight forward way to that. One might expect code like the following to work:

var legend = new ILLegend("Line 1", "Line 2");
legend.Background.Color = Color.FromArgb(200, Color.White);

Continue reading Getting to know your Scene Graph

Fun with HDF5, ILNumerics and Excel

It is amazing how many complex business processes in major industries today are supported by a tool that shines by its simplicity: Microsoft Excel. ‘Recently’ (with Visual Studio 2010) Microsoft managed to polish the development tools for all Office applications significantly. The whole Office product line is now ready to serve as a convenient, flexible base framework for stunning custom business logic, custom computations and visualizations – with just a little help of tools like ILNumerics.

ExcelWorkbook1In this blog post I am going to show how easy it is to extend the common functionality of Excel. We will enable an Excel Workbook to load arbitrary HDF5 data files, inspect the content of such files and show the data as interactive 2D or 3D plots. Continue reading Fun with HDF5, ILNumerics and Excel

Performance on ILArray

Having a convenient data structure like ILArray<T> brings many advantages when handling numerical data in your algorithms. On the convenience side, there are flexible options for creating subarrays, altering existing data (i.e. lengthening or shortening individual dimensions on the run), keeping dimensionality information together with the data, and last but not least: being able to formulate an algorithm by concentrating on the math rather than on loops and the like.

Convenience and Speed

Another advantage is performance: by writing C = A + B, with A and B being large arrays, the inner implementation is able to choose the most efficient way of evaluating this expression. Here is, what ILNumerics internally does: Continue reading Performance on ILArray

Uncommon data conversion with ILArray

ILNumerics Computing Engine supports the most common numeric data types out of the box: double, float, complex, fcomplex, byte, short, int, long, ulong

If you need to convert from, let’s say ushort to float, you will not find any prepared conversion function in ILMath. Luckily, it is very easy to write your own:

Here comes a method which implements the conversion from ushort -> float. A straight forward version first:

        /// <summary>
        /// Convert ushort data to ILArray&lt;float>
        /// </summary>
        /// <param name="A">Input Array</param>
        /// <returns>Array of the same size as A, single precision float elements</returns>
        public static ILRetArray<float> UShort2Single(ILInArray<ushort> A) {
            using (ILScope.Enter(A)) {
                ILArray<float> ret = ILMath.zeros<float>(A.S);
                var retArr = ret.GetArrayForWrite();
                var AArr = A.GetArrayForRead();
                int c = 0;
                foreach (ushort a in A) {
                    retArr[c++] = a;
                }
                return ret;
            }
        }

Continue reading Uncommon data conversion with ILArray

Plotting Fun with ILNumerics and IronPython

Since the early days of IronPython, I keep shifting one bullet point down on my ToDo list:

* Evaluate options to use ILNumerics from IronPython

Several years ago there has been some attempts from ILNumerics users who successfully utilized ILNumerics from within IronPython. But despite our fascination for these attempts, we were not able to catch up and deeply evaluate all options for joining both projects. Years went by and Microsoft has dropped support for IronPython in the meantime. Nevertheless, a considerably large community seems to be active on IronPython. Finally, today is the day I am going to give this a first quick shot.

Continue reading Plotting Fun with ILNumerics and IronPython

Dark color schemes with ILPanel

I recently got a request for help in building an application, where ILPanel was supposed to create some plots with a dark background area. Dark color schemes are very popular in some industrial domains and ILNumerics’ ILPanel gives the full flexibility for supporting dark colors. Here comes a simple example:

Continue reading Dark color schemes with ILPanel

Large Object Heap Compaction – on Demand ??

In the 4.5.1 side-by-side update of the .NET framework a new feature has been introduced, which will really remove one annoyance for us: Edit & Continue for 64 bit debugging targets. That is really a nice one! Thanks a million, dear fellows in “the corp”!

Another useful one: One can now investigate the return value of functions during a debug session.

Now, while both features will certainly help to create better applications by helping you to get through your debug session more quickly and conveniently, another feature was introduced, which deserves a more critical look: now, there exist an option to explicitly compact the large object heap (LOH) during garbage collections. MSDN says:

If you assign the property a value of GCLargeObjectHeapCompactionMode.CompactOnce, the LOH is compacted during the next full blocking garbage collection, and the property value is reset to GCLargeObjectHeapCompactionMode.Default.

Hm… They state further:

You can compact the LOH immediately by using code like the following:

GCSettings.LargeObjectHeapCompactionMode = GCLargeObjectHeapCompactionMode.CompactOnce;
GC.Collect(); 

Ok. Now, it looks like there has been quite some demand for ‘a’ solution for a serious problem: LOH fragmentation. This basically happens all the time when large objects are created within your applications and relased and created again and released… you get the point: disadvantageous allocation pattern with ‘large’ objects will almost certainly lead to holes in the heap due to reclaimed objects, which are no longer there, but other objects still resisting in the corresponding chunk, so the chunk is not given back to the memory manager and OutOfMemoryExceptions are thrown rather early …

If all this sounds new and confusing to you – no wonder! This is probably, because you are using ILNumerics :) Its memory management prevents you reliably from having to deal with these issues. How? Heap fragmentation is caused by garbage. And the best way to handle garbage is to prevent from it, right? This is especially true for large objects and the .NET framework. And how would one prevent from garbage? By reusing your plastic bags until they start disintegrating and your eggs get in danger of falling through (and switching to a solid basket afterwards, I guess).

In terms of computers this means: reuse your memory instead of throwing it away! Especially for large objects this puts way too much pressure on the garbage collector and at the end it doesn’t even help, because there is still fragmentation going on on the heap. For ‘reusing’ we must save the memory (i.e. large arrays in our case) somewhere. This directly leads to a pooling strategy: once an ILArray is not used anymore – its storage is kept safe in a pool and used for the next ILArray.

That way, no fragmentation occurs! And just as in real life – keeping the environment clean gives you even more advantages. It helps the caches by presenting recently used memory and it protects the application from having to waste half the execution time in the GC. Luckily, the whole pooling in ILNumerics works completely transparent in the back. There is nothing one needs to do in order to gain all advantages, except following the simple rules of writing ILNumerics functions. ILNumerics keeps track of the lifetime of the arrays, safes their underlying System.Arrays in the ILNumerics memory pool, and finds and returns any suitable array for the next computation from here.

The pool is smart enough to learn what ‘suitable’ means: if no array is available with the exact length as requested, a next larger array will do just as well:

public ILRetArray CreateSymm(int m, int n) {
    using (ILScope.Enter()) {
        ILArray A = rand(m,n); 
        // some very complicated stuff here...
        A = A * A + 2.3; 
        return multiply(A,A.T);
    }
}

// use this function without worrying about your heap!
while (true) {
   dosomethingWithABigMatrix(CreateSymm(1000,2000)); // one can even vary the sizes here!
   // at this point, your heap is clean ! No fragmentation! No GC gen.2 collections ! 
}

Keep in mind, the next time you encounter an unexpected OutOfMemoryException, you can either go out and try to make use of that obscure GCSettings.LargeObjectHeapCompactionMode property, or … simply start using ILNumerics and forget about that problem at least.