Main Content

This example shows how Simulink selects a vector or a 2-D matrix from table data. In a 2-D table, the output vector can be a column or a row depending on the model configuration setting **Use algorithms optimized for row-major array layout**. In this example, the Direct Lookup Table algorithm is optimized for row-major array layout. The Direct Lookup Table algorithm that is optimized for column-major array layout is also presented as a reference. The code generated by using row-major interpolation algorithm performs with the best speed and memory usage when operating on table data with row-major array layout. The code generated by using column-major algorithm performs best with column-major array layout.

In this example, you:

Output a vector or a plane by using direct lookup with a column-major or a row-major algorithm.

Preserve semantics when switching from a column-major algorithm to a row-major algorithm.

Generate code by using a row-major algorithm and an array layout.

Open example models `rtwdemo_row_dlut3d_selvector`

and `rtwdemo_col_dlut3d_selvector`

.

open_system('rtwdemo_row_dlut3d_selvector'); open_system('rtwdemo_col_dlut3d_selvector');

**1.** By default, Simulink configures a model with column-major algorithm and column-major array layout. The model `rtwdemo_col_dlut3d_selvector`

is preconfigured to use column-major algorithms. Simulate the model and observe the output stored in workspace variable `yout`

.

**2.** To enable row-major algorithms, open the Configuration Parameters dialog box. On the **Math and Data Types** pane, select the configuration parameter Use algorithms optimized for row-major array layout Alternatively, in the MATLAB Command Window, enter:

set_param('rtwdemo_col_dlut3d_selvector','UseRowMajorAlgorithm','on');

**3.** On the **Simulation** tab, click **Run** to simulate the model. Observe the change in output dimension and numeric values logged in workspace variable `yout`

.

The column-major and row-major algorithms differ semantically in the selection of output vector. For example, in a 2-D table, Simulink selects a column vector as output for column-major algorithm and a row vector for row-major algorithm. In a table with a 3-D or higher dimension, Simulink selects the output vector from the first dimension of the table for a column-major algorithm and from the last dimension of the table for a row-major algorithm. The elements of the selected vector are contiguous in the table storage memory. In this example, the last dimension is the third dimension of the 3-D table. Due to semantic change, column-major and row-major direct lookup table algorithms output different vector size and numeric values.

These illustrations compare the vector output of row-major and column-major direct lookup table algorithms in a 3-D table.

For a direct lookup table that outputs a vector or 2-D matrix, the model semantics change when you switch from a column-major algorithm to a row-major algorithm. To preserve the semantics or ensure the same output given the same block I/O connections, you must permute the table data. Otherwise, Simulink propagates incorrect dimensions to downstream blocks.

**1.** The block `rtwdemo_col_dlut3d_selvector/Direct Lookup Table (n-D)`

has 3-D table data **T3d = reshape([1:24], 3,2,4)** and two input ports with value **0** and **1** (both are 0-based indices). The selected output vector is **T3d(:,1,2)** (1-based index) for a column-major algorithm. To preserve the semantics for a row-major algorithm on the same model, that is, select the same vector with same index port inputs, permute the table as **T3d_p = permute(T3d, [2,3,1])**. For a row-major algorithm, the selected vector is **T3d_p(1,2,:)**.

T3d_str = get_param('rtwdemo_col_dlut3d_selvector/Direct Lookup Table (n-D)','Table'); set_param('rtwdemo_col_dlut3d_selvector/Direct Lookup Table (n-D)','Table',... ['permute(',T3d_str,',[2,3,1])']);

**2.** When you import table data from a file, you must permute the table data in the file before importing it. This permutation keeps the table tunable throughout the simulation and code generation workflow.

After permuting the table data, Simulink configures the model `rtwdemo_col_dlut3d_selvector`

for row-major simulation. The model is equivalent to the preconfigured model `rtwdemo_row_dlut3d_selvector`

that has permuted table data and uses a row-major algorithm.

**1.** To set up these models for row-major code generation, open the Configuration Parameters dialog box. In addition to enabling the **Use algorithms optimized for row-major array layout** configuration parameter, on the **Code Generation > Interface** pane, set the configuration parameter Array layout to the `Row-Major`

option. This configuration parameter enables the model for row-major code generation. Alternatively, in the MATLAB Command Window, enter:

% For model 'rtwdemo_col_dlut3d_selvector' set_param('rtwdemo_col_dlut3d_selvector', 'ArrayLayout','Row-major'); % For model 'rtwdemo_row_dlut3d_selvector' set_param('rtwdemo_row_dlut3d_selvector', 'ArrayLayout','Row-major');

**2.** In the Direct Lookup Table (n-D) block dialog box, examine the permuted 3-D table data.

**3.** Change your current folder in MATLAB® to a writable folder. On the **C Code** tab, click **Build** to generate C code. In the generated code, the `memcpy`

function replaces the `for`

loops. Using `memcpy`

reduces the amount of memory for storing data. This optimization improves execution speed.

open_system('rtwdemo_row_dlut3d_selplane'); open_system('rtwdemo_col_dlut3d_selplane');

**1.** Open the example model `rtwdemo_row_dlut3d_selplane`

that outputs a plane or 2-D matrix from a 3-D table.

**2.** Simulate and generate code from the model by repeating the steps performed on `rtwdemo_col_dlut3d_selvector`

. The row-major and column-major direct lookup algorithms that output a 2-D matrix from a 3-D table are illustrated here.

close_system('rtwdemo_row_dlut3d_selvector',0); close_system('rtwdemo_col_dlut3d_selvector',0); close_system('rtwdemo_row_dlut3d_selplane',0); close_system('rtwdemo_col_dlut3d_selplane',0);