Store 
xPC Target Benchmarks
The following figures are the result of executing xpcbench at the MATLAB command prompt. The function xpcbench displays benchmark results or runs the benchmarks on your actual target hardware. The benchmark results indicate the smallest base sample time that an xPC Target application can achieve for several Simulink models. The benchmark results compare the performance realized on several different sets of PC hardware, with different CPUs and clock rates.
xpcbench
and provides detailed information about the benchmarks.
XPCBENCH Execute xPC Target benchmarks and show result
XPCBENCH is very similar to the famous BENCH utility which comes with MATLAB. BENCH benchmarks this machine using MATLAB features and compares
the result with pre-stored benchmark results of other machines. By executing BENCH on this machine you get an impression of how this utility
works. XPCBENCH, as the name implies, is a utility to benchmark xPC Target performance. In other words, it measures the execution performance of
real-time applications on your target PC.
The pre-stored benchmarks results are labeled with CPU type and CPU clock rate of the PC which was used for this particular
benchmark. All pre-stored benchmarks have been built with Visual C/C++ 6.0 Professional. Some of the pre-stored benchmarks
results are based on builds with other compilers or versions. The benchmark labels reflect this:
.NET : Visual C/C++ .NET Professional
.NET03 : Visual C/C++ .NET 2003 Professional
XPCBENCH without an argument displays the pre-stored benchmark results without running the benchmark on this xPC Target environment. The first
figure displays the results of the 5 benchmarks executed on several target PCs with different types of CPUs. The benchmark results are displayed in
microseconds and stand for the minimal achievable base sample time for that real-time application (benchmark) executed on that specific target
PC. The second figure displays a summary of the benchmark results (relative performance) using a bar graph.
The 5 benchmarks are:
- Minimal: This is based on a "minimal" model just consisting
of three blocks (Constant, Gain, Termination). The model has neither
continuous nor discrete states. The result of this benchmark gives an impression about the target PC specific latencies.
- F14 : This is based on the standard Simulink example model
'f14'. Type f14 at the MATLAB command line prompt to open
the model and get an impression of the model complexity. The model contains
62 blocks and defines 10 continuous states.
- F14*5 : This is based on the standard Simulink example model 'f14'
as well, but this benchmark model consists of 5 f14 systems
modeled in subsystems. This benchmark is therefore 5 times more demanding
than benchmark 2.
(310 blocks, 50 continuous states)
- F14*10 : This benchmark contains 10 f14 systems (620 blocks, 100 continuous states)
- F14*25 : This benchmark contains 25 f14 systems (1550 blocks, 500 continuous states)
XPCBENCH('this') runs the benchmarks on your xPC Target environment.
At the time you invoke the benchmark execution xPC Target has to be
fully set up and functional; therefore, a target system has to be connected
and the xPC Target test XPCTEST has to have successfully
run. The benchmark execution takes about 10 minutes. This includes
the generation of
the Simulink benchmark models, the xPC Target application
build and the automatic search for the smallest achievable sample
time for all
5 benchmarks. After the execution of the benchmarks the
result will be displayed along with the pre-stored benchmark results
found for other machines.
See also XPCTEST.
Volunteer for usability testing
You are invited to participate in the usability testing of prototype software, which includes alpha, beta, or other prerelease versions of software under various stages of development by The MathWorks, Inc. This testing helps to gauge the ease of learning and using the software and to make recommendations for product enhancements.
