User`s guide

Instrumentation Performance
Instrumentation 3-5
❏ Statistics are accumulated in 32-bit variables on the target and in 64-bit
variables on the host. When the host polls the target for real-time
statistics, it resets the variables on the target. This minimizes space
requirements on the target while allowing you to keep statistics for long
test runs.
❏ You can specify the buffer size for LOG objects. The buffer size affects
the program’s data size and the time required to upload log data.
❏ For performance reasons, implicit hardware interrupt monitoring is
disabled by default. When disabled, there is no effect on performance.
When enabled, updating the data in statistics objects consumes between
20 and 30 instructions per interrupt for each interrupt monitored.
3.2.1 Instrumented Versus Non-instrumented Kernel
It is possible to disable support for the kernel instrumentation by changing the
global properties of the application. Within the Configuration Tool, the Global
Settings module has a property called Enable Real Time Analysis. By
unchecking this checkbox, you can achieve optimal code size and execution
speed. This is accomplished by linking with a DSP/BIOS library that does not
support the implicit instrumentation. However, this also has the effect of
removing support for DSP/BIOS Analysis Tools and explicit instrumentation
such as the LOG, TRC, and STS module APIs.
The Table 3-1 presents examples of code size increases when working with
the instrumented versus non-instrumented kernel. These figures provide a
general idea of the amount of code increase that can be expected when
working with the instrumented kernel. Table 3-1 uses as samples two
example projects that are shipped with Code Composer Studio software
which utilize many of the DSP/BIOS features. By including DSP/BIOS
modules, the example applications incorporate the instrumentation code.
Therefore the following numbers are representative of the amount of code
size incurred by the instrumentation, and are not affected by the size or
variations among users' applications. The first example, Slice, contains the
TSK, SEM, and PRD modules, while the second example, Echo, uses the
PRD and SWI modules. Neither example application is specifically designed
for minimizing code size.
For information on DSP/BIOS kernel performance benchmarks, including a
comparison of the instrumented versus non-instrumented kernels’
performances, see Application Report SPRA662, DSP/BIOS Timing
Benchmarks on the TMS320C6000 DSP.