| Embedded Target for Texas Instruments C6000 DSPs | |
TI C6000 Code Generation Options
From this category, you choose from four options that define the way your code gets generated:
Incorporate DSP/BIOS determines whether the build process incorporates DSP/BIOS features in your generated code. When you select Incorporate DSP/BIOS, the build process inserts the DSP/BIOS options and files (the .cmd file that contains DSP/BIOS configuration information) in the generated code. The resulting code include instrumentation based on DSP/BIOS objects. Introducing DSP/BIOSTM provides details about the changes that occur to your generated code when you opt to include DSP/BIOS.
For profiling your generated code, the code generation options include the Profile performance at atomic subsystem boundaries option. When your model includes atomic subsystems, you can select this option to have Embedded Target for TI C6000 DSP generate a run-time report about the way your generated code performs when you run the code on your target. For more information about using code profiling, refer to Profiling Generated Code.
To allow you to specify whether the functions generated from blocks in you model are used inline, or by pointers, Inline DSP Blockset functions tells the compiler to inline DSP blockset function. Inlining functions can make your code run more efficiently (better optimized) at the expense of using more memory. As shown, the default setting uses inlining to optimize your generated code.
When you inline a block function, the compiler replaces each call to a block function with the equivalent function code from the static run-time library. If your model use the same block four times, your generated code contains four copies of the function. While this redundancy uses more memory, inline functions run more quickly than calls to the functions outside the generated code.
The final option in this category is Use target specific optimization for speed (allow LSB differences), which determines whether Embedded Target for TI C6000 DSP attempts to optimize the code generated from your model to make it run more quickly on your selected target. This option may not make any difference in some models.
Notice that selecting target specific optimization allows your generated code to differ from your simulation results in the least significant bit (LSB) for the outputs of optimized blocks. You should review the results of the optimized and simulation-true code to see that they are sufficiently close for your needs. For many models, the LSB differences do not matter. Clearing this option results in generated code whose results match your model simulation results.
The preferred way to use Use target... is to create your model, generate code from the model, and run the code on your target with profiling enabled. After you have your model and code running the way it should (generating the correct answers), try selecting Use target specific optimization for speed (allow LSB differences) and regenerating your code. Again, run the new code with profiling and compare the profile reports to see whether optimization improved the performance.
| | TI C6000 Target Selection | TI C6000 Compiler Options | |
