Embedded Target for Texas Instruments C6000 DSP Release Notes |
|
Known Software and Documentation Problems
This section includes a link to a description of known software and documentation problems in Version 2.0.
***UPDATE LINK WHEN GECKO HAS PRODUCT IN LIST***
You can see a list of known software and documentation problems in Version 2.0. If you are viewing these Release Notes in PDF form, please refer to the HTML form of the Release Notes, using either the Help browser or the MathWorks Web site and use the link provided.
For a list of bugs reported in the previous release that remain open, see Known Software and Documentation Problems.
Linking and Code Composer Studio Limitations
The following entries apply to linking operations.
- Having CCS visible on your desktop during CCS read or write operations slows down the data transfer process. Generally, transfer rates fall by a factor of four when CCS IDE is visible. When you do not need to interact with CCS IDE, use the function visible to force CCS to the background to avoid the slowdown, as shown in this example.
- When your work requires that you interact with CCS IDE during linking operations, place CCS IDE in the background temporarily while your program transfers large amounts of data in either direction. For example, the following code demonstrates changing the CCS IDE visibility state before and after read and write operations.
cc = ccsdsp:
.
. % Interact with CCS IDE.
.
vstate = isvisible(cc); % Store visibility state for CCS.
cc.visible(0); % Force CCS to the background.
x = cc.read('FF00','int32',[20 20],20); % Large read transfer.
cc.write('FF00',int32(randn(20)),20); % Large write transfer.
cc.visible(vstate); % Return CCS to its previous visibility
state.
.
. % Continue to interact with CCS IDE.
.
.
- CCS limits CCS target memory read and write data transfers to 32 KB. This restriction does not affect RTDX links and data transfers.
- Before you use the links, run the GEL initialization that is appropriate for your target processor. For RTDX communications to work properly on the C5402 DSK, you must initialize the board. To initialize your C5402 DSK, select C54x submenu --> C5402_DSK_Init from the GEL option on the CCS menu bar.
- To automate this process as part of GEL reset, edit C5000.gel as follows:
- In the .
gel
file, locate the section beginning hotmenu C5402_DSK_Init.
- In the hotmenu section, highlight and copy all instructions between the braces, { }.
- Paste your copied text into the startup function at the beginning of the .gel file.
- Your subsequent GEL resets run the C5402 DSK initialization.
- During memory transfer operations, you may receive a time-out error telling you that the time allowed for the operation expired. In most cases, the transfer was successful in spite of the error message. Once the transfer operation starts, it cannot be stopped or changed. Time may expire if the transfer completes but the completion message is delayed. Try setting the time-out value to a large number of seconds to avoid the error message.
- When you configure your target code to perform RTDX write operations (write data from the target to the host) on the C5402 DSK, you must poll the "successful emptying of the write buffer" status. The C6000, which allows polling or host-initiated interrupts, does not require polling.
- The following code provides an example of the polling syntax, where ochan is the write channel.
- On some occasions, CCS opens information dialogs in the background. You must respond to the dialogs before you can continue to work in MATLAB or CCS. Until you respond to the dialogs, MATLAB will not respond to input.
Targeting
- For reasons related to the Texas Instruments C compiler, you cannot use certain Simulink blocks for targeting. They are:
- Singular Value Decomposition block in the library dspfactors
- Pseudoinverse block in the library dspinverses
- SVD Solver block in the library dspsolvers
- These blocks are not included in the run-time libraries and do not build when you generate code from your model.
- For this prerelease of the Embedded Target, the LED overrun indicator has been removed.
Demos
In MATLAB Demos (select Help->Demos from the MATLAB menu bar), the Developer's Kit demonstration program LMS Adaptive Filtering runs slowly on DSP Starter Kit (DSK) boards. Design features of the board cause the slow processing across the RTDX link. Using the Texas Instruments XDS 510 JTAG/PC Controller, or an equivalent emulator, to connect your host to the DSK can alleviate the problem.
Data Types
C5000 and C6000 family processors handle integer data types differently. On C5000 family processors, 8-bit values do not have unique addresses. The C5000 family processors use 16-bit addresses natively. When you read and write to C5000 family processors, 8-bit integers in MATLAB are stored as 16-bit integers on the processor. In detail, the differences are:
- When you write an 8-bit integer from MATLAB to a C5000 family processor, MATLAB pads the value to 16 bits and the processor stores the value as a 16-bit integer.
- When you read an 8-bit integer from the processor into MATLAB, the read function takes 16 bits from processor memory, discards the eight most significant bits, and returns the remaining eight bits to MATLAB as the integer value.
For these reasons, scripts that run on C6000 family processors may not run without modification on C5000 family processors.
General Target Operations
When your target digital signal processor is running, CCS uses up to 99% of the CPU cycles on your PC. Consequently, some other operations may be very slow, such as running the MATLAB help system or processing other applications.
Documentation
In the Embedded Target for TI TMS320C6000 DSP Platform User's Guide, you can read about DSP/BIOS and targeting. For this prerelease, the DSP/BIOS features are not available in the product.
 | Upgrading from an Earlier Release | |