Release Note

When using dual processors any multi-threading or inter processor
communications which are not thread safe will cause a crash. These
crashes will happen on a single processor eventually, but usually will
be very very rarely and unrepeatable as they are completely dependent
on system timing, which is dependent on all the processes running on
machine and what they are doing at the moment. However with a dual
processor as that each thread can run on a different CPU the
likelihood of hitting the timing such that the thread communications
are unstable increases to the point where the problem is generally
repeatable by issuing the offending commands only a couple of times.

Release Note

TargetRun command may not function properly from 2nd time onwards when
TargetRun->Reset->Restart sequence is used in a loop.

Release Note

When re-registering a plug-in (by developers usually) an extra copy may
show up in the component manager.
 
To make sure this doesn't happen, whenever significant changes are made
to a plug-ins registry handling (for example, in the UpdateRegistry
function) the following method can be used to clear the registry before
running the new version of the plug-in:
1. Go to the directory where the plug-in is located. Move it (this is
only temporary so make sure you keep track of the location) to another
name (without the OCX extension).
2. Run CCStudio. You can close it right away. Alternatively you can
launch the comp_mgr.exe. Running the application after moving the OCX
will clean up the registry and the component manager's INI file. You
can confirm in component manager that the plug-in is gone.
3. Move the file back or compile the new version.
4. Run "regsvr32" on the file. From a prompt this would look like:
 
"regsvr32 [pluginname].ocx"
 
5. Launch CCStudio. A message box should be shown indicating that new
components were detected. Select "Yes" to enable them. Confirm in
component manager that there is now just one copy of the plug-in.

Release Note

Opening any project after a uninstall last update from CCS 2.2 causes
the following error
 
" No build tools found . Cannot open the project: Try re-installing the
application"
 
Reason:
The component manager does not manage the build tools in the same way
it manages the plug-ins, though the component manager shows the build
tools in its view.
 
Work-around:
To get around this problem, the user needs to re-register build tools.
 
Re-register the build tool dlls for ccs 2.12 re-enables the IDE to open
the projects.
 
For different ISAs, there are different build tool dlls. They dlls
start with "bti" prefixes and are located under the  installation.
For exampe,
c6000: bti6000.dll
c5500: bti5500.dll
c5400: bti5400.dll
 
Once the proper dlls are located, register the dlls in a dos command
window by using the regsvr32 program.
For example,
regsvr32 bti6000.dll
 
 
 
IMPORTANT NOTE of 2.21 fix:
The "fix" is not visible when the user upgrades 2.x to 2.21 and then
rollback to the previous version, because the bug is not fixed in the
uninstaller or component manager.
 
The user will see the same project-loading problem after the rollback.
This is because the user is still using the previous project server and
BTI after the rollback, and the previous project server is not able to
register the bti dll before loading it.
 
In this case, the user still have to perform the same work-around to
solve the project loading problem after the rollback. That is, to
regsvr32 btixxxx.dll
 
The fix is "visible" when the user upgrades 2.21 to a higher version.
 

Workaround

Re-register the BTI dlls for ccs 2.12 re-enables the IDE to open the
projects.
 
For different ISAs, there are different BTI dlls and they all start
with "bti" prefixes.
For exampe,
c6000: bti6000.dll
c5500: bti5500.dll
c5400: bti5400.dll
 
Once the proper dlls are located, register the dlls in a dos command
window by using regsvr32 program.
For example,
regsvr32 bti6000.dll

Release Note

Sometimes when trying to launch Update Advisor you get an error that
the page cannot be found.  The problem stems from a limitation in the
length of the url allowed.  This typically occurs if you have not yet
logged into MyTI and the MyTI server has to add information to the url
inorder to complete the login process.

Release Note

On a multiprocessor environment, CCStudio may occasionally not work as
expected after an emulator reset.  After the target has reconnected,
but
before the programs are reloaded onto the target, CCStudio seems to be
busy, or stuck, trying to finish the Reset Emulator operation.
 
If a disassembly window right-click, 'Start Address...' is performed,
or
some other GUI operation, CCStudio will get out of this situation and
resume, finishing the reload program operations.
 

Release Note

If a board power cycles while a program is loading or a program is
about
to start loading then the user will get data verification errors.
These
errors sometimes interfere with the emulator reset causing CCStudio to
go into a bad state and lock up.

Release Note

Not possible to access the debugger during standard input.
 
#include "stdio.h"
 
void main(void)
{
        char buffer[20];
 
        printf("\nEnter anything >");
        gets(buffer);
 
}
 
 

Release Note

After selecting some toolbar buttons the buttons appear
"highlighted" or "raised."  This does not affect the usability of the
buttons as they still display their "enabled" or "disabled" status
correctly.
 

Release Note

Inconsistent file extension names when using the Open File and Add
Files to Project.
 
For example, Open File has *.c for C Source Files, but Add Files to
Project has *.c or *.ccc extension.
 

Release Note

Customer Enhancement Request
 
When using GEL_System() gel command, spaces are not allowed in the DOS
command name. For example, I can use GEL_System("C:\\myProgram.exe")
but not GEL_System("C:\\my Program.exe").  While this may not seem like
a big problem, it is very limiting on file and path names.
 

Release Note

Volatile memory option isn't specifiable in GEL_MapAdd

Release Note

Terminal Font size 5 doesn't work in almost all windows
 
There are no plans to fix this bug.

Release Note

When deleting characters in the edit window the character may not
appear to be deleted (although it actually is) when using some bold
fonts.  It is recommended that bold fonts are not used in the edit
window.

Workaround

Please describe the workaround for this problem.

Release Note

When switching between documents in the edit area, sometimes the
document being switched to may not appear on the same page that it was
previously.
 

Release Note

CodeMaestro attempts to complete variable names using previously
defined and deleted variable names.  As you can see in the attached
screen shot CodeMaestro attempts to complete the variable "number2"
using "number" because "number" was previously used and was deleted.
Rebuilding the project also does not resolve this problem.
 
typedef struct number2
{
int n1;
int n2;
};
 
void main()
{
   a = 10;
   numb   /*CodeMaestro suggests number instead of number2*/
}

Release Note

After the 'Find in Files' selection is used from the toolbar, the
processor loading increases greatly on the PC and never decreases back
to the normal levels.

Release Note

The keyboard shortcuts do not work if you remap any of the already
assigned keys to new keys. As an example, I have attached 2 .key files,
the standard cc_app.key and new_app.key. The only difference between
cc_app.key and new_app.key is that the function key F5 has been
remapped to be Edit-Copy instead of Debug-Run. However, upon restarting
CCS using new_app.key map, CCS always defaults F5 to Debug-Run instead
of copy.
 
In order to duplicate the problem, load the file test_board.key from
the Options->Customize->Keyboard->filename. After doing this open a
project with a source file and try using F5 key to do Edit->copy. It
will still default to Debug->Run despite the fact that the options show
F5 being mapped to Edit -> Copy.
I have tried mapping Edit ->Copy to a function key that was unassigned
like F4 and this seems to work. However, if you change the mapping for
any key assigned at default this DOES NOT work.

Release Note

The file search dialog box doesn't show all of longer paths

Workaround

Note: By "file search" I presume the reference is to the standard File
Open/Save dialogs. These dialogs are not implemented by the CCS team
but rather, are standard Windows file I/O dialogs (hence why they are
similar to the ones seen in Dev Studio).
 
In Win2K and WinXP, Microsoft now allows the user to resize the width
of the window and in turn, the width of the dropdown list also
increases, allowing users to see more of the file path. (Win98/Me lets
the user resize the dialog but this doesn't cause the dropdown list to
increase in width and WinNT doesn't allowing resizing of standard
dialogs at all -- hence only on newer OSs can the user see the
additional path information).
 

Release Note

Sometimes after changing the options for the color of the text,
comments, and the background, the comment color seems to extend past
the
end of the comment.

Release Note

Enhancement request for automatic CR/LF at the end of all C/C++ and
Assembly files

Release Note

Change in functionality (from 2.12 to 2.2).  In CCS 2.12, the source
file is opened through the external editor when the customer clicks on
the file in the project window, however, in v.2.2 this is not
happening.  The file is opened through the CCS editor when the same
steps are performed.  The same functionality is seen when double
clicked on a syntax error.

Release Note

Highlighting text and scrolling with mouse wheel does not always work.

Release Note

Cut and Paste of text does not always repaint screen correctly, pasted
text will have characters missing but really there.

Release Note

When stepping through code monitoring a value with yellow cursor over
pop-ups, the value in pop-up does not change.

Release Note

When highlighting a work in one code window, like to copy a word, then
switch to another code window in CCS and paste the word, when you
return to the first code window the text is still highlighted, but
cursor is at another location, thus typing does not replace
highlighted text.

Release Note

Each time you type the structure followed by the period "." to access
a member the cursor jumps to another place in my code. Actually to the
start of previous function and then highlights that function
definition and starts over writing.

Release Note

In CCS 2.20.03 with Windows XP, opening a source file in a project and
then scrolling down with PgUp/PgDown keys leads CCS to crash

Workaround

Switching on CodeMaestro, when the source file is already open, does
not lead to a crash.

Release Note

File i/o does not take 0x0000 as the valid starting address. It works
 
 
without any problem for any other address other than this.
 
 

Release Note

Find/Replace dialog box sometimes hides found text underneath the
dialog box.
 
 

Workaround

Scroll the text or move the dialog box.

Release Note

When using large fonts in the desktop display settings, CCS does not
display the least significant digit of the column display on the status
bar when viewing a file with the cursor on column 10 or higher.
 
 

Workaround

Use small fonts in the display settings.
 

Release Note

Request for Find in files gives inconsistent results.

Release Note

The following GEL functions are ignored if they are in the StartUp()
routine:

GEL_MemoryLoad
GEL_MemorySave
GEL_PatchAssemblyFunction
GEL_SrcDirAddFunction
GEL_SrcDirRemoveAllFunction

 

Release Note

GEL_MemoryLoad writes only 800 bytes of memory of the COFF(.out) files
generated from CCS projects

Release Note

In CCS 2.20 for OMAP using a 1510 or a 1610, if you open the CPU
regsiters window and attempt to edit/modify any register such as PC,
SP or R0-R12 to any value greater than 0x7FFFFFFF or try to change the
sign bit, CCS returns an error "Invalid Input for Register R0 =
0x80000000. Enter a value in the range 0...0x0".
 
The same beavior seems to occur on any register, not just CPU
registers.

Workaround

Enter the register's you wish to work with in the watch window.  They
can even be placed in a new tab (right click and select add tab) and
they should be able to be saved with the workspace.
 
The watch window does not have this limitation when editing registers.

Release Note

When CCStudio is installed in a directory that has spaces in the
pathname (i.e. ) the menu
link "Help->CCS on the Web->Register OnLine" does not work correctly
(The spaces are removed from the path string).

It's still possible to register online by manually launching the HTML
file: "\docs\releasenotes\ccs_registration.html"
 

Release Note

CCS Generates "Cannot Add Two Pointers" message when a pointer and a
symbolic constant are added in the memory address window.

Release Note

Global Symbols having double underscores in their names will not get
displayed completely in the Symbol Browser

Release Note

"Invalid Line" error message is generated when using profile points.
Profile all funcions, then view the source code in Mixed Mode. When you
right-click on the profile points again, they now say "Invalid Line."
 

Release Note

Large disparity observed between the Excl. Total and the Incl. Total
when profiling a function in CCS 2.0 C6000.

Release Note

The CCS profiler cannot profile single functions that are greater than
32kb in size. Attempting to profile a function greater than this limit
will result in an "Out of memory" message box appearing.

Release Note

The profiler gives different Incl. Count for Function Profile Debug(-
gp)
and Full Symbolic Debug(-g) for a function.
 
It shows Incl. Count for the case of -gp option as 24 and for -g option
as 31.

Release Note

build project, add 2 functions to the
profile session, run the project, make small adjustments to code,
rebuild, my 2 functions have disappeared from the profile session, so
I re-add them, and run again.  This procedure will eventually cause a
fatal crash.
 
 

Release Note

Workspace fails to save disassembler type and enable clock status.  When
workspace is reloaded, these have their default values.
 
No plans to fix this bug.  The clock will become an obsolete item

Release Note

If you keep the profiler window open on C6000 CCS v2.20 (and maybe
2.21) and reload an .EXE, CCS sometimes crashes.

Release Note

When selecting "ALL" in the Reload Symbols Submenu that contains files
loaded with offsets, an error stating that the file cannot be found and
suggesting that it might have been moved or deleted would appear.

Workaround

Load each file indiviual through the Reload Symbols Sub menu.

Release Note

ccapp doesn't load COFF file after build

Release Note

CCS displays forward slashes in some paths

Release Note

If you are using the SDK 2.00 to open a project in Code Composer Studio
2.00 there are two methods available. It is recommended that you use
the
following method:
 
- use CCApplication to get the active CCDebugControlWindow.
- use CCDebugControlWindow to get CCDspTask.
- use CCDspTask to get CCProjects2.
- use CCProjects2::OpenProject to open the existing project.

Workaround

See release note.

Release Note

If a project(.out) and dependent project(.lib) have a file with same
file name then debugger may not be able to point to the correct file
 
Do not use same file names in the dependent projects.
 
The sequence of looking for source files is
- currently opened files
- active project
- all other projects
- debug source directories in the customize dialog.
 
If the user knows the file in advance, he can close the opened files of
same file name first and then set a particular project be active. Then
stepping through source code in a wrong file will not happen.

Workaround

Do not use same file names in the dependent projects.
 
The sequence of looking for source files is
- currently opened files
- active project
- all other projects
- debug source directories in the customize dialog.
 
If the user knows the file in advance, he can close the opened files of
same file name first and then set a particular project be active. Then
stepping through source code in a wrong file will not happen.

Release Note

Some of the hint text (at the bottom of CCS interface) in the Status
Bar is incorrect for the Register Window.  This was displaying
erroneous text.
 

Release Note

After the XDS560 is uninstalled, references to the XDS560 CCS drivers
will still appear in the center pane of CC setup.  Although visible,
these XDS560 drivers will be marked invalid.

Workaround

Please describe the workaround for this problem.

Release Note

Processor name specified in CCS setup does not reflect in ccBrd*.dat
file.
 
Open CCS setup and choose any board for System configuration. Select
the board under 'My System', and then modify the name of the processor
(for eg: Generator).
 
This name of the processor shows up in the PDM when you invoke CCS.
But the name of the processor does not get updated in the generated
ccBrd*.dat file. Here it still shows up as "cpu_0" etc.
 

Release Note

Watch window fails to show the structure name in the type column. It
displays this "as struct $$fake0".

Release Note

Expanding an array in Quick Watch doesn't show array values in CCS 2.20.
 

Release Note

The problem is with watch windows. If a watch window is open and
the .out file is loaded a second time. The *opening of structure* in
the watch window causes the CCS to crash.

Release Note

Every call of quick watch dialog in CCS 2.20 adds a new selection in
source window.

Release Note

The file ReleaseNote.txt, which contains install information for
the SDK, incorrectly tells you you copy the contents from
 
"ti\SDK\CCActXWiz\" to
 
"C:\Program Files\DevStudio\SharedIDE\bin\ide" .
 
For version 6.0 of the C++ development studio, the
correct path is:
 
"C:\Program Files\Microsoft Visual Studio\Common\MSDev98\Bin\IDE"
 
 

Release Note

The currently loaded program name is not displayed on the title bar if
the program is loaded via GEL file (with the GEL_Load function). If the
same program is loaded through CCS Load Program, then the name is
correctly displayed.

Release Note

Some of the files in the exclude.dat files do not exist, e.g. malloc.h
and types.h. The files are not installed by CCS, therefore, they should
not be included in the default exclude.dat file.
 
 

Release Note

The Project Load indicator on Code Composer Studio's status bar does
not correctly indicate the status of the load process.  This indicator
should display a blue-block (no a solid bar) progress bar.

Release Note

After using "Find in Files" functionality in CCS (menu) Edit -> Find in
Files, CPU load (from Task Manager) will remain close to 100%.

Release Note

Assembler .include directive does not like files such as --
2d1cfg.asm.
 
The assembler will generate an error message when trying to include
thile file --
 
"vectors.asm", ERROR!   at line 1: [E0003] Invalid constant
specification
                .include 2d1cfg.s62
Errors in Source - Assembler Aborted
Build Complete,
2 Errors, 0 Warnings, 0 Remarks.
 
Add quotations around the filename in the .asm file --
 
        .include "2d1cfg.s62"
 
 

Workaround

Put quotation marks around the filename:
 
   .include "2dicfg.s62"
 

Release Note

CCS does not show the correct assembly source code display after
loading the project that has the entry point and another assembly
header file included in it.
The debugger shows the control in the header file even if there is no
executable code in it, on program load.

Workaround

Do not include assembly header files in assembly source code, instead
write the code in source code itself and access it in another source
files by using the .ref directive for the data variables.
 

Release Note

Assembly optimizer comments out .equ directives and when the
initialized symbols are used out side of ".cproc" and ".endproc"
it generates the following symbol undefined errors.  This used to
assemble fine with version 4.00 of the assembler.
 
1 Assembly Error, No Assembly Warnings
"test.asm", ERROR!   at EOF: [E0300] The following symbols are
undefined:
  DATA_VALUE
 
Errors in Source - Assembler Aborted
 
Compile Complete,
  2 Errors, 0 Warnings, 0 Remarks.
 
************test.sa  start************************
 
DATA_VALUE  .equ        0x1000
       MVKL DATA_VALUE, A2
 
;_LABEL:             .cproc
;                    MVKL DATA_VALUE, A2
;                    .endproc
 
************test.sa  end***************************

Release Note

Program level optimizations do not work correctly in a multi-level
directory program. The assembly files object files are always put in
the base directory in default mode or the object directory if defined.
Program Level Optimization looks for them in the object directory + the
directory where the object files are defined to go.
 

Release Note

If the rts library is needed by a project and it is not added to the
project as a file, the user should specify the -x linker option in
addition to the -i or -l linker options. Otherwise, linker errors
may occur because the rts library will get processed by the linker
before the obj files.
 

Workaround

If the rts library is needed by a project and it is not added to the
project as a file, the user needs to specify the -x linker option in
addition to the -i or -l linker options. Otherwise, linker errors
may occur because the rts library will get processed by the linker
before the obj files.
 
 

Release Note

When using the 4.10 compile tools in a solaris environment, if an
object file with the same name exists but you do not have write
permissions to it (it might be owned by someone else)
the compiler will fail to create a new obj file even if it
has write access to the directory.
 
=====================================================================
 
A user needs to ensure that pre-existing .obj/.lst files in their
operating system can be overwritten by the user before invoking the
compiler to generate .obj/.lst files of the same name.

Release Note

The -fr option in CCS does not take into account the spaces in the path.
For eg. The option -fr "c:\my documents\debug" gets converted to
-fr "c:\mydocuments\debug"
 
This is a bug.
 
----------------------------------------------------------------
 
12/17/2002 (Tom Suchyta)
 
The C24x codegen tools were designed for the DOS environment and spaces
in file names or directories are not supported.

Release Note

Watch Window points to the same static variable if the there are 2
differenct static variable with same name. The static variable are
declared in the same file, but in diffrent function locally.
 
Also (minor issue) The watch window is able to displays a local static
variable even though you are not in the scope of the variable. You just
have to be in the same file as where the local variable is defined.

Release Note

Compiler generates a warning message "symbol 'xxxx' redeclared with
incompatible type" when -pm option is selected in the project build
options.
 
This happens when mixing C and C++ source code where the C file
defines a variable (i.e. int x;) and the C++ file references that
variable (i.e. extern int x;).
 
 

Release Note

1) The statement
_mem2(&a[7]) = 0x1234;
does load the values into the array in in both big endian and little
endian modes.
 
2) In Big endian mode
        as = _mem2(&a[1]);
'as' will not gett loaded with correct value.

Release Note

The compiler should generate an error when it sees a function call
without any prototype. As Generated code for a function call without a
prototype will generate wrong code.

Release Note

Compiler generates "INTERNAL ERROR: no match for UMULT"
on compiling the below code.
 
void main ()
{
 
}
 
void f( void )
{
   int params[ 3 ]={0,0,0};
 
    params[2];
 
}

Release Note

Compiler may store double word values in non-doubleword aligned address
 
========================================================================
 
User needs to ensure that if they are using their own code for memory
allocation that variables of type double, long double, or long are
aligned to a double-word boundary.

Release Note

Application error occurs and Dr Watson window pops up when compiling
the following source file:
 
command for compiling:
cl6x -c -mv6400 exec.c
 
The source file(exec.c) contents is as follows:
----------------------------------------------
#include 
void main()
{
  printf("Hello\n");
}
void f() {}
interrupt void i() {f();}
interrupt void vwxyz00() {f();}
----------------------------------------------

Workaround

Any of the following actions results in a clean compile:
 
        1)      Remove the "-mv6400" switch.
        2)      Remove the either "interrupt" keyword from exec.c.
        3)      Remove the either "f()" call.
        4)      Alter the identifier "vwxyz00" in one of the
                following ways:
 
                a)      insert or delete a character before the "00"
                b)      alter the "00" in any way

Release Note

Compiler may not handle reinterpret_cast keyword correctly. Following
code generates different results.
 
#include 
#include 
 
using namespace std;
volatile int a = 0;
char *cc = "oooo/n";
 
void isrnextdata(void)
{
  a = 0x25;
}
 
int main(void)
{
  cc = "toto/n";
  char *ca = reinterpret_cast(reinterpret_cast(0x100));
  printf("ca: %08X\n", ca);
  char *cb = reinterpret_cast(reinterpret_cast
(isrnextdata));
  printf("cb: %08X\n", cb);
  printf("reinterpret_cast: %08X\n", (reinterpret_cast
(reinterpret_cast(isrnextdata)))
                   - (reinterpret_cast(reinterpret_cast
(0x100))));
  printf("Direct          : %08X\n", cb - ca);
}
 
=======================================================================
 
The C/C++ standard does not allow for conversion between function
pointers and data object pointers. Therefore it is illegal to convert
between a function pointer and a void pointer and will result in
undefined behavior.

Release Note

New[ ] operator does not put 8 byte header on size of allocated memory

Release Note

Dividing zero by zero results in INF(Infinite) rather than NaN(Not a
Number)
 
const double x = 0.0 / 0.0;
const double y = 1.0 / 0.0;
results in x being assigned the value +Inf (0x7FF0000000000000) rather
than NaN of some flavour (e.g. 0x7FFC000000000000); y is correctly set
to +Inf.

Release Note

Using a file name that begins with numbers (22.obj) and a long enough
list of options passed to the linker --
 
cl6x -z -llnk.cmd -lrts6200.lib 22.obj -lrts6200.lib -lrts6200.lib
-lrts6200.lib -lrts6200.lib -lrts6200.lib -lrts6200.lib -lrts62.lib00
-lrts6200.lib -lrts6200.lib -lrts6200.lib
 
The linker will generate an error message about not being able to open
file '.obj' or file 'obj.obj'.
 
 

Workaround

Use a shorter command line, with less options for the linker.
 
Change the name of the file or directory to not start with a number or a '.'
 
For non-C++, try invoking the linker directly.

Release Note

The Optimiser, for c6x CCS2.2 CGT v 4.32, appears to assume that any
floating point value is equal to itself. This is not true - NaN is not
equal to itself. Thus, as a very simplified example, the function
 
int isnan(double val)
{
  return (val != val);
}
 
just returns zero on all occasions: which is wrong.

Release Note

Adding a conditional branch around a loop in linear assembly code
where the target label does not contain any code before the end
of the linear assembly function will cause the assembly optimizer
to generate multiple uses of the loop label.  This will be flagged
as an error by the assembler.
 

foo:   .proc  A4, B4, A6
       ...
 [!reg1]  B    go_back
 
loop:
       <linear assembly loop code>
 [reg2]   B    loop
 
go_back:
        .endproc

Release Note

Unexpected pre-incrementation of pointer occurs with -mv6400
 
void RLIB_swap(unsigned char *uPdem,
               const unsigned int nSymbols)
{
  unsigned int symbInd;
  unsigned int* restrict temp_p;
 
  temp_p = (unsigned int*) uPdem;
  for (symbInd = 0; symbInd < nSymbols * RLIB_MAX_N_RFR; symbInd++)
  {
    *temp_p++ = _rotl(*temp_p, 16);
0000077C 023C82E6            LDW.D2T2      *+SP[0x4],B4
00000780 00006000            NOP           4
00000784 02908942            ADD.D2        B4,0x4,B5
00000788 02BC82F6            STW.D2T2      B5,*+SP[0x4]
0000078C 029402E6            LDW.D2T2      *+B5[0x0],B5
00000790 00006000            NOP           4
00000794 029607B2            ROTL.M2       B5,0x10,B5
00000798 00000000            NOP
0000079C 029002F6            STW.D2T2      B5,*+B4[0x0]
000007A0 00002000            NOP           2
  }
}
 
=======================================================================
 
This operation is undefined. The increment can occur before or after the
_rotl takes place. Please see the C standard Section 6.5.2.4 ,# 2

Release Note

Compiling linear assembly code with the debug option may not show
you the right registers when you attempt to watch your variables
defined with the .reg directive in Code Composer Studio.

Release Note

Compiled the code below for C67x.
 
void main()
{
    unsigned int i, j;
    float f;
 
    i = 0x81000000;
    f = (float)i;
    j = (unsigned int)f;
    printf("i = 0x%8x, f = %f, j = 0x%8x\n", i, f, j);
}
 
The compiler generates a SPTRUNC instruction for "j = (unsigned int)f;".
This will put an incorrect value in j, as the SPTRUNC instruction will
return 0x7fffffff or 0x80000000 if any kind of overflow (f > INT_MAX
or f < INT_MIN) occurs.  So j will get the value 0x7fffffff instead of
0x81000000.

Release Note

The compiler generates code that accesses memory out of order with -o2
-mv6400 on the following code --
 
       void foo(unsigned char *e) {
            unsigned int r = 16, i, j;
            unsigned char *p = e+15;
 
            for (i = 0; i < 2; i++) {
                *(p++) = *(e+r++); *(p++) = *(e+r++);
                *(p++) = *(e+r++); *(p++) = *(e+r++);
                for (j = 10; j < 21; j++) {
                    *(p++) = *(e+r++); *(p++) = *(e+r++);
                    *(p++) = *(e+r++); r++;
                }
            }
        }

Release Note

Assembly optimizer may remove multiplies when machine registers are used
in the linear assembly source code. In the following loop the optimizer
replaces with one mpy instruction and so half the values of the array b
will be wrong.
 
_scale: .cproc  a, b, scaleF, len
        .reg    cnt
loop:
        LDDW    *a++, A1:A0
        MPY     scaleF, A1, A1
        MPY     scaleF, A0, A0
        STDW    A1:A0, *b++
        BDEC    loop, cnt
 
 

Workaround

Use only variables as follows.
 
_scale: .cproc  a, b, scaleF, len
        .reg    cnt, x:y
loop:
        LDDW    *a++, x:y
        MPY             scaleF, x, x
        MPY             scaleF, y, y
        STDW    x:y, *b++
        BDEC    loop, cnt
        .endproc

Release Note

File: C6x.h
version: 4.32
Product: CCS 2.20
 
This file needs to contain support for the Debug Interrupt Enable
Register (DIER) and is missing.
 
The DIER is required in order to select RT interrupts from within the
application.  These are the interrupts that the application selects to
notify the on-chip emulation logic that, when the debugger has
configured the target in realtime mode and when the target is halted,
must continue to be serviced.
 
If this bug is encountered in a release then it can be worked around
in the following manner:
 
Add the following lines to the c6x.h file.
 
#ifdef _TMS320C6400
extern __cregister volatile unsigned int DIER;
#endif
 

Workaround

If this bug is encountered in a release then it can be worked around in
the following manner:
 
Add the following lines to the c6x.h file.
 
#ifdef _TMS320C6400
extern __cregister volatile unsigned int DIER;
#endif

Release Note

Compiling C++ code that contains templates with -o3 may cause cg6x to
abnormally terminate.
 
 

Release Note

std::type_info::operator==() may return FALSE for equivalent
fundamental or derived types when typedefs for them are used.

Release Note

It seems as if the linker is selecting a library irrespective of the
order mentioned. If a same function present in 2 libraries , lib A and
lib B . If lib A is specified in the build options and lib B in a .cmd
file. The library B is called when the function is evoked. Ideally the
lib A should be called.
 
Always the function present in .cmd get a higher precedence than what
is mentioned in the build option. Ideally the build option should get a
higher precedence.
 
----------------------------------------------------------------------
In actuality the linker is not really doing this. For example:
 
If you compile on the command line, the lib specified first is
searched first, ie;
 
cl6x -qq hello.c -z -llib1.lib -lrts6700.lib my.cmd -o hello.out
load6x -q hello
 
lib1 is activated!
 
However,
 
cl6x -qq hello.c -z my.cmd -lrts6700.lib -llib1.lib -o hello.out
load6x -q hello
 
lib2 is activated!
 
Within code composer, the cmd file is read first, hence the lib
specified in it is always used. Please also see the Section on "Object
Libraries" in the Linker Manual.
 

Release Note

The linker did not combine '.bss' and '.bss:c' sections (the same
for '.cinit' and '.cinit:c' sections) when the -ar linker option was
used.  This can cause the linker to generate relocation error messages.
 
 

Workaround

The user and/or BIOS can force '.bss:c' to get combined into
the '.bss' output section (and similarly for '.cinit:c'):
 
SECTIONS
{
    ...
    .bss: { *(.bss) } > IDRAM
    .cinit: { *(.cinit) } > SBSRAM
}
 
 

Release Note

The Linker(cl6x ver 4.32) crashes when Linking the object files in a
particular order.
 
On Windows 2000 the following error message pops up:
"lnk6x.exe has generated errors and will be closed by Windows.  You
will need to restart the program. An error log is being created."
 
On WinNT the the following error appears:
"unknown software exception (0xc00000fd) ..."
 
Workaround:
The bug can be avoided by using the -b linker option to disable the
type-merging pass.

Release Note

When compiling code with the -o2 option, the following error message is
generated, INTERNAL ERROR: no match for ICALL.
 
void foo(..., struct1S* const restrict p5, ...)
{   ...
    unsigned int a;
    signed int tmp;
    ...
    for (k = 0; k < X; k++) { ...
        tmp = ((signed long)a * (signed long)tmp +
               (signed long)(8 - a) *
               (signed long)p5->r[k][0]) >> 3;
        ... }
}

Release Note

Compiler uses the CMPGT instruction instead of the CMPGTU instruction
to compare unsigned values. This becomes a problem because the compiler
then chooses to use CLR.

Release Note

Compilation terminated with the message
"ERROR - OUT OF MEMORY - OPTIMISER ABORTED" when using -o2/o3
compilation option.
 
This occurs on files that contain switch statments containing dozens
of cases containing non-trivial code for each case.
 
 

Release Note

Optimiser is terminated by windows with the message "opt6x.exe has
generated errors and will be closed by windows. You will need to
restart this program"

Release Note

The optimizer abnormally terminates when using either -o2 and -o3
optimisation level and the source code uses an UNROLL pragma to ask
that a for loop be unrolled hundreds of times.
 
  #pragma MUST_ITERATE(702, 702, 702);
  #pragma UNROLL(702);
  for(i = 1; i < 702 - 1; i++) {
    *p += diff >> k;
    x = *(p = (diff > 0) ? p1 : p2);
    if(x < min) x = min;
    *p = x;
    p1++;
    p2++;
  }

Release Note

Compiling with -o2 causes incorrect code to be generated for the
following example --
 
#include 
 
void main()
{
    int i = 0;
 
    while (i < 4)
    {
        if      (i==0) { printf(" Statement ZERO , i=%d \n",i); }
        else if (i==1) { printf(" Statement ONE ,  i=%d \n",i); }
        else if (i==2) { printf(" Statement TWO ,  i=%d \n",i); }
        else
        {
           printf("Else, i=%d \n",i);
           i++;
           continue;
        }
 
        printf("Iteration done, i=%d \n\n",i);
        i++;
    }
    printf("\nEnd of test, i=%d \n", i);
}
 
A workaround would be to delete the 'continue;' statement and add an
if-statement around the last 2 statements of the while loop --
 
        if (i <= 2)
        {
            printf("Iteration done, i=%d \n\n",i);
            i++;
        }
    }
    printf("\nEnd of test, i=%d\n, i);
}

Release Note

The Optimizer doesn't realize arguments can change if passed though
function pointer.

Release Note

The optimizer(-o2/-o3) may not calculate the index value of an array
correctly in some scenarios and may return wrong value.

Release Note

If a FOR loop is completely unrolled (using #pragma UNROLL) and
optimised -o3, then the loop can exit prematurely. This occurs if the
only statement within the FOR loop is an IF statement. If the IF
statement returns false, then the compiler has generated code that
will exit the FOR loop prematurely.
 
Removing the '#pragma UNROLL' fixes the problem.
 
for (i = 0; i < X; i++) if (Mask & (0x1<
Workaround
Removing the '#pragma UNROLL' seems to fix the problem.

Release Note

The compiler with optimization enabled (-o2 or -o3) for C67x generates
incorrect code for the following --
 
    extern float b[], c[];
    float a, *ptr, *p;
    for (i = 0; i < X; i++)
    {
        a = 0;
        p = ptr++;
        for (j = 0; j < Y; j++) a += b[j] * *p--;
        c[i] = a;
    }

Release Note

With -o2 or -o3 optimization, the compiler generates incorrect code
which reads the 'i' attribute before it is defined.  This causes the
operation a.b.get_i() to return 0 instead of 1234.
 
class b_c {
public:
  b_c(void) : i(0) {}
  template void f(T& t)
    {
      if (i == 0) { i = 1234; }
      else        { i = 4321; }
    }
  int get_i(void) const { return i; }
 
protected:
  int i;
};
 
class a_c {
public:
  b_c b;
};
 
int main(void)
{
  a_c a;
  int x = 1;
 
  a.b.f(x);
  printf("i = %d\n", a.b.get_i());
  x = a.b.get_i();
  printf("x = %d\n", x);
  return 0;
}

Workaround

Add the volatile keyword to the attribute 'i'.
 
class b_c {
public:
  b_c( void ) : i( 0 ) {}
  template void f( T& t )
    {
      if (i == 0) { i = 1234; }
      else        { i = 4321; }
    }
  int get_i(void) const { return i; }
 
protected:
  volatile int i;  /* ADDED VOLATILE KEYWORD */
};

Release Note

The compiler may generate incorrect code when copying input data to a
local structure pointer with -o2 or -o3.
 
    typedef struct { ... } Y;
    void foo(unsigned x, ...) { Y *z = (Y*)(&x); ... }
 

Release Note

Optimizer may not return correct values when handling function
pointers. The function bugfunc() always returns zero in -o1 and -o2
optimizations.
 
#include 
typedef void (*FuncPtr)(int *);
typedef struct _Cinfo {
        FuncPtr func1;
        int foo;
} Cinfo;
 
int bugfunc(Cinfo *cinfo)
{
    int retval;
    retval = 0;  /* Set retval to zero. */
    (*cinfo->func1)(&retval);
    return(retval);  // Expected to return 42
}
 
void myfunc(int *iptr)
{    *iptr = 42;  }
 
main()
{
    Cinfo cinfo;
 
    cinfo.foo = 57;  /* Initialize the cinfo structure. */
    cinfo.func1 = myfunc;
    cinfo.foo = bugfunc(&cinfo);// The return value is always zero in -
o1 and -o2 optimziations
 
    printf("%d\n", cinfo.foo); /* At this point, cinfo.foo should be
equal to 42. */
}

Workaround

If -o1 or -o2 is necessary, you can add the volatile keyword in front of
'retval' (and added to the typedef -- void (*FuncPtr) (volatile int *);).
 
Use -o0/-o3 optimziations otherwise.

Release Note

Optimizer may give different results

Release Note

Optimzier terminates abnormally for -o2 and -o3 optimizations for the
following code.
 
#include 
#include 
#include 
#include 
#include 
#include 
 
void test_floor (void);
main ()
{
        test_floor();
}
 
static inline double xxx(double x)
{
        double y;
        return modf(x, &y);
}
 
void test_floor()
{
  double      f;
  long        i;
  int      x[] =  {1,2};
 
  for (i = 0; i < 2; i++)
  {
    f = xxx(x[i]);
    if (f == 1) {    f = f+1;    }
  }
}
 

Workaround

Use "int  i" in place of "long  i" in the function test_floor()

Release Note

Optimizer may wrongly pass address of structure member, when structure
member is passed to a function.

Release Note

The code generated with INLINE is not effectively optimized.
Optimization result is not the same as in the case of manually in
lining the code.

Release Note

With -o2/-o3 optimization, the compiler may generate incorrect code.
This occurs when casting a variable to another class type when
assigning
to a class object and immediately using that class object.
 
For example --
 
struct X { int * rNum; int nSize; };
static X y;
 
class K { ... }
 
K K::operator >> (int n) const { ... }
 
void foo()
{
    unsigned int out[2];
 
    out[1] = bar( &y );
    out[2] = bar( &y );
    /* out may not get copied into temp before temp is used */
    K temp = *((K*)(out));
    printf("%d ", (unsigned int)(temp >> 32));
    printf("%d ", (unsigned int)(temp));
}
 

Release Note

Long and complex template names can cause the compiler to overflow its
buffers when compiling with optimization enabled.

Release Note

Code generation tools (v 4.32 and 4.36) generate incorrect code with
options -mv6400 and -o2/-o3 optimization,
 
int foo(unsigned x, unsigned y, unsigned *z)
{
    unsigned a = x * y, b = a & ~0x03, i;
    int Sum = 0;
    for( i = 0; i < b; ) {
        Sum += w[z[i++]]; Sum += w[z[i++]];
        Sum += w[z[i++]]; Sum += w[z[i++]];
    }
    for( ; i < a; i++) Sum += w[z[i]];
    return Sum;
}

Release Note

Long bitfield types are not supportd in C++ programs on the C6x.
 

Release Note

Using pointer to member operations in C++ (.* or ->*) with the -o2
or -o3 option can cause the compiler to lose some memory dependency
information, leading to incorrect code generation.
 
 

Release Note

C++ allows enumerators larger than an int.  Currently, the C6x
compiler forces all enumerators to ints.

Release Note

Covariant return types of virtual functions are not fully supported.
 

Release Note

"Extern inline" functions are not supported by the C6x C/C++ compiler.
The V3.00 compiler/code generator does not create globally accessible
code for functions which are declared inline. A simple example is:
 
inline int x() { return 1; }
int y() {return 2;}
 
When compiled with cl6x -k -c test.c a warning is produced:
"test.c", line 1: warning: function "x" was declared but never
referenced
 
and the resulting assembler file (test.asm) does not contain any code
for x().
 
See section "Workaround" for a workaround to this problem.
 
 

Workaround

As a workaround that involves the least amount of code modification, I
suggest that the user define a macro to be "static inline" and use that
macro as the inline modifier for these functions.  In the normal case,
these functions will be inlined.  To generate callable versions of these
functions, also compile the header files containing the inlined
functions with this new macro undefined.
 

For example:
 
----------------------------------------------------------------------
header.h:
 
#ifdef NO_INLINE
#define STATIC_INLINE
#else
#define STATIC_INLINE static inline
#endif
 
/* Replace the inline modifier with STATIC_INLINE for all inlined */
/* functions which you also want to generate a callable version.  */
STATIC_INLINE int x() { return 1; }
 
----------------------------------------------------------------------
foo.c:
 
/* Inline occurs normally in this file */
#include "header.h"
 
int y() { return x(); }
 
----------------------------------------------------------------------
body.c:
 
/* Compile this file to generate callable versions of all inline */
/* functions which use the STATIC_INLINE macro.                  */
#define NO_INLINE
#include "header.h"

Release Note

When stepping through a C 'switch' statement the PC (program counter)
register will always jump into the last 'case' statement before
returning to the correct 'case'.  Program flow will continue normally.

Release Note

The compiler does not correctly align an array of STRUCT_ALIGNed
structures.  It does, however, place the structures the correct
distance apart.

Release Note

The implementation of va_start() and va_arg() cannot handle a variadic
(passed through "...") argument that is a struct, nor any variadic
argument when the last known argument is a struct. This results in
incorrect values from va_arg().
 

Release Note

The qsort prototype in stdlib.h is incorrect.

Workaround

Manually change the declaration of qsort in stdlib.h to be
 
 _CODE_ACCESS void qsort(void *_base, size_t _nmemb, size_t _size,
                         int (*_compar)(const void *, const void *));

Release Note

There is a problem with the path mentioned under the linker tab for the
example found in C:\ti\c6000\ssl\examples\6211dsk\mcbsp1.
 
The library search path indicates c:\ti\c6000\ssl\include instead of
c:\ti\C6000\ssl\lib.

Release Note

There's a bug in the BIOS Configuration cdb GUI for C64xx.cdb.
 
On the C64x the Limit on the EDMA - Table Numbers is 0-60, this was the
C6x1x range, on C64x the range should be 0-18.  This is in the
Parameter RAM table setup
(i.e. when setting up a Table Handle to use a Table Pointer.)
 
It is also documented in help under EDMA_allocTable API.
 

Release Note

The EMIF SDRAM Refresh Period will not accept values outside 1...4095

Release Note

Enhancement request to allow multiple header files entries under CSL
Extern Declaration.

Workaround

The user can create their own header file(suppose myfile.h). He/She can
include all the header files in myfile.h. This myfile.h can then be
entered under the properties of the CSL Extern Declaration. This
would solve the problem of entering multiple header files.

Release Note

In the CSL, insert a new "DMA Global Register" object and access its
properties. The entry field wording of the "Global Index" tab is a bit
misleading because the values to enter are not "Count"s. The entry
field labels should be changed to "Offset" or something similar.

Release Note

Requested for a  message box instead of a comment field to enter the
comments in DMA and MCBSP cfg properties in the CSL so that the intended
operation about the DMA or MCBSP setup can be entered as a paragraph.

Release Note

EDMA_OPT_PRI_DEFAULT is implemented as 0x0 for non-C64x. But this
priority must not be used for C6x1x EDMA because it is reserved for L2
transfers. The CSL default value should be something other than 0.

Release Note

Enhancement for the BIOS config tool:
Create a new .cdb from a C6x11 seed. Insert a new EDMA configuration.
Go to the properties and select the "Operation Mode" tab. The pre-
selected (default) priority level is "Urgent(64x)". When this priority
is left unchanged on the C6x11, the EDMA transfer doesn't start (PRI=0
is not allowed). Hence, the pre-selected default value for the C6211
and C6711 .cdb seeds in the edmaCfg properties should be something
other than "Urgent(64x)".

Release Note

PCI is actually not available on the 6416 TEB/DSK
by default and neither on the simulator.
PCI is multiplexed with HPI.
The customer might be running the code on one of
these, mentioned above.
 
I verified that PCI_intEnable is actually working fine
by running the testcase on NVDK, which is the video kit,
where the PCI is brought out and is enabled and
HPI is disabled.

Release Note

The SDRAM Control Register in EMIF control Registers in the cdb file
shows TRC calculation as
TRC = trc/ (clock period - 1)
 
but actually the value is calcuated as
TRC = (trc/clock period) - 1
 
Similarly TRP and TRCD also needs to be modified.

Release Note

EDMA_enableChannel() call will not be generated when channel is
disabled.

Release Note

The ELM & FRM IDX values entered are checked using with correct
condition to accept all signed 16 bit nos.

Release Note

The code for PCI_intEnable is wrong.
 
The problem is that the PCI_intEnable code does an 'add' of the current
value and the bit to enable, instead of a bitwise OR.  The semantics of
these two operations are very different.
 
 
 

Release Note

CACHE_setL2Mode() is modified to use 'wbInvL1d' function instead
of 'invalidate' function.

Workaround

Workaround:
Manually implement CACHE_setL2Mode according to SPRU610:
- Set .cdb -> Global Settings -> Properties -> C641x -> C641x L2 Mode -
CCFG(L2MODE) to "4-way cache (0k)"
- Insert the following function before main():
static void configureCacheTo256K();
void configureCacheTo256K()
{
        unsigned int gie;
        gie = IRQ_globalDisable();
        // write back data from L1D to L2
        // which is in the L2 address range to be used as cache now
        // (was missing in 2.20.03 CSL)
        // last parameter in following function call is word count, not
byte count
        // needs to be changed for other cache sizes than 256K
        CACHE_flush( CACHE_L1D, (void *)0x000C0000, 0xFFFF );
        // Clean L2 and wait for completion
        CACHE_clean(CACHE_L2ALL, 0, 0);
        // Last parameter of following macro needs to be changed for
other
        // cache sizes than 256K
        CACHE_FSET(CCFG,L2MODE,CACHE_256KCACHE);
        CACHE_FGET(CCFG,L2MODE);
        asm(" NOP 4");
        asm(" NOP 4");
        IRQ_globalRestore(gie);
}
- Insert a call to this function as the first thing in main().
 
 
 

Release Note

IOM_ABORT and IOM_FLUSH (for input) should reset the CIRC buffer
pointers so I/O is completely discarded.  For output, any characters in
the CIRC buffer will still be transferred to the UART.  For input, the
characters in the CIRC buffer will be input for the next I/O call.
 

Release Note

On the C6713, the number of systems events reaching the EDMA is more
than 16, but EDMA supports only 16 events at any point of time. One
needs to select the events by programming the event select register.
 
The McBSP events are selected by default, but one could modify them by
reprogramming the event select register in order to select other
events.
 
CSL supports an EDMA event allocation mechanism for the C6713. It is
something similar to the interrupt selector API.
 
The ideal usage of EDMA transfers is,
 
1.Query EDMA CSL for event slots.
2.Use that event for doing EDMA transfers.
 
The C6x1x generic McBSP driver does not do the above step to program
the EDMA transfers. It just assumes that event select register is
still a default value. This assumption could be wrong, so it needs to
follow the above steps before programming the EDMA transfers.  However
this is applicable only to C6713 device and not to other devices.

Release Note

"." in ".text" and .align 32 directive are missing in the single
precision exp function (expsp.asm) in fast67x rts library

Workaround

The line with .sect text should be modified to read:
 
.sect .text
 
 
 

Release Note

adddp and subdp used in the fastrts62x.lib causes undefined symbols.

Workaround

Symbols defined with capitalization of all letters. Needs to be all
lower case.
 
 

Release Note

It is apparent that the first sentence is missing in spra653 page 4
paragraph 4,
 
Furthermore Figure 1. Block Diagram of the AUDIO DEMO Example needs to
be updated according to latest changes in audio example.

Release Note

One shot PRD require explicit call to PRD_start.
 
The specified bug occurs as PRD_start was not called explicitly.

Release Note

The values displayed for the BIOS SWI 'KNL_swi' in the 'Statistics
View' window (selected from "DSP/BIOS->Statistics View" in CCS) are
inaccurate and should not be considered to hold any meaningful
information regarding the SWI's actual ready-to-completion times.
 
This is a by-product of the special way in which this distinguished SWI
is used by DSP/BIOS.  Please ignore all values for 'KNL_swi'.

Release Note

Fixed bug in STS view that causes it to report inaccurate statistics
when RTDX is not enabled.

Release Note

The current DSP/BIOS configuration files do not support the 6713.

To workaround this problem:
0. Set chip type in Global Settings to “other”.

1. Uncheck the "621x/671x - Configure L2 Memory Settings" check box
   located in the Global Settings Properties page under the
   "621x/671x" tab.

2. Then use the following CSL API's to correctly setup the CACHE.
   This function then can be plugged into the User init function
   in the Global Settings properties.

   CSL_init();
 
   CACHE_setL2Mode() - Will setup the L2 CACHE control register.
   CACHE_enableCaching() - Will setup the MAR registers.

The Global init function will be called during system intialization
time and will initialize the CACHE before other BIOS initialization.

3. Link your program with the csl6713.lib library by adding this to
your project or adding it to the linker options in the build options.
 

Workaround

To workaround this problem:
1. Uncheck the "621x/671x - Configure L2 Memory Settings" check box
   located in the Global Settings Properties page under the
   "621x/671x" tab.
 
2. Then use the following CSL API's to correctly setup the CACHE.
   Cache_setL2Mode()-Will setup the L2 CACHE control register.
   Cache_enableCaching()-Will setup the MAR registers.

Release Note

devid type declarations are inconsistent in dev.h
 

Release Note

This problem has been fixed.

Release Note

In the 'PRD Function Manager' properties window, The 'Microseconds/Tick'
field is just an information filed. So this field should be greyed out.
 
Also if 'Use CLK Manager to Drive PRD' is checked, the
'Microseconds/Tick'
should reflect the value as set by the CLK Manager properties.
 
If 'Use CLK Manager to drive PRD' is unchecked, the 'Microseconds/Tick'
filed should not reflect the values set by CLK Manager and should
refelct some value like zero.

Release Note

The config tool does not place object declaration for LCK module
objects into the xxxcfg.h file.

Release Note

When configuration file is used to configure the CSL objects and not
bound to any peripherals, the generated .c file has an empty
CSL_cfgInit
function. In this case linker error is generated when the "-pm -op2"
options are used. With these compiler options, the empty CSL_cfgInit
function to be eliminated and results in an undefined symbol error.
 
To avoid this problem, the gconf generated C file should declare the
function to be externally called, then the function would not be
eliminated by the optimizer. The call would be:
 
    #pragma FUNC_EXT_CALLED(CSL_cfgInit)
 

Workaround

Build the project to get a link error. Now insert the line
 
    #pragma FUNC_EXT_CALLED(CSL_cfgInit)
 
in the generated .c file and perform an incremental build. This will
help to work around the problem.
 
******Another workaround :-
 
Conditionally reference the CSL_cfgInit() somewhere, e.g. in main, but
prevent the compiler from optimizing the condition by using the
volatile keyword:
void main()
{
        volatile unsigned char neverTrue = 0;
        if( neverTrue )
        {
                CSL_cfgInit();
        }
        ...
}

Release Note

When using "-fg" compiler option (which uses C++ conventions for .c
files), the #pragma in the cfg_c.c file causes errors.
 
Work-around is to use 'file-specific' options for this one file
(remove the -fg).

Workaround

Use 'file-specific' options for this one file
(remove the -fg).
 
 

Release Note

"DSP type" field in the global settings of Dsk6713 seed is represented
as 6711, but it should be 671x.
 

Release Note

Selecting the mask "ALL" in the gconf for a HWI shows the mask as
0xFFFF which includes the mask for Reset and NMI.

Release Note

The CPU load graph utility sometimes freezes after the CPU load has
overrun.

Release Note

The CCS save workspace does not save the Message Log Options.
The Option Maximum number to display shows 0 even when the workspace is
saved with 2048.

Release Note

Some cdb files do not update correctly when moving to new version of
CCS.  Assembling the generated configuration assembly file will
generate an error if IDL objects are not part of the cpu load
calibration.
 
The workaround is for the user to open the updated cdb file
"check" and then "uncheck" the
"Include in CPU load calibration" check box.
 

Workaround

The workaround is for the user to open the updated cdb file
"check" and then "uncheck" the
"Include in CPU load calibration" check box.

Release Note

When upgrading from 2.1x to 2.20 the cache configuration in the global
properties of  the cdb file (62x/671x) may be grayed out after
upgrading, if "other" is selected as the CSL library file.

Release Note

Setting breakpoints inside assembly files in prebuilt projects does not
work.  The project must be rebuilt first.

Workaround

Build the project then set the breakpoint.

Release Note

If the CCS installation path contains white spaces e.g. C:\Program
Files\ti, gmake for examples breaks.

Workaround

Release Note

The IDRAM object in the attached .cdb can not be deleted.
It has dependencies to the HST host channel manager, Memory
section manager which can easily removed but the dependencies
to RTA_fromHost and RTA_toHost cannot be removed.

Release Note

DSP/BIOS generated *cfg_c.c cannot be used with C++ examples due to
badly performed pragma. This is because the #pragma statements in the
*cfg_c.c cause incorrect compilation when interpreting the file as a
C++ file.
 

Workaround

Workaround for this would be to replace the line
 
#pragma CODE_SECTION(CSL_cfgInit,".text:CSL_cfgInit") in *cfg_c.c with
the following lines and perform an incremental build.
 
#ifdef __cplusplus
#pragma CODE_SECTION(".text:CSL_cfgInit")
#else
#pragma CODE_SECTION(CSL_cfgInit,".text:CSL_cfgInit")
#endif
 
 

Release Note

Gconf shows under the Advanced tab of attached .cdb a CTL reg of 0x280;
the Clock Control tab shows "Two Clock cycles", This is wrong. The CTL
value should be 0x290

Release Note

For Timer Object, cdbcmp creates the following output for that:
prog.module("TIMER").instance("timerCfg0").timerCtlPwid = "Two clock
cycles"
 
The value has a typo; it should read "Two Clock cycles" (capital 'C' in
Clock) instead.

Release Note

DSP goes to reset vector when SIO_start is called from main. When an
SIO_startup is called after a dynamic SIO_create in main, the programs
tries to access an invalid memory and goes to the reset vector.

Release Note

The callback version of DIO may fail if the driver code returns
IOM_COMPLETED from its submit function.  In the DIO_cbIssue function,
if IOM_COMPLETED is return from the driver submit function, the frame
is placed into the fromDevice QUE and then the DIO_cbCallback function
is called and it also puts the same frame into the fromDevice QUE a
second time.  No frame should be placed into fromDevice QUE twice.
The DDK provides the source code for DIO driver.

To fix the problem, in DIO_cbIssue() remove the line:

QUE_put(device->fromdevice, frame);

Workaround

The only possible workaround at this time is to take the dio_cb.c
file and add it into your project build and remove the QUE_put call
in DIO_cbIssue()

Release Note

When a HWI function posts a SWI object having a priority of 1, that SWI
object won't run after the HWI function is processed if the HWI
function was invoked from inside a SEM_pend() call that is blocking.
 
The SWI will be scheduled but won't run immediately after the HWI
object processing, and instead it will run when the SWI scheduler is
invoked outside the SEM_pend() API.

Workaround

Avoid using priority 1 for SWI objects.

Release Note

Inside a kernel if an interrupt occurs and posts a semaphore, the task
waiting on the semaphore may not immediately wake up.

Release Note

 
There is a bug in KNL_enter/KNL_exit that is exposed when using the non-
instrumented BIOS libraries.  The bug happens only when STS-for-SWIs is
enabled in TRC_R_mask.
 
The bug is reached when a TSK that was previously preempted is restored
by virtue of a TSK blocking because of a kernel call.
 

Workaround

 
When using the non-instrumented DSP/BIOS libraries (RTA disabled),
ensure that the TRC mask is 0.  This is achieved by either:
    - disabling all TRC events in the configuration tool;
    - setting TRC_R_mask = 0 in main().
 

Release Note

If the GEL OnPreFileLoaded() handler in the start-up GEL file contains
a GEL_Reset() function call and a BIOS plug-in is being used, CCStudio
may not operate correctly when a .out file is reloaded.

If a target reset is required before loading a .out file, it can be
performed using the Debug->Reset CPU menu item. Alternatively, the BIOS
plug-ins can be closed before reloading the .out file.

Release Note

Calling config.hasReportedError in a Tconf Script causes the
Rhino Debugger to crash -- e.g. when stepping over the line
 
if (config.hasReportedError == false)
 
The error occurs only within the debugger.
 
However, the config.hasReportedError call still performs the correct
function when running Tconf outside of the Rhino debugger shell without
the -g flag.
 
So for someone who really wishes to step over this code, simply comment
out the above if statement as a workaround. One can then step over the
line
 
prog.gen();
 
to generate the cdb file.
 

Release Note

Usage of cdbcmp without the -a switch can lead to creation of
incorrect
DSP/BIOS configurations. Worse, the configuration is valid...its just
not what the user intended.
 
Many SFSs, FAEs are now encouraging customers to do the following
procedure :-
a) take your project's cdb
b) run 'cdbcmp app.cdb > appcfg.tcf'
c) edit/clean-up appcfg.tcf
d) run 'tconf appcfg.tcf' to produce the new cdb.
Their goal is to produce a new cdb that's exactly the same as what
they
started with except having gone thru (a) -> (d) they've not had to do
a
potentially error-prone cdbupdate.
 
Hence, quality/correct cdbcmp output becomes a critical need.
 
In this case, customer's original *cfg_c.c file had the line :-
void CSL_cfgInit()
{
   ...
   EMIF_config(&vs110mem);
   ...
}
 
However after doing...
- cdbcmp vs130_62xx_220.cdb > vs130_62xx_220cfg.tcf
- minor mods to script to make it build
- tconf vs130_62xx_220cfg.tcf
...this line was *not* in the generated *cfg_c.c file.
 
The reason is that EMIF config required more than just creation of an
instance, it also required pre-initialization as a cdb logic
precondition to generating the EMIF_config() call. ie
// value changed:
//    old HEMIF.Enable Pre-Initialization = 0
//    new HEMIF.Enable Pre-Initialization = 1
prog.module("HEMIF").emifEnablePreInit = 1;
 
// value changed:
//    old HEMIF. Pre-Initialize with = prog.get("EMIF_NOTHING")
//    new HEMIF. Pre-Initialize with = prog.get("vs110mem")
prog.module("HEMIF").emifPreInit = prog.get("vs110mem");
 
If instead of doing ...
cdbcmp vs130_62xx_220.cdb > vs130_62xx_220cfg.tcf
u do ...
cdbcmp -a vs130_62xx_220.cdb > vs130_62xx_220cfg.tcf
...then it will generate the complete differences.
 
The -a flag compares all objects and fields including those
that may be unconfigured or hidden. Without this flag cdbcmp will
ignore any mods to properties of modules which have no instances or
hidden objects.
 
Several points of note for BIOS Eng in conclusion are :-
 
1. cdbcmp without the -a flag may give incorrect results.
SDSsq20890 - "cdbcmp does not pick up PRD 'Use CLK Manager to drive
PRD' changes". Although this was fixed, the general problem of cdb
changes to no-instance modules remains.
 
2. cdbcmp -a should perhaps be the encouraged default. However it
produces a high signal to noise ratio!
 
3. I would like to see the set of switches available on the command
line (e.g. -a), fully documented in spru007.pdf -- the tconf manual.
Unfortunately I had to look in internal xdc tree sources.
 
4. Usage of cdbcmp and its output in...
(a) regeneration of a new cdb file
(b) movement to next gen build environments
...should be a key test case for BIOS Eng. Efforts should be made to
ensure that cdbcmp output is semi-clean and largely works out of the
box with a straight call to 'tconf mycfg.tcf'
 
5. A good test case for #4 is the attached cdb. It is a highly
populated cdb -- if cdbcmp can produce output that automatically works
when calling tconf on the resulting cdbcmp output tcf then we're in
good shape.
 

Release Note

Some buttons on the toolbar appear to be active when they are not.
After halting a program, the Halt icon appears to still be active
because it is being framed as any other active icon.
 

Release Note

Columns in the Profiler window can no longer be rearranged. Under
Profiler->Viewing Profile Data->To Read the Proile Data, the second
sentence of the second paragraph reads: "To change the order in which
the tabs appear within the Profile View, click-and-drag the desired
tab
to a new location."
 
This feature is not implemented in the current release.
 
 

Release Note

Opening Help a second time can cause a "Help File Not Found" message.

Release Note

The option "Do Not Scan Dependencies When Loading"  basically checks
for dependencies when opening(loading) a project(.pjt) in CCS. For
example a project has been created using a plugin and if it has more
than 1000+ files included in that project then it will take too much
to
open the project. This can be avoided by checking the option "Do Not
Scan Dependencies When Loading".

Release Note

Compiling maxminmath available under
tutorial/sim64xx/maxminmath can cause a compilation error
Build error: cannot run c:\ti\c6000\cgtools\bin\cl6x This happens if
CCS is not installed in the deafult path ie c:\ti.

Workaround

Modify filespecific option for averageValue.c. In the general tab
please modify the build command which shows now
c:\ti\c6000\cgtools\cl6x  to the correct install path of CCS. Or you
can directly call cl6x.

Release Note

The TMS470R1x Assembly Language Tools User's Guide (SPNU118), Appendix
A, contains a description of the Section Header Flags (bytes 40 through
43) in a COFF file, but the description is not complete. Out of the 32
bits of the section header flag, bits 0-7 specify the section type, but
there is no explanation of what bits 8-11 mean.
 
Bits 8-11 actually specify the alignment of the section, as a power of
2.

Release Note

As per page 7-32 of SPRU 187G, the TMS320C6000 Optimizing Compiler
User's Guide,the line:
.bss: fill = 0x00;
 
in the SECTIONS section.
 
Should be .bss fill = 0x00

Release Note

Example on page 8-7 of the CPU & Instruction guide is incorrect.  The
CPU & Instruction guide says that this type of branch needs to have a
label as the 'target'.  So, the bug is actually in the example
itself.

Release Note

Table A–5. Section Header Flags
WRONG:   STYP_PASS 0x2000 Section should pass these unchanged
CORRECT: STYP_PASS 0x200 Section should pass these unchanged
 
WRONG: The flags listed in Table A–5 can be combined; for example, if
the flag’s word is set to 024h, both STYP_GROUP and STYP_TEXT are set.
 
CORRECT: STYP_GROUP  0x0000 0004

Release Note

Add bullet to section of C6x Optimizing C Compiler User's Guide
describing legal and illegal scenarios of using the .if/.endif
assembler directives in and around .(c)proc/.endproc regions.

Release Note

sym directive explantion wrong for the value and size field

Release Note

 
 
According to page 2-20 of spru198g.pdf, intrinsic _mem2 should be
supported on any 6000 series target.  In addition, the help page USING
INTRINSICS in CCS2.1 makes the same assertion.
 
But actulally intrinsic _mem2 is only supported on c64x target.

Release Note

The CCS Help page titled TMS320C6000 C/C++ Intrinsics and the page
titled TMS320C6400 C/C++ Compiler Intrinsics seem to imply that the
&_amem2  &_amem4 &_amemd8 intrinsics are available on only for c64x
series devices.
 
This is contradicted on the page titled Using Intrinsic that says that
above intrinsics can be using with all c6x devices.
 
The information given in "Using Intrinsic" in CCS Help on &_amem is
correct.

Release Note

SPRU403 documents the .Order property only for CLK instances, but not
for PRD, SWI, TSK, IDL. The output of cdbcmp shows that the PRD, SWI,
TSK and IDL module instances also have a .Order property.
 
This needs to be documented in SPRU403

Release Note

PCI_xfrTest documentation in SPRU403e
 
1. Bookmark for the function is not working.
2. Description ("Tests if the current transaction is complete.") is
not
complete, There is no info on the meaning of values returned.
3. Return Value is not only "0 or 1", but really consists of 2 bits.

Release Note

Documentation problems for SEM_post() and TSK_yield() as they relate
to TSK_disable/enable.
 
(1)  The constraint for SEM_post() API should be updated from:
 
"SEM_post cannot be called from within a TSK_disable/TSK_enable block."
 
to:
 
"If SEM_post() is called from within a TSK_disable/enable block, the
semaphore operation will not be processed until TSK_enable() is
called."
 
(2) The following constraint should be removed from the TSK_yield API:
 
"TSK_yield should not be called from within a TSK_disable/TSK_enable
block."
 
 

Release Note

The Documentation SPRU187 mentions the C64x specific registers DIER
(Delay interrupt enable register), EM (Exception mask register), and ER
(Exception register), however they aren't yet fully documented in
SPRU189e, CPU and Instruction Set Reference Guide' or otherwise.

Release Note

The user guides'links in online help of ATK in "Related Documentation"
section may not work sometimes. In such cases, the user guides needs
to
be invoked from the  \docs\pdf  directory.

Release Note

The online RTDX documentation incorrectly lists a set of libraries
as "extended addressing versions".  The C6x does not support "extended
addressing".  The "rtdxe.lib" listed is actually a (JTAG) RTDX big-
endian target library.

Release Note

The RTDX online help reference to RTDX_interrupt_mask is missing from
the "RTDX->Interface->User Interface" page.  It can be located
under "RTDX->FAQ->What is RTDX_interrupt_mask?"
 
Further, the description for this variable is incorrect.  It should
state:
 
The variable RTDX_interrupt_mask is a user specified bit-mask which
indicates which interrupts are to be disabled within RTDX critical
sections.  The enable bit for the specified interrupts will be
restored upon exiting the critical section.
 
RTDX functions may be called from within one or more Interrupt Service
Routines (ISRs).  Thus, it is possible for an RTDX function to be
interrupted by another.  The mechanism for protecting RTDX global data
structures uses this mask to temporarily disable only those interrupts
that may call an RTDX function so that RTDX data structures may be
manipulated atomically.  This provides interrupt safety for RTDX
critical sections and limits the time that interrupts are globally
disabled to just the critical section entry/exit code itself.
 
Upon entry to an RTDX critical section, this mask is ANDed with the
interrupt control register (see family specifics below) thereby
disabling only those interrupts the user specified.  Upon exiting the
critical section, the interrupt control register is restored to its
prior value.  Failure to specify this mask correctly will NOT protect
RTDX global data structures and may result in undefined RTDX behavior.
 
The safest value for this mask is zero.  However, this is also the
most intrusive value as this results in ALL interrupts being disabled
within RTDX critical sections.  To use the mask effectively, the user
only need clear the bit in the mask which corresponds to those
interrupts which are known to call an RTDX function.  (This includes
any RTDX interrupt handlers.)  Thus, high-priority interrupts which do
not call RTDX interrupts are unaffected.
 
However, if using RTDX within a pre-emptive multi-threaded operating
system (such as DSP/BIOS) the user must mask not only those ISRs which
call RTDX directly, but also those ISRs which may signal other threads
which call RTDX directly.  (See SDSsq25620)
 
RTDX provides example linker control files which specify this mask at
link time in "\ti\examples\\rtdx\shared".  This mask may also
be specified from within the DSP/BIOS Configuration Tool for DSP/BIOS
applications.  For more information on the DSP/BIOS Configuration Tool
see the DSP/BIOS online help.
 
Details per target family:
* On C6x: The 32bit mask is applied to the IER register.
* On C54x: The 16bit mask is applied to the IMR register.
* On C55x: The 32bit mask is applied to the IER0/1 register where the
lower 16 bits of the mask correspond to IER0, and the upper 16 bits
correspond to IER1.
* On ARM: The RTDX_interrupt_mask is unused as RTDX can make no
assumptions about the interrupt control scheme of the user's system.
For this target the user must determine and implement the most
appropriate mechanism based on the file:
   "\ti\examples\\rtdx\shared\rtdx_ints.c".
This file contains 2 functions that are used to bracket RTDX critical
sections.
 

Release Note

When using the C6201 EVM or C54x device (connected using the PCI bus)
and  executing Tools -> AET -> Event Sequencing, an error is
received: "AET could not be initialized properly"

Technical explanation:

Advanced Event Triggering is only available for devices with enabled
silicon capabilities. Only the C6211 and C6711 are enabled with
Advanced Event Triggering and the ARM7 and ARM9 for the TMS470.
Advanced Event Triggering will also be available in future devices,
but will not be retrofitted into the C54x, or C620x or C670x devices.

Regarding documentation, there is a full tutorial on Advanced
Event Triggering, a complete description of Advanced Event Triggering
in the on-line help and a demo.

The AET option should be grayed out.
 

Release Note

Users may get an "Out of Environment Space" error when attempting to
autoconfigure 6X11 DSK.  This can occur during installation or normal
usage. If this error occurs during installation, the user must complete
the installation process.
 
 

Workaround

The following steps will alleviate the problem:
 
1. Go to the Start menu and locate the "Configure 6X11 DSK" shortcut.
2. Right-click and select Properties, then click the Memory tab.
3. Select 4096 from the drop-down menu for the Initial Environment and
   click OK.
4. With the 6X11 DSK powered on and reset, select the "Configure 6X11
   DSK" shortcut.
 
Your DSK should now be configured.
 
 
 

Release Note

While running a standard C6711 DSK with CCS 2.0 that is connected to a
PC with a parallel cable, an error will occur if you pull the parallel
cable.
 
A warning pops up indicating the failure condition.
Clicking on "Exit" leads to a dialog:
"Are you sure you want to exit Code Composer?".
Clicking on "Yes" doesn't exit CCS.

Release Note

To Profile an High-Speed RTDX data link application the user must be
aware of two problems:
 
First, on the C6211 and C6711 devices, the EMU0 pin is used by the
profile counter.  So, the user must switch the EMU pin used by
High-Speed RTDX from EMU0 (the default selection) to EMU1 in order to
avoid conflict.
 
Second - The High-Speed RTDX data link must be started before Profiling
is enabled.  Profiling reserves both EMU0 and EMU1 when it starts even
though it only uses EMU0. This keeps the High-Speed RTDX data link from
working at the same time. To avoid this, after loading code or
performing a restart, start execution of your program and either run to
main or to a software breakpoint and then start profiling.  When you
halt, remember to disable profiling until you have ran again to get
the High-Speed RTDX data link enabled.
 
 

Workaround

The user must switch the HSRTDX pin from the default (EMU0) to EMU1 in
order to avoid HSRTDX/profiling conflict.
 
The user must also start the HSRTDX mechanism before Profiling is
enabled:  After loading code or performing a restart, start execution
of
your program and either run to main or to a software breakpoint and
then
start profiling.  When you halt, remember to disable profiling until
you
have ran again to get HSRTDX enabled.

Release Note

The 6711 DSK does not provide complete descriptions for all errors.
Some of the obscure errors only provide an error number.

Release Note

The ability to program the Event Sequencer and the Event Analysis
Chained Breakpoint or Watchdog Timer jobs while the target is running
is broken.
 

Workaround

If you stop the target, program the sequencer, or the Event
Analysis jobs, and then restart the cpu, the jobs will perform as
expected.
 

Release Note

"Cache coherency issue with software breakpoints on 641x devices"

Due to a cache coherency issue with emulation memory accesses on 641x
class devices, the setting or clearing of software breakpoints may be
unsuccessful.  The conditions that must be met to encounter this
problem are:

+ Software breakpoint location exists in the most recently fetched 64-
byte cache line

+ Software breakpoint location is in the same 64-byte cache line as a
data section

In general the solution to this problem is to ensure that any data
access made is not made to a location that is in the same 64-byte
aligned portion of memory as executable code.   The suggested approach
to solving this problem will vary, depending on the application type.
There are two classes of applications that will be discussed here:

+ Non-BIOS application with user-defined linker command file

+ BIOS application with linker command file auto-generated from
configuration file

References to data memory and executable code memory will be made in
the sections below.  It should be noted that the compiler generated
sections correspond to data or executable code as shown below:

Executable Code Sections:
·  .text, .cio
 
Data Sections:
·  .data, .const, .switch, .bss, .far, .stack, .sysmem, .cinit

 
Non-BIOS Application with user-defined linker command file

In the case of a user-defined linker command file, the following steps
should be taken against the linker command file to ensure that data
memory and executable code memory do not reside in the same 64-byte
aligned portion of memory:

a) Never interleave data and executable code sections.  Examine any
data and executable code sections that are adjacent to one-another.  If
they do not share a 64-byte aligned portion of memory then step ‘b’ may
be skipped.  If the data section is not a data buffer section then
step ‘c’ may be skipped.

b) If a data section shares a 64-byte aligned portion of memory with an
executable code section, consider using the .align directive to force
64-byte alignment of the section that is allocated to the higher
portion of memory:

.data   > MEMX          ; .data section
.text   > MEMX, align(64)       ; start .text section on a 64-byte
boundary

If multiple code sections are defined then the linker command file
should also ensure that each data section and adjoining executable code
section be analyzed to determine if alignment is necessary.

c) If a data section that precedes an adjoining text section is a data
buffer and the ‘-mhn’ compiler option is used then consider the
following change:

.data   > MEMX                  ; .data section
.text   { .+= n;}> MEMX, align(64)      ; pad by n and ensure 64-byte
alignment

BIOS Applications with linker command file auto-generated from
configuration file

In the case of a linker command file that is automatically generated
based on a configuration file, the following steps should be taken
against the configuration file to ensure that data memory and
executable code memory do not reside in the same 64-byte aligned
portion of memory:

a) Define an ISRAM_TEXT and SDRAM_TEXT section that will be used
exclusively for user and BIOS text sections.

b) Modify the default ISRAM and SDRAM to be used for user and BIOS data
sections only.

c) If memory padding is required, as is the case for the ‘-mhn’
compiler option, then the suggested solution is to leave some memory
space (worst-case: n + 64) between the ISRAM and ISRAM_TEXT sections as
well as the SDRAM and SDRAM_TEXT sections.

Release Note

Reloading the COFF file may not be successful when the profiler is On.
The problem is observed with the following environment
 
CCS 2.20.03 + 6416 TEB rev 1.1+ XDS510 PPplus +  WindowsXP
 
The profiler seems to cause problems resulting in abnormal behavior of
CCS when working with BIOS project having many number of tasks.

Release Note

The C6211 examples provided with CCS2.10 does not work with XDS560 on
Windows XP. The examples worked fine with Parallel Port Connection, but
failed when XDS560 was used. Observerd this behaviour for Volume2 and
Bigtime examples.
 
The problem is specific to C6211DSK + Windows XP + XDS560
 
This is a bug.

Workaround

The problem occurs due to GEL file not being loaded.
Make sure that the '.\cc\gel\DSK6211_6711.gel' file is loaded.
 
The reason that the 6711 is working and the C6211 isn't is because of
the EMIF is being programmed as part of the POST.  The C6711 runs
through the POST at power-up (LEDs flashing).

Release Note

Bit "0" for RESET in the IER register is "read only" and should always
be read as "1".  In CCS, it always reads as "0".
 
- Update
 
The LSB (Bit 0) of the IER (Interrupt Enable Register) is a
READ ONLY bit.  The value of this bit is permanently set to 1
in silicon. But this bit is not available for debugger access.
 
The IER register is effectively a 15-bit register. The value
of the LSB is non-deterministic and is simply set to 0 for
display purposes.
 
Modification of this bit by the user is an action that is NOT
SUPPORTED by silicon.  CCS will allow modification of this bit
but the state of the bit should not be interpreted, because it
is not be a true reflection of the state of the target.
 
In summary the user should simply know the following:
 
- Bit 0 of IER is not accessible to the emulation logic.
- The state of this bit is permanently set to 1 in hardware
- CCS sets the bit to 0.
- Modification of this bit has no effect on the target hardware.
 
 
 
 
 
 

Workaround

Bit 0 of IER register is always tied to 1 on Silicon. Reading IER
register from emulator always displays 0 for bit 0. This is because the
IER register(16bits) is built up from the real 15 register latch bits
from the IER and a NULL bit, bit 0. Emulator software treats the NULL
bit as a space holder and displays it as 0 when read. There is nothing
wrong with the silicon behavior. We decided to document this issue.
This
is something the user can simply ignore it.
 
Update:
- See Release note documented under this DDTS entry
 

Release Note

Stepping over or debugging the project by placing break point on 6416
TEB will lead to RESET vector and starts BIOS from BIOS_init. The
problem may appear on any 64xx processor. Simulator works without any
problem.
The problem was observed with
CCS 2.12/CCS 2.20 + TEB6416 + XDS560 + WindowsNT/Windows2000
 
 

Workaround

At present there is communication problem between XDS560 with 64x
processor. There is no workaround except to use XDS510 for debugging on
6416TEB.
 

Release Note

The root cause was identified to be a problem in the software.
The header file was in error and needed the following correction.
 
In defines.h;
 
    Change from:
      #define REG_FAUCR  50
      #define REG_FMCR   51
      #define REG_FADCR  52
    To:
      #define REG_FADCR  50
      #define REG_FAUCR  51
      #define REG_FMCR   52
 
 
 

Release Note

  CCS may display incorrect data when attempting to directly view the
DXR or DRR registers of the McBSP serial port.

Release Note

The register names listed in the registers window for cache registers
do not match those in SPRU 609/610
 

Release Note

The External memory interface register 'CESEC0B' is mapped to
0x01a00048 instead of 0x01a80048 which is the correct memory address .
 

Release Note

When viewing registers using the CCS view registers drop down menu the
6711, 6712 and 6713 devices show the correct register list. However,
6711C, 6712C and 6713 devices will show 67xx in the CCS title bar.
6711C and 6712C devices will incorrectly show the 6713 register set
when registers are viewed with CCS.  There are no operational defects
and no problems with 6711C and 6712C devices.  The additional registers
should be ignored.
 

Release Note

The problems were related to heterogeneous multiprocessor systems.  The
two identified problems are:
 
1) xdsprobe does not complain when the ccbrd0.dat file does not match
the actual heterogeneous multiprocessor hardware configuration.
 
2)xdsprobe correctly identifies the number of processors in a
heterogeneous multiprocessor system but when it compares the ccbrd0.dat
file to the actual HW configuration, xdsprobe always thinks there is
only a single processor in the system.

Release Note

Profiler on 6416 DSK(si rev 1.1) may not show correct results when
tasks are created dynamically. 6416 Device Simulator works without any
problem.

Release Note

CCS can't be installed or used without administrator privileges.
 
 
 
 
 

Workaround

On Windows NT, 2000, and XP administrator privileges are needed to use
CCS or modify the "Texas Instruments" registry keys to allow "full"
control (read and write) capability.
 
 

Release Note

When installing a full version of CCS 2.2 over top of CCS 2.0 + CCS
2.1
+ XDS560 patch an error appears complaining that it can't find the
jtagdata.dll.

At the end of the installation, cc_setup.exe is executed to place
available driver names into the middle pane of CCSetup.  During this
execution, an error message occurs stating "jtagdata.dll" not found.

Workaround:
1.  Once the error occurs, click OK and allow the installation to
complete.

2.  After a reboot, open CCSetup.

3.  Click File -> Import

4.  From the Import Configuration dialog box, choose Advanced>>

5.  Browse for the file \drivers\import\_init.ccs

6.  Click OK

This will place each supported Board/Simulator Type into the middle
pane of CCSetup.
 
 
 
 

Release Note

Updating to CCS 2.20 for the C6000 through the 2.0 to 2.2 flow via
Update Advisor. After installing the CCS 2.20 patch, the driver for the
C6713 did not work properly.
 
When installing the full version of CCS 2.20, this problem is not seen.

Release Note

This is a Microsoft Windows refresh issue.  In certain instances,
shortcuts will not be removed properly.

Release Note

On Windows 98, the uninstall of the XDS560 software may leave
xds560.pnf in the /windows/inf directory and/or it may leave
xds560.inf
in the /windows/inf/other directory.  These files may be deleted by
hand
if desired.  Leaving the files in place does not impact the integrity
of
the system.
 
 

Release Note

DASM command is not supporteed in the Command Window

Release Note

Certain executables do not contain banner or copyright information.
 

Release Note

The following process causes an AET job to be accidentally removed:
 
1. On the line where the AET job is set, hit the return key. This
creates a blank line.
2. Hit backspace twice. The AET job is removed.
 
The only workaround is to resubmit the AET job.

Release Note

Setting a hardware breakpoint through Tools->Advanced Event Triggering-
>Event Analysis will generate an error dialog when you exit CCS which
does not happen when the beakpoint via right mouse click with no Event
Anlaysis window open.

Workaround

Set the Breakpoint through the right mouse click.

Release Note

When creating a project with the Project Wizard plugin, the .cdb file
created by the Project Wizard contain an older BIOS version number.
The resultant project will not build. However, there is a workaround.

Release Note

When calling the CCS Scripting tool from Excel Macro, the following
warning message appears:
 
"Microsoft Excel wait end of OLE action execution from an other
application"

Release Note

FlashBurn gives the following error when trying to program a Motorola S
Format hex file to C6711 DSK.
 
WRITE ERROR: Attempting to write locations
0x70000000 to 0x70000100
which is outside the logical bounds of Flash Memory:
0 to 0x20000

Release Note

RF 3 previously imported a dsk6711.cdb seed which set the DSP Target
Type to C6711. This prevents you from running code on a DSK6211 since
you cant run floating point code on a fixed pt device.
 
RF 3 now imports a dsk6211.cdb seed which sets Target Type to C6211.
This fixed pt code can run on either DSK6211 or DSK6711. User can
still change up to C6711 for full floating pt support if required.
 

Release Note

When the C++ compiler is enabled, the RFs build with an error for the
file *cfg_c.c file, saying "error: badly formed pragma". This is
because a the syntax of the CODE_SECTION pragma used in the file for
CSL_cfgInit is incompatible with C++. C++ supports a different syntax
where the pragma has to be directly located before the function it
addresses. The workaround here is to right click on *cfg_c.c in the
Generated Files of the project view, select Files Specific Options and
uncheck "Treat C files as C++ files" under Parser of the Compiler Tab.
 

Workaround

The workaround here is to right click on *cfg_c.c in the Generated
Files of the project view, select Files Specific Options and
uncheck "Treat C files as C++ files" under Parser of the Compiler Tab.

Release Note

All C6000 modules and applications are shipped with optimization turn
on (-o2).  The provided project (*.pjt) don't use -o2 flag.  All
modules and applications can be re-built using the provided project to
remove optimization.
 

Release Note

ICC_linearDelete caused an assertion when argument is non-NULL. This is
a bug in the source code. The line
 
    UTL_assert(linearIccHandle == NULL);
 
should really be
 
    UTL_assert(linearIccHandle != NULL);
 
This has now been fixed.

Release Note

All new ddk v1.1 IOM drivers expect version # and interrupt masks
passed in as part of their device parameters structure. Hence when
using the Reference Frameworks, one must make sure that the DEVPARAMS
structure has the correct number of fields to maintain correctness in
the code.

Workaround

There are 2 possible workarounds:
 
1. Pass in NULL as the device params ptr in the udevCodec user-defined
device instance in the cdb file (or in XXXXXXX_udevCodec.tci). This
would ensure that the default (correct) set of parameters gets used by
the driver. This is already being done in RF1.
 
2. Modify the file XXXXXXX_devParams.c (where XXXXXXX is the name of
your target) to reflect the changes in the new param structure, e.g.
for DSK5416:
 
DSK5416_DMA_PCM3002_DevParams DSK5416_DEVPARAMS = {
    DSK5416_DMA_PCM3002_VERSION_1, /* Drivers version uses by
application  (NEW!!!)*/
    DSK5416_DMA_PCM3002_RXDMAID_DEFAULT,  /* Uses DMA channel 4 */
    DSK5416_DMA_PCM3002_TXDMAID_DEFAULT,  /* Uses DMA channel 5 */
    0x1FF,                      /* Control Register 0 on PCM3002
device
*/
    0x1FF,                      /* Control Register 1 on PCM3002
device
*/
    0x0,                        /* Control Register 2 on PCM3002
device
*/
    0x0,                        /* Control Register 3 on PCM3002
device
*/
    DSK5416_DMA_PCM3002_INTR_MASK_DEFAULT, /* Receiver interrupt mask
(NEW!!!)*/
    DSK5416_DMA_PCM3002_INTR_MASK_DEFAULT, /* Transmitter interrupt
mask (NEW!!!)*/
};
 
See the appropriate header file (e.g. dsk5416_dma_pcm3002.h for the
dsk
5416) for the actual structure definition.
 

Release Note

loading a COFF file in CCS v2.20 crashes the CCS.

Release Note

Using Microsoft Visual C++ 6.0, you might get an error when building a
RTDX host client that claims that the following RTDX constants are
undefined:
 
 * Success
 * Failure
 * Warning
 * ENoDataAvailable
 * EEndOfLogFile
 
This is caused by a bug in the Visual C++ 6.0 compiler, which has been
fixed in Service Pack 3 and above. More details about this problem are
described in the Microsoft Knowledge Base under article Q217755.
 
The Microsoft Knowledge Base is located at the following URL:
http://search.support.microsoft.com/kb/c.asp
 
Link to Q217755
http://support.microsoft.com/support/kb/articles/Q217/7/55.ASP?LN=EN-US
 
Link to Visual Studio Service Pack 3 (required to fix problem)
http://msdn.microsoft.com/vstudio/sp/vs6sp3/default.asp
 
 

Release Note

RTDX flowcontrol is not resumed after "Debug/Reset, Restart, Reload",
thus the host client may never recieve any more data from the target.
This is most especially true when using an XDS560 emulator - but is
also possible using XDS510 class emulators (including parallel port).
 
FlowControl is a feature of RTDX is Continuous Mode only, and when the
host requests the target to temporarily suspend sending data to the
host as the host buffers are becoming full.  After the host client has
consumed enough RTDX data from the buffers, the host will then request
the target resume sending target-to-host data.
 
This flowcontrol feature is only applicable to when RTDX is used in
Continuous mode.
 
To work around this issue, the user must exit and re-start
CodeComposer if this occurs, or use Non-Continous mode.
 
(See RTDX online documentation for more info on FlowControl and
Continuous & Non-Continous recording modes.)
 

Workaround

This appears to only happen when using the C6711 DSK, the XDS560
emulator, HSRTDX, and CCS 2.20 beta or when you use the C6711 DSK via
parallel port and CCS 2.12.
 
To workaround this issue, you can use the C6711 DSK, the XDS560, and
CCS 2.20 beta.

Release Note

On the C6000 platform, RTDX data transfer stops if the target
temporarily disables interrupts for longer than 1 second.
 

Workaround

Data flow will resume if the target application is restarted or
reloaded and restarted.
 

Release Note

While in a running RTDX MP processor system, a code load onto one of the
non-running DSPs will corrupt the data received from the running DSPs.
Running DSPs (communicating over RTDX) must be stopped before performing
a code load on any of the target in the MP system, otherwise data may be
corrupted.

Release Note

RTDX does not support more than 255 processors within a multi-
processor configuration.  Attempts to configure more than 255
processors within Code Composer Setup will result in an RTDX
initialization failure when Code Composer is invoked.
 

Release Note

There's a problem with JTAG-RTDX and LOG_printf if used in a
multi-processor setup. The LOG window doesn't update for all processors.

Release Note

RTDX Tutorial examples shipped for CCS 2.20 is not working on 6201EVM

Workaround

Adding a 6201 DSP/BIOS cdb file to the tutorial example works.

Release Note

Setting a breakpoint in the RTDX_Poll ISR can cause RTDX to fail - even
if the breakpoint is then removed.  This means that the RTA plug-ins
will no longer update while the DSP is running.

Release Note

The RTDX_writing does not become NULL and so the control does not come
out of the following while loop in the TSK.
 
while (RTDX_writing != NULL)
TSK_sleep(10);
 
This occurs only after the program is run second time or more. The
first time the target executes fine without any problem.
This occurs at around 200 transfers (RTDX_write).

Release Note

Putting a breakpoint at an instruction next to the multi cycle nop
instruction gives a view that the multi cycle nops are being jumped
over without completion of requisite number of nops. However the
internal functionality will be proper though the view of these steps is
incorrect.

Release Note

When using the C6x0x, C6x1x device simulators and C64xx cache
simulators variables are updated one cycle late in the watch window.
 
 
 
 

Release Note

The C62xx and C64xx fast simulators do not support Memory Maps.  Thus
user cannot set memory maps using the CCS GUI or GEL.  This is not
documented in the C6000 simulator user guide.

Release Note

Writes to config address space in the simulator may not be reflected in
the memory window after the corresponding instruction(viz, store instr)
execution(and halt) because the simulator halts the system as a whole,
potentially with writes pending to corresponding config memory/register
space. This is not the case in hardware because on a halt,
only the CPU is halted and the system clocks keep ticking, which inturn
update the memory/register with writes, if any. Non Config writes
behave the same both on hardware and simulator since simulator provides
the CPU(user) view for these writes.
 

Release Note

The piece of code below does not work on CCS2.2 simulator, although
it works fine on the DSK 6713.
 
 
// CODE //
  Float myFloatValue = -1.2456f;
  Int32 myIntValue;
 
// Setting .L units rounding mode to negative infinity
  FADCR |= 0x06000600;
 
// This should round the value down to -2.
// On the 6713 simulator, the value is rounded up to -1.
  myIntValue = _spint(myFloatValue);
 

Release Note

When the VCP is configured with the following configuration it
returns all zeros for computed data and fails decode check
 
Configurations tested
test case       64x Design Simulator    PG 1.1 SI       PG 1.03 SI
CCS simulator
FL=78, Rate = 3, K = 7, HD,Convergent, R = 78, C = 36
Pass            Pass            Fail            Fail
returned all zeros
FL=78, Rate = 3, K = 7, HD,Convergent, R = 336, C = 36
Pass            Pass            Fail            Fail
 
VCP should return
0x9A03FA6F
0xF126E546
0x000000BB
 

Release Note

C6xxx Simulators have specific problem in handling of RTDX messages
from DSP to Host whose sizes (in bytes) are  NOT multiples of target
word sizes.
 
C6x has 4 bytes to a word, while handling messages like:
RTDX_write(&output_channel, &bytes_to_send, 3);
Simulator changes the word size from 3 to 4 and sends it to the host.
 
 

Release Note

TCP module of C6416 Device Simulator is not fully cycle accurate.
Applications that use TCP with smaller frame lengths result in fewer
cycles on the C6416 Device Simulators.
 

Release Note

C62xx simulators may sometimes generate spurious interrupts, if events
are enabled in simulator analysis plugin.
 
 
 
 

Release Note

The example code in the algFree() description in SPRU360
is missing the setting of memTab[0].base and memTab[0].size attributes.
 
 

Release Note

The algorithm standard Code Generator windows(Instance Creation
parameters and Status Creation Parameters) cannot copy(^C) the text.
But ^X works without any problem.
 
---------------------------------------------------------------------
 
The template code generator has been replaced by Component Wizard tool
in the XDAIS Dev Kit v2.5 for CCS v2.1 or later. This component is
available for download from dspvillage.

Workaround

To copy and paste in another window, first cut the text message then
paste it in the windows wherever it's needed.

Release Note

The template code Generator expects space between * and the function
name in the algorithm method. Otherwise * is treated as a function
name.
        Int * AlgMethod2( Int *pnArg )
generates the correct code. But not the below method
        Int *AlgMethod2( Int *pnArg )

Workaround

Please describe the workaround for this problem.

Release Note

The "G.723 Options" window in the Algorithm Standard demo loses focus
and becomes hidden behind the CCS main window when clicking elsewhere
on the CCS window.
 
 
 

Release Note

When having selected text in an input field of the Tools -> Algorithm
Standard -> Template Code Generator plugin, CCS' Edit menu is grayed
out even though it should be possible to cut/copy/paste/delete text.
 
-----------------------------------------------------------------
 
This tool is depricated and superceded by Component Wizard for
eXpressDSP from Hyperception in XDAIS Developers Kit v2.5. The bug is
however open in CCS v2.2 - Rajesh