DSP/BIOS Task compile issue

balasub...

the ID in the write task (only one write task) needs to be set to contain the
thread number, not the task number.

R. Williams

---------- Original Message -----------
From: balasubramanian vaidhyanathan
To: William C Bonner , Richard Williams

Cc: c...
Sent: Thu, 25 Sep 2008 07:35:16 -0700 (PDT)
Subject: Re: [c6x] Issue understanding this code

> Hi Richard
>
> Are the three Ids three different tasks. I added three tasks in the
> cdb file and connected this function _write to the properties and
> executed the code still i get only one id.. Can you tell me how to get
> three id printed out?.....
>
> --- On Wed, 9/24/08, Richard Williams wrote:
>
> From: Richard Williams
> Subject: Re: [c6x] Issue understanding this code
> To: i...@yahoo.com, "William C Bonner"
> Cc: c...
> Date: Wednesday, September 24, 2008, 11:32 PM
>
> balasub,,,
>
> I ran the code, I now see why the 'freeQueue' was being initialized
> with
> pointer to messages/
> I.E.
> The QUE_Get() and QUE_Put() only transfer a pointer to a message,
> rather than the actual message content.
>
> So I am wrong in the elimination of the freeQueue.
> Sorry about that.
> BTW:
> The code is also missing the following, probably just after the
> #include statements: void reader(void); // declaration for the reader task
> void writer(Int taskId); // declaration for the writer task
>
> The code is missing the following statements:
> extern SEM_Obj sem;
> extern QUE_Obj msgQueue;
> extern QUE_Obj freeQueue;
> extern LOG_Obj trace;
>
> And each of the writer_Msg variables that I wrote, need to be changed
> to a pointer, etc.
>
> If you want, I can send along a more realistic version.
>
> R. Williams
>
> ---------- Original Message -----------
> From: balasubramanian vaidhyanathan
> To: William C Bonner
> Cc: c...
> Sent: Wed, 24 Sep 2008 15:13:18 -0700 (PDT)
> Subject: [c6x] Issue understanding this code
>
> > Hi
> >
> > Below is the code I yanked out of TI BIOS User guide. Now what i
> > cant understand is this, the writer task has an id. In the document it
> > says there are three writers. So does that mean i have to create three
> > tasks and connect them to _writer and give same priority and run this
> > code so that i can have 3 writers?. I tried running with 2 writers and
> > the code does not behave as told. I ran with only one task and in the
> > reader() task i removed the 2 in the loop and the code exited
> > properly. Please explain.I have highlighted my confusing points in
> bold....
> >
> > regards
> > Bala
> >
> > #include
> > #include
> > #include
> > #include
> > #include
> > #include
> > #include "queuecfg.h"
> > #define NUMMSGS 3 /* number of messages */
> > #define NUMWRITERS 3
> > typedef struct MsgObj {
> > QUE_Elem elem; /* first field for QUE */
> > int id;
> > Char val; /* message value */
> > } MsgObj, *Msg;
> >
> > /*
> > * ======== main =======> > */
> > Void main()
> > {
> > int i;
> > MsgObj *msg;
> > Uns mask;
> > mask=TRC_LOGTSK|TRC_LOGSWI|TRC_STSSWI|TRC_LOGCLK;
> > TRC_enable(TRC_GBLHOST|TRC_GBLTARG|mask);
> > msg=(MsgObj *)MEM_alloc(0,NUMMSGS*sizeof(MsgObj),0);
> > if(msg==MEM_ILLEGAL)
> > {
> > SYS_abort("Memory allocation failed \n");
> > }
> > for(i=0;i > > {
> > QUE_put(&freeQueue,msg);
> > }
> > }
> > /* reader code */
> > void reader()
> > {
> > Msg msg;
> > Int i;
> > for(i=0;i > > {
> > SEM_pend(&sem,SYS_FOREVER);
> > msg=QUE_get(&msgQueue);
> > LOG_printf(&trace,"read '%c' from (%d)
> .",msg->val,msg->id);
> > QUE_put(&freeQueue,msg);
> > }
> > LOG_printf(&trace,"reader done\n");
> > }
> >
> > /* writer code */
> > void writer(Int id)
> > {
> > Msg msg;
> > Int i;
> >
> > for(i=0;i > > {
> > /* Get msg from the free queue*/
> > if(QUE_empty(&freeQueue))
> > {
> >
> > SYS_abort("Empty free queue\n");
> > }
> > msg=QUE_get(&freeQueue);
> > /*fill in value*/
> > msg->id=id;
> > msg->val=(i& 0xf)+'a';
> > LOG_printf(&trace,"(%d) writing
> '%c'...",id,msg->val);
> > QUE_put(&msgQueue,msg);
> > SEM_post(&sem);
> > }
> > LOG_printf(&trace,"Writer is done\n");
> > }
> >
> > --- On Wed, 9/24/08, William C Bonner wrote:
> >
> > From: William C Bonner
> > Subject: Re: [c6x] DSP/BIOS Task compile issue
> > To: "balasubramanian vaidhyanathan"
>
> > Cc: c...
> > Date: Wednesday, September 24, 2008, 11:04 AM
> >
> > When you couldn't get your project to compile, was it giving you
> > errors, or was it simply ignoring your code?
> >
> > I seem to remember something like this from when I was starting with
> > the TI compiler. I am pretty sure I set a flag to cause it to
> > recognize the .cpp extension as a c++ file. The extension used for
> > recognizing a C++ code file varied between compilers for a long time.
> > Many recognized .cc or .C while using .c for standard C code.
> >
> > balasubramanian vaidhyanathan wrote:
> > > I got my trial project to compile... Only because my file name was
> > > .cpp and i changed to .c it compiled.... i find this strange in
> > > DSP/BIOS... Has anyone faced this issue before?
> > >
> > >
> ------- End of Original Message -------
------- End of Original Message ------- Start a New Thread

If the SEM_pend is happening sometime during the main function then the
behavior is expected. Semaphores and locks do not function until after the
main function returns and the DSP/BIOS finishes initializing.

On Thu, Sep 25, 2008 at 10:43 AM, balasubramanian vaidhyanathan <
i...@yahoo.com> wrote:

> Hi all My code is supposed to pend on a semaphore and i checked the
> semaphore count alsoit is zero. but the code still goes beyond SEM_pend when
> i step through. Why would it behave like this?. If the semaphorecount is
> zero then it should pend forever because ihave time out as SYS_FOREVER....
> Have u faced this issue?> regardsBala
>
>
>
>
>
>
--
Gregory Lee Start a New Thread

To add little more to it ,
If application is having only one task with Sem_pend in that and posting is
done from main function , this will cause abnormal behavior , for semaphore
pending and task . If you check the Execution graph details it will
show *"ERROR:
Invalid thread state transition"*

But if you add one more task to above application , i mean empty task with
not even single line of code in that, then Sem_pend will work fine and no
error message also . Sem posting from main function works normally .

This goes to prove that ,Semaphore usage is for task-to-task sync only , ie
only when multiple tasks exists .

- Jai

On Fri, Sep 26, 2008 at 1:56 PM, Gregory Lee wrote:

> If the SEM_pend is happening sometime during the main function then the
> behavior is expected. Semaphores and locks do not function until after the
> main function returns and the DSP/BIOS finishes initializing.
>
> On Thu, Sep 25, 2008 at 10:43 AM, balasubramanian vaidhyanathan <
> i...@yahoo.com> wrote:
>
>> Hi all My code is supposed to pend on a semaphore and i checked the
>> semaphore count alsoit is zero. but the code still goes beyond SEM_pend when
>> i step through. Why would it behave like this?. If the semaphorecount is
>> zero then it should pend forever because ihave time out as SYS_FOREVER....
>> Have u faced this issue?> regardsBala
>>
>>
>>
>>
>>
>>
> --
> Gregory Lee
>
> Start a New Thread