FIR16 aligned to 0x100

Vishal, thanks for your answer.

I am using F2808.

I have tried to use this library and have some problem:
when I use inly RAM memory to debuging - it`s ok.
but when I use FLASH and RAM memory to work - processor freezes (maybe going to ILLEGAL_ISR).

This is my init. code...

/* ===== FIR16 Filter initialization ===== */
/* Define the Delay buffer for the N order filter and place it in "firldb" section */
#pragma DATA_SECTION(dbuffer_DC, "firldb");
int32_t			dbuffer_DC[(FIR_ORDER_DC+3)/2];
const int16_t	COEFF_FIR16_DC[FIR_ORDER_DC+2];
int16_t			revCoeff_DC[FIR_ORDER_DC+2];
FIR16 			FIR16_DC;

#pragma DATA_SECTION(dbuffer_AC, "firldb");
int32_t			dbuffer_AC[(FIR_ORDER_AC+3)/2];
const int16_t	COEFF_FIR16_AC[FIR_ORDER_AC+2];
int16_t			revCoeff_AC[FIR_ORDER_AC+2];
FIR16 			FIR16_AC;

void InitFIR16 (void)
{
	Uint i;

	// DC
	FIR16 FIR16_DC=FIR16_DEFAULTS;

	if((FIR_ORDER_DC & 0x01) == 0)
	{
		revCoeff_DC[FIR_ORDER_REV_DC-1] = 0;
	}
	// Reorder the coefficients
	for(i = 0; i < FIR_ORDER_DC + 2; i = i + 2)
	{
		revCoeff_DC[FIR_ORDER_REV_DC-i-1] = COEFF_FIR16_DC[(i>>1)+(FIR_ORDER_REV_DC/2)];
		revCoeff_DC[FIR_ORDER_REV_DC-i-2] = COEFF_FIR16_DC[(i>>1)];
	}
	// Initialize FIR16 object
	FIR16_DC.order=FIR_ORDER_DC;
	FIR16_DC.dbuffer_ptr=&dbuffer_DC[0];
	FIR16_DC.coeff_ptr=(int32_t *)&revCoeff_DC[0];
	FIR16_DC.init(&FIR16_DC);

	// AC
	FIR16 FIR16_AC=FIR16_DEFAULTS;

	if((FIR_ORDER_AC & 0x01) == 0)
	{
		revCoeff_AC[FIR_ORDER_REV_AC-1] = 0;
	}
	// Reorder the coefficients
	for(i = 0; i < FIR_ORDER_AC + 2; i = i + 2)
	{
		revCoeff_AC[FIR_ORDER_REV_AC-i-1] = COEFF_FIR16_AC[(i>>1)+(FIR_ORDER_REV_AC/2)];
		revCoeff_AC[FIR_ORDER_REV_AC-i-2] = COEFF_FIR16_AC[(i>>1)];
	}
	// Initialize FIR16 object
	FIR16_AC.order=FIR_ORDER_AC;
	FIR16_AC.dbuffer_ptr=&dbuffer_AC[0];
	FIR16_AC.coeff_ptr=(int32_t *)&revCoeff_AC[0];
	FIR16_AC.init(&FIR16_AC);
}
// -------------------------------------- //// -------------------------------------- //

And this is my linker .cmd...

PAGE 0:    /* Program Memory */

RAML0       : origin = 0x008000, length = 0x001900

...

PAGE 1 :   /* Data Memory */

RAML1       : origin = 0x009700, length = 0x000700

...

ramfuncs : LOAD = FLASHD, 
RUN = RAML0, 
LOAD_START(_RamfuncsLoadStart),
LOAD_END(_RamfuncsLoadEnd),
RUN_START(_RamfuncsRunStart),
PAGE = 0

...

firldb align(0x100) > RAML1,     PAGE = 1

...

I changed location PAGE 0/1 and RAM L/M 0/1 for firldb, it also  do not work.

Also in fir16.asm I specified, that FIR16 calculation doing in RAM memory:

.sect "ramfuncs"

_FIR16_calc:

...

but if I remove .sect "ramfuncs" it also do not work.

What I do wrong?

When the execution goes to illegal_isr its because the c28x tries to execute some garbage opcode. Its trying to execute something that is not code. I would single step through each line of the C code and see where the execution goes to illegal_isr.

The FIR code should be able to run out of Flash.

Dear Vishal.

I debug this code in FLASH, and...

... I am going to calculation FIR16_DC, and...

...jumping to here. It is infinite loop (from 0x000724 to 0x00072C)

What is this location? I never specify this addresses.

Hmm, so it looks like XAR7 contains the value 0x722 just before LCR (the call to the FIR routine). So two things to check for
1. check the value of FIR16_DC.calc - it should hold the address of the FIR routine. go to that location in the disassembly window and make sure the FIR code is there
2. MOVL XAR7, @0x2a is acutally loading the value at address (DP*0x40 + 0x2a) i.e. the data page pointer times 64 words plus the offset. This address should correspond to the location of the object element FIR16_DC.calc. So make sure the DP pointer is pointing to the correct page, ifs it not then this instruction is loading the value at the wrong address.

If you find that point #2 is the problem then my follow up question would be - do you have DP load optimization turned on? if yes then does turning it off fix your problem? if it does then please let me know as this might be a compiler issue.

Vishal,

Yes, you was right. I have the wrong address, because I specify FIR16 FIR16_DC=FIR16_DEFAULTS; in initialization function (look at the my second message). When I specify it in header - it all ok.

But then I have a new problem =)

Do I understand correctly: if I want to use several filters (I want to use 5 filters) I should specify separate memory aligned sections? Because if I specify several dbuffer in one section - the filters is working not correctly (the point address of dbuffer is damaged). When I specify separate aligned sections for each filter - it work ok.

I am using Filter order = 50. I look at disassembly and on memory browser - filter is not used all 0x100 memory, but when I decrease this memory - I have a similar error (the point address of dbuffer is damaged).