byte order microblaze

[Frank]

Hi,

I've a small program in bram which has a pointer to sdram.

static volatile BYTE * const p_sdram = (BYTE *) XPAR_SDRAM_CTRL_BASEADDR;

in my code I try to fill the sdram (by the way, it's 32-bits wide databus):

for (i = 0; i < 16; i++) {
*(p_sdram+i) = i;
}

Now comes the problem: when I read back the data, each 2 bytes are
exchanged. Reading back is done with the same byte pointer:

for (i = 1; i <= 16; i++) {
xil_printf("%02X ", (BYTE) *(p_sdram+i-1));
}

The output is:

01 00 03 02 05 04 07 06 09 08 0B 0A 0D 0C 0F 0E

What is wrong?! If I have a byte pointer, the big or little endian story
doesn't matter, does it? I'm posting this here, because I have the feeling
that it has to do with the opb sdram controller? Can somebody explain the
above?

Frank

 

[Vasanth Asokan]

I have this piece of code. I have just used char instead of BYTE. I get
the proper output. I do not see the interleaving that you specify.
If you can give more info about the design you are seeing this problem
with, we can take a look.


=====
int main()
{
static volatile char * const p_sdram = (char *) 0xe000000;
int i;

for (i = 0; i < 16; i++) {
*(p_sdram+i) = i;
}

for (i = 1; i <= 16; i++) {
xil_printf("%02X ", (char) *(p_sdram+i-1));
}

while (1);
}
======

Frank wrote:

Hi,

I've a small program in bram which has a pointer to sdram.

static volatile BYTE * const p_sdram = (BYTE *) XPAR_SDRAM_CTRL_BASEADDR;

in my code I try to fill the sdram (by the way, it's 32-bits wide databus):

for (i = 0; i < 16; i++) {
*(p_sdram+i) = i;
}

Now comes the problem: when I read back the data, each 2 bytes are
exchanged. Reading back is done with the same byte pointer:

for (i = 1; i <= 16; i++) {
xil_printf("%02X ", (BYTE) *(p_sdram+i-1));
}

The output is:

01 00 03 02 05 04 07 06 09 08 0B 0A 0D 0C 0F 0E

What is wrong?! If I have a byte pointer, the big or little endian story
doesn't matter, does it? I'm posting this here, because I have the feeling
that it has to do with the opb sdram controller? Can somebody explain the
above?

Frank

 

[Anita Schreiber]

Frank,

I've also used char in my code when testing the OPB SDRAM controller and not
seen a problem.

There is a bit-ordering difference that is really only important in the address
bus simply because when setting the SDRAM mode register, the address bus
contains the data for that register.

So I reccommend checking the bit-ordering connections. There are some examples
in the design spec.

Thanks!
anita

Vasanth Asokan wrote:

I have this piece of code. I have just used char instead of BYTE. I get
the proper output. I do not see the interleaving that you specify.
If you can give more info about the design you are seeing this problem
with, we can take a look.

=====
int main()
{
static volatile char * const p_sdram = (char *) 0xe000000;
int i;

for (i = 0; i < 16; i++) {
*(p_sdram+i) = i;
}

for (i = 1; i <= 16; i++) {
xil_printf("%02X ", (char) *(p_sdram+i-1));
}

while (1);
}
======

Frank wrote:

Hi,

I've a small program in bram which has a pointer to sdram.

static volatile BYTE * const p_sdram = (BYTE *) XPAR_SDRAM_CTRL_BASEADDR;

in my code I try to fill the sdram (by the way, it's 32-bits wide databus):

for (i = 0; i < 16; i++) {
*(p_sdram+i) = i;
}

Now comes the problem: when I read back the data, each 2 bytes are
exchanged. Reading back is done with the same byte pointer:

for (i = 1; i <= 16; i++) {
xil_printf("%02X ", (BYTE) *(p_sdram+i-1));
}

The output is:

01 00 03 02 05 04 07 06 09 08 0B 0A 0D 0C 0F 0E

What is wrong?! If I have a byte pointer, the big or little endian story
doesn't matter, does it? I'm posting this here, because I have the feeling
that it has to do with the opb sdram controller? Can somebody explain the
above?

Frank

 

[Frank]

I also tested the same code with a pointer to bram, that was working good (I
got the expected output). I am using the Virtex II development board. I will
check the connections.

Thanks,
Frank

"Anita Schreiber" <anitas@xilinx.com> wrote in message
news:3FDA3B4A.E206AF06@xilinx.com...

Frank,

I've also used char in my code when testing the OPB SDRAM controller and
not
seen a problem.

There is a bit-ordering difference that is really only important in the
address
bus simply because when setting the SDRAM mode register, the address bus
contains the data for that register.

So I reccommend checking the bit-ordering connections. There are some
examples
in the design spec.

Thanks!
anita

Vasanth Asokan wrote:

I have this piece of code. I have just used char instead of BYTE. I get
the proper output. I do not see the interleaving that you specify.
If you can give more info about the design you are seeing this problem
with, we can take a look.

=====
int main()
{
static volatile char * const p_sdram = (char *) 0xe000000;
int i;

for (i = 0; i < 16; i++) {
*(p_sdram+i) = i;
}

for (i = 1; i <= 16; i++) {
xil_printf("%02X ", (char) *(p_sdram+i-1));
}

while (1);
}
======

Frank wrote:

Hi,

I've a small program in bram which has a pointer to sdram.

static volatile BYTE * const p_sdram = (BYTE *)
XPAR_SDRAM_CTRL_BASEADDR;

in my code I try to fill the sdram (by the way, it's 32-bits wide
databus):

for (i = 0; i < 16; i++) {
*(p_sdram+i) = i;
}

Now comes the problem: when I read back the data, each 2 bytes are
exchanged. Reading back is done with the same byte pointer:

for (i = 1; i <= 16; i++) {
xil_printf("%02X ", (BYTE) *(p_sdram+i-1));
}

The output is:

01 00 03 02 05 04 07 06 09 08 0B 0A 0D 0C 0F 0E

What is wrong?! If I have a byte pointer, the big or little endian
story
doesn't matter, does it? I'm posting this here, because I have the
feeling
that it has to do with the opb sdram controller? Can somebody explain
the
above?

Frank