Mity CPU Platforms » MityDSP-L138 (ARM9 Based Platforms)

Hi,

I have a custom board with
-- MityDspl-138F module (with FPGA)
-- No Ethernet port
-- UART,USB,SD CARD interface
-- uPP transfer for data process
We are using MMC card for booting linux and to copy application software files to board.

I have developed a C++ CCS 6.1.3 project with compiler version 7.4.16 which uses base of MDK_2014-01-13. I am using \MDK_2014-01-13\sw\DSP\templates\criticallink\platforms\MityDSP-L138\platform.tci. In .map file generated, I can see IRAM section as below

MEMORY CONFIGURATION

name            origin    length      used     unused   attr    fill
----------------------  --------  ---------  --------  --------  ----  --------
  IRAM                  11800000   00020000  00000000  00020000  RWIX

I received data from uPP. To process this data faster I want to use IRAM for some of my buffers. For that I used #pragma DATA_SECTION ("IRAM"); before buffer. For ex.

#pragma DATA_SECTION ("IRAM");
float test_array¹⁶⁰⁰⁰;

After that I can see used IRAM as below.

MEMORY CONFIGURATION

name            origin    length      used     unused   attr    fill
----------------------  --------  ---------  --------  --------  ----  --------
  IRAM                  11800000   00020000  0000fa00  00010600  RWIX

But If I put my buffer in IRAM , the process time reduced and the output get distorted.

following things I observed in .map file is as below.

SECTION ALLOCATION MAP

output                                  attributes/
section   page    origin      length       input sections
--------  ----  ----------  ----------   ----------------
.vers      0    00000000    00000044     COPY SECTION
                  00000000    00000044     HelloworldDSPcfg.obj (.vers)
.prd       0    11800000    00000000     UNINITIALIZED

11800000 PRD_A_TABBEG
11800000 PRD_A_TABEND
00000000 PRD_A_TABLEN

I can not understand the .prd section and its address is same as IRAM i.e 11800000.

My queries are as below

1) Is this the correct way to use IRAM as I defined with pragma?

2) If I want to copy one buffer to another, and one buffer is in DDR2 and another one in IRAM, Is there any specific process I need to take care?

3) copy from/to IRAM/DDR2 is ok? Because IRAM and DDR2 operates on different speed.

4) Anything else is using IRAM as in .map file it is showing .prd section at same memory location 11800000.

If my all buffer is in DDR2, my output is smooth but it takes more time to process data. I am not able to understand the reason of distorted output when I put one of the buffer (which copies data from uPP buffer) in IRAM. Please suggest.

Thanks,
Vivek.