I'm not able to use RAM_D2 section in STM32H743 as frame buffer

[Svenn Dahlstrom]

Everything is working fine having the framebuffer in RAM_D1 (0x24000000).

This is what I have done trying to move my frame buffer to RAM_D2 (0x30000000):

• I have setup the "TouchGFX_Framebuffer" in the .ld file to RAM_D2

 TouchGFX_Framebuffer :

 {

  *(TouchGFX_Framebuffer)

  *(TouchGFX_Framebuffer*)

  . = ALIGN(4);

 } >RAM_D2

• I'm enabling RAM_D2 clock in the beginning of main()

 __HAL_RCC_D2SRAM1_CLK_ENABLE();

 __HAL_RCC_D2SRAM2_CLK_ENABLE();

 __HAL_RCC_D2SRAM3_CLK_ENABLE();

• Configuring the MPU, setting the RAM_D2 region same as RAM_D1 region

What do I miss out for this to work?

[MM..1]

Hi, you dont write how big is your buffer , but i mean 128k here is low ?

And how your TouchGFX_Framebuffer is used in code?

[Svenn Dahlstrom]

Hi, my framebuffer is (320*240*3) 225KB and the RAM_D2 area is 288KB (using STM32H743ZI) so this should be enough.

From TouchGFX in TouchGFXGeneratedHAL.cpp:

namespace {
    // Use the section "TouchGFX_Framebuffer" in the linker to specify the placement of the buffer
    LOCATION_PRAGMA("TouchGFX_Framebuffer")
    uint32_t frameBuf[(320 * 240 * 3 + 3) / 4] LOCATION_ATTRIBUTE("TouchGFX_Framebuffer");
    static uint16_t lcd_int_active_line;
    static uint16_t lcd_int_porch_line;
}

[MM..1]

Maybe this help you How to properly enable all RAMs on a single core STM32H7?

[Svenn Dahlstrom]

Thanks @MM..1 for helping out.

Yes, I have seen this and tried it but with no luck!

I do believe this is the issue here, so I think I'm not implementing this correctly.

I will do some debugging tomorrow to get more answers.

[Svenn Dahlstrom]

I can confirm that the RAM_D2 is working, I can read and write to the RAM_D2 area.

The frame buffer is setup to use RAM_D2 address @0x30000000.

But the screen is black, nothing is showing up on the screen...

In the linker file, this works:

TouchGFX_Framebuffer :
  {
    . = ALIGN(4);
    *(TouchGFX_Framebuffer)
  } >RAM_D1

This don't work:

TouchGFX_Framebuffer :
  {
    . = ALIGN(4);
    *(TouchGFX_Framebuffer)
  } >RAM_D2

The frame buffer contains the same data in both cases, so it seams like it's not capable to push data to display when it's in the RAM_D2 area.

Does anyone have any clue why it's not working from RAM_D2?

Is it a DMA issue her not working with RAM_D2?

[MM..1]

You dont write how your LCD is connected and ... , but maybe you set new address only for TouchGFX and not hw layer. Your declaration in namespace

uint32_t frameBuf ...?

Try check in debuger how data is in framebuff and what is send to display.

[Svenn Dahlstrom]

I use the STM32H743ZI on a custom board and I use parallel interface LTDC with ChromART (DMA2D), pixel format RGB888.

When I debug I can se that the LTDC frame buffer CFBADD is set to 0x30000000 (RAM_D2) and also the DMA2D memory address MA is also using the RAM_D2 (0x30000000 to 0x30038400, frame buffer size 225K).

The application is running fine (except for there is no picture on the LCD) and also if I have the application to change the screen picture I can see the RAM_D2 (0x30000000 to 0x30038400) content is changing.

Remember the only thing I have to do is change which RAM (RAM_D1 or RAM_D2) in the linker where the frame buffer (TouchGFX_Framebuffer) shall be, and using RAM_D1 it works fine!

[Svenn Dahlstrom · Accepted solution]

This https://community.st.com/s/article/FAQ-DMA-is-not-working-on-STM32H7-devices pointed me to the conclusion that this is not possible.

The DMA2D can copy data from QSPI to RAM_D2 (this I can confirm from debugging) but the LTDC can not take data from RAM_D2 because there is no connection between them. And therefor using RAM_D2 for frame buffer is not possible on STM32H7. Take a look at the memory architecture below.

Is my conclusion correct?

[MM..1]

Yes when isnt RAM_D2 accessible over AXI SRAM , then you have right about LTDC. Otherwise MPU is boss.