SPI with uDMA cause delay

These times are in ns - according to the documentation and it matches the real word.

This is one SPI transfer.

The next transfers have a delay of about 128us shown in the next picture.

The Logger show 4 delays of bout 30.5us, so all together 122us. I think those 4 delay cause my problem of a fast SPI communication.

Update:

For illustration I set a GPIO (violet) just before SPI_transfer() (BLOCKING) and cleared it right afterwards.

Hi Vincent

Yellow is SPI data

Green is SPI clock.

The 32kHz timebase for the times, also explains why the offset is always  ~30.5us.

But from the violet pin we see, the delay happens after the SPI transmission and I don't know why.

Hi Vincent

I used the instrumented Library, because I discovered the delay using the default one.

I'm using BIOS.LibType_Custom and also with noInstumentation and Logger off - I still have the delay.

Could it be caused by an error I explored in the ROV -> BIOS -> Scan for errors:

ti.sysbios.knl.Semaphore : Basic : (0x20002ed0) : pendElems : Error: Problem scanning pend Queue: JavaException: java.lang.Exception: Target memory read failed at address: 0xbebebebe, length: 8This read is at an INVALID address according to the application's section map. The application is likely either uninitialized or corrupt.

Regards

Armin

Hi Vincent

I did merge my transfers to make it faster, but I have a problem while the slave is sending an interrupt. There I have to read a register and write the same data back to another register of the slave. And with the delayed SPI communication I'm to slow.

This would be my .cfg file.

0677.custom.cfg

To reproduce my problem, I investigated your sample application "lcdSmartRF06EB_CC1310DK_7XD_TI_CC1310F128".

I made a simple modification to send data to the SPI very quickly:

/* Write welcome message to buffer and send it to the LCD */
LCD_bufferClear(lcdHandle, 1);
LCD_bufferPrintString(lcdHandle, 1, "Hello SmartRF06EB!", 0, LCD_PAGE0);
LCD_bufferPrintString(lcdHandle, 1, "Low Priority Task", 0, LCD_PAGE1);
LCD_bufferPrintString(lcdHandle, 1, "Writing to Buffer 1", 0, LCD_PAGE2);
while (1) {
LCD_update(lcdHandle, 1);
}

while (1) {
/* Toggle Board_LED1 */
PIN_setOutputValue(pinHandle, Board_LED1, !PIN_getOutputValue(Board_LED1));

I made a endless while loop over LCD_update(). I looked up the code and LCD_update() mainly does a for loop over LCD_PAGE_COUNT and sends LCD_gotoXY() and LCD_sendData() witch are 2 SPI_transfers.

Here you can see the packages send to the display (yellow = /CS, green = SPI clk, violet=SPI mosi:

Zooming in we see a transfer of 3 bytes for LCD_gotoXY() needs also 87.7us

Would be nice if we could speed up the SPI-DMA driver to be able to communicate to slave devices in "real time".

Best regards Armin