TMS570LS1224: APIs from Halcogen spiTransmitData, spiTransmitAndReceiveData taking 10us to execute at 10MHz baud rate for 4 bytes

Hi,

How did get 9.6us? Did you use scope or PMU? 

When is the start point and end point of the 9.6us interval in your measurement? 

You can measure how many CPU cycles the API takes using PMU:

/*Start Cycle Counter */
_pmuStartCounters_(pmuCYCLE_COUNTER);

/* SPI API */

spiTransmitData(...);

or

spiTransmitAndReceiveData(...)

/*Stop Cycle Counter */
_pmuStopCounters_(pmuCYCLE_COUNTER);

/* read the CNNT cycle */
cycles= _pmuGetCycleCount_();

I tried to toggle a PIN before this API is called and after. I will try this using PMU.

Toggling a GIO pin is good enough to measure the time consumed by the SP API.

Can you think of any reasons why it is taking 10us to execute?

Please let me know which section in your code takes 10us to execute?

The else part of the code with out CRC.

I will check it 

Hi Sachin,

I just check the CPU cycles taken by the SPI APIs: 

1. spiTransmitAndReceiveData(spiREG1, &dataconfig1_t, 1, TX_Data_Master, RX_Data_Master);

    This takes 313 CPU cycles to transmit and receive 1 word. 

2. spiTransmitData(spiREG1, &dataconfig1_t, 1, TX_Data_Slave);

    This takes the same amount of CPU cycles.

The following is the code for measurement:

The CPU frequency is 160Mhz, so the two APIs used in your else section take 626 CPU cycles: 626*1000/160=3.9us.

Hi Sachin,

Per today's conf call, this is the example of MibSPI+DMA. This example was implemented on TMS570LC4357 Launchpad. 

TMS570LC4357_MibSPI1&3_DMA.zip