Hi Robert,

Can you tell me about how you made this comparison? Specifically please share the part number of the OMAP-L138 EVM SOM you used in the test?  You can use the following link to find the part number:

http://www.logicpd.com/pa...rial-number-sticker/

Please also share the details about the test including the command output logs if possible. If you can share enough information to allow us to repeat the Logic portion of the test we can look into your issue.

Thanks

.

Hi Jemiaha,

Thanks for replying, the part number is 1017855 REV A. The same code (see below) is running on both SOMs using ThreadX. The code is being run via an emulator on both SOMs

The count value that is being output on the Mitty DSP - L138B by critical link ~= 666000

The count value that is being output on the OMAP-L138 EVM                         ~= 222000

#define TIME_SINCE(x)        (tx_time_get() - x)

void DebugThread(ULONG thread_input)
{
     DWORD Count = 0;
     ULONG StartTime = tx_time_get();

     ConfigData.DebugLink = true;
     while(1)
     {
          if( DebugCountOn )
          {
               StartTime = tx_time_get();
               while( TIME_SINCE(StartTime) < TX_1_SEC )
               {
                    Count++;
               }

               sprintf( (char*)&FaultMessageBuffer[0], "\r\n Count = %ld \r\n", Count );
               SendTerminalMessage(&FaultMessageBuffer[0]);
               Count = 0;
          }
          else
          {
               tx_thread_sleep( 1000);
          }
     }
}

Regards,

Rob

Hi Robert,
So the SOM you have has 128MB of DDR and 8MB of Flash. I assume that you're running out of the RAM. Does your Mitty DSP-L138F have the same amount of RAM? (I think it has a 256MB option) How are you setting and checking the Clock rate on the Logic SOM? To be honest I can't think of a good reason for why the code should be running slower if you're correctly setting the clock to 300MHz.

Looking at your code I do have a question: You said this was running via an emulator. What do you mean by that? If you're running ThreadX then this code won't be an accurate judge of the code execution speed unless you can ensure that the threads have run for the same amount of time on both SOMs. The Count variable is only incrementing while the DebugThread is running. You could work around that by enclosing the following code in a critical section where preemption and interrupts are disabled:
StartTime = tx_time_get();
while( TIME_SINCE(StartTime) < TX_1_SEC )
{
Count++;
}

You'll need to be sure your ThreadX build can tolerate a critical section one second long. Shouldn't be a problem if the timer that tx_time_get() is based on is free-running. Can you check to see how that effects the results?

Thanks,
Jemiah
.

Hi Jemiah,

The difference in speed looks like it is caused by the Mitty DSP-L138F having a ubl binary file and u-boot binary file in spi flash. When I then restored the EVM spi flash with a ubl binary file and u-boot binary file the code then ran at the same speed. I had been running with just a single application image or via Code Composer Studio V5.3 via an in circuit emulator. I take it that U-boot must be doing some extra initilisation that I will need to do in my single application image. Is this to do with setting up DDR memory.

Regards,

Rob