TM4C123GH6PM: what is tm4c123 microcontroller maximum speed at GPIO pins

Danny F said:

simply flipping a pin takes at least two instructions. or 40Mhz on a 80Mhz mcu. that's max you can get. anything less than that is possible.

There are limitations beyond instruction clock. (Loop overhead, instructions per cycle, access times etc...) Bruno measure the fastest he could achieve at 20 MHz

https://e2e.ti.com/support/microcontrollers/tiva_arm/f/908/p/506050/1840865#1840865

PWM should be capable of faster. Whether that's useful is another question. Certainly if all you are doing is making a square wave it would be a waste of resources to use the processor over a PWM.

Bitbanging an output at ~2MHz may be doable (although I'd question the necessity).  Doing it at exactly 2MHz may be a fools errand.

Robert

"Bruno measure the fastest he could achieve at 20 MHz"

obviously he is doing it wrong:

0x000006D0 F8C02080  STR      r2,[r0,#0x80]
    61:                 LED_PORT->RESERVED0[LED] = LED; 
0x000006D4 F8C01080  STR      r1,[r0,#0x80]
    62:                 LED_PORT->RESERVED0[LED] =~LED; 

from good old Keil 3.24. STR is a single-cycle instruction. So if you fill the loop with that, you will be running at 1/2 of main clock.

Might it be (reasonably) expected that poster's application will extend (beyond) a single, endless loop? If so - the brutality & limitation of "such method" will quickly become apparent.     In addition - producing a proper 40MHz output waveform - from this class MCU - proves extremely unlikely!    

Assuming (some) real MCU work is to be achieved - use of one of the MCU's (many) Timers - configured into PWM mode - should achieve a "stable" (hardware based) 2MHz clock output - WHILE enabling multiple other (normal) MCU operations & processes...     Which (effectively) meets "Real-World" usage requirements...

Actually, i want to generate a 12 bit counter which runs from all 0's to all 1's and again back all 0's(endless loop) at precisely 1.024Mhz  rate..For this application also will PWM work ? if yes how do i configure PWM for 12bits counter at 1.024Mhz frequency?

siddharth goyal said:

Actually, i want to generate a 12 bit counter which runs from all 0's to all 1's and again back all 0's(endless loop) at precisely 1.024Mhz  rate.

It is ALWAYS "distressing" - even "Crazy Making" - to receive such "Actually" ...  late in the Requesting Sequence.      Should this detail not have been presented - initially?

Your "newly introduced" (Actuality) bears little resemblance to your thread's Subject/Title - and the new detail was NOWHERE presented (nor hinted) earlier.    Not good!

Bruno Saraiva said:

you land here by parachute on a gusty day

Oh Bruno - its NOT (that) gusty!       And (lowercase poster) Danny DID - "Stick his landing!"

Earlier - one here - noted that this class MCU's output structure is "highly unlikely" to generate (anywhere near) proper - 40MHz output...

(overheard: ... ...  "Is that a flack-jacket you're wearing - or are you (happy) to see me?")

siddharth goyal said:

Actually, i want to generate a 12 bit counter which runs from all 0's to all 1's and again back all 0's(endless loop) at precisely 1.024Mhz  rate..

You want to have a 12 bit parallel output? Or are you just feeding a clock to an external counter? In either case why? What's your ultimate goal?

Precisely at 1.024MHz? How about accuracy and repeatability? Bitbanging this is unlikely to produce the latter especially if you do anything else on the micro.

siddharth goyal said:

For this application also will PWM work ?

Can you produce a MHz signal with PWM? Yes. However, especially given the lack of detail so far whether you can easily meet you specifications is unknowable.

Robert

Danny F said:

obviously he is doing it wrong:

0x000006D0 F8C02080  STR      r2,[r0,#0x80]
    61:                 LED_PORT->RESERVED0[LED] = LED; 
0x000006D4 F8C01080  STR      r1,[r0,#0x80]
    62:                 LED_PORT->RESERVED0[LED] =~LED; 

OK a few things here

• First this is also obviously wrong since it does not include the required looping. Yes you can unroll but that can only go so far
• Second have you tested? As I noted there is more going on than the instructions. Even just in the microprocessor before we get to the peripherals. It's been my experience that in micros like the ARMs toggle speed is limited by other than instruction timing.

Robert

Danny F said:

So you think at some point single cycle instructions would run in single cycles? The mcu will just eat some of them randomly?...

Silly man, where did I ever imply that?

No, there are other factors that slow instructions down. Pipeline effects, cache effects, peripheral timing. I'm sure there are more.

 Robert

Danny F said:

obviously there are, like a pins current drive....

Try rephrasing that

Danny F said:

But do you think those instructions will arrive at the port in the same order as issued by the core?

Instructions don't arrive at the port. I think I understand what you are trying to say. The outputs will arrive at the port in the same order that the instructions execute*. The timing of that may be restricted by things other than the instruction cycle count, extra cycles may be inserted.

Danny F said:

Will the instructions cause the pins to swing, albeit may not sufficiently?

Yes

Danny F said:

Would it be logical to think that if the pins can swing at 40Mhz when driven by a peripheral, the pins have the ability to swing at 40Mhz when driven by the core?

No. It's quite routine to observe otherwise actually. The peripheral will bypass the port structure and connect directly with the pin drivers**. As such the peripheral can have a quite different rate limit than the port.

Robert

* I'll even grant that in this case the instruction will execute in the order written.

** I don't think I've ever come across a device that multiplexes the pin drivers.

Danny F said:

So it ***can*** have. But does it have?

What can have what? Words are your friend, there is no shortage.

Robert