How do I set input pullups on the 4809 using registers?

BitSeeker · May 4, 2025, 3:05pm

On the 328P, this is easy. Just suppose I wanted to enable pull-ups on 5 of the pins on port d:

PORTC |= 0b00111111;

which would enable pull-ups in PD0-PD4.

I am struggling to accomplish the same on the 4809. Its a different processor, and looking at the datasheet I think I need to write to PORTD.PINnCTRL, but how to I write a 1 to the PULLUPEN bit) of that? I am presuming that I need to specify a port number for 'n', for example PORTD.PIN2CTRL and then somehow write a 1 to bit 3, so would something like this do it:

PORTD.PIN2CTRL = PORTD.PIN2CTRL | (1<<3);

The other problem is that this does not lend itself to iteration so it seems I can't do this in a loop for all pins but would need 8 individual commands for 8 pins?

Surely there must be an easier way?

Yes, it can probably be done with a for loop and pinMode() if you had a pin map, but I can't believe that setting the register directly should be THAT complex?

alto777 · May 4, 2025, 3:21pm

As much research as I will take time to do suggests that this is, indeed, the case.

Here's a macro chatGPT wrote, I cannot test it

#define ENABLE_PORT_PULLUPS(port)                 \
    do {                                          \
        (port).DIRCLR = 0xFF;                     /* Set all pins to input */ \
        for (uint8_t i = 0; i < 8; i++) {         \
            (&(port).PIN0CTRL)[i] |= PORT_PULLUPEN_bm; \
        }                                         \
    } while (0)

So you can

ENABLE_PORT_PULLUPS(PORTA);  // Enables pull-ups on all PA0–PA7

Part of it seems to be moving away from confusing multiple meaings of writing like we are accustomed to doing PORTD = 0xff;

HTH

a7

BitSeeker · May 4, 2025, 3:54pm

I appreciate your reply alto777.

Its an interesting idea to set up a macro, but I am little puzzled by what AI has provided here although it may just be a neat trick:

Suppose I pass it PORTD, then its taking the address of PORTD.PIN0CTRL register which would be a pointer, and then iterating from that pointer to find the subsequent pin registers and changing the PULLUP bit for each. If it actually works, that could be a very neat solution.

I also just found a helpful discussion here:

github.com/SpenceKonde/megaTinyCore

megaavr/extras/Ref_DirectPortManipulation.md

master

# Direct Port Manipulation
When using Arduino functions, like digitalWrite and digitalRead, is too slow, there are two ways to get much faster access to I/O pins.

The first (and more traditional, but more complicated) method is to directly access the registers that control the pins; that is the option that most of this document is concerned with. Within this way, there are two sets of registers, the full service PORTx registers which get accessed with LDS/STS (if you're just accessing one member of the PORTx struct) or LD/LDD, ST/STD if you're using a pointer to the port, which can set all of the options that are "owned" by that port, So read, and/or with constant, write is either 3, 1, 2 clocks or 2, 1, 1 clocks. The other set of registers are the high performance VPORTx registers. They get 4 1-byte registers per port - the direction, output value, input value, and interrupt flags - but these are located in the most favorable part of memory; they work best when you are writing or testing a single bit, known at compile time, within a VPORT register known at compile time.

The second - equally performant when an apple-to-apple comparison is possible - method is the digitalWriteFast()/digitalReadFast()/etc functions provided by the core. These functions are detailed in [the digital I/O reference](Ref_Digital.md). This method is recommended when both are viable as code tends to be more readable and less likely to contain bugs. Fast Digital I/O requires the pin to be compile time known and constant, and prefers if the second argument is as well.

Both of these methods compile to the same machine code:
```c
digitalWriteFast(PIN_PA4,HIGH); /* 1 clock: sbi 1, 4 */

VPORTA.OUT |= 1 << 4; /* Same as above */

/* Similarly */
if (digitalReadFast(PIN_PA4) == HIGH) { /* rendered as sbis 2, 4; rjmp afterfunctionbody; typically. Overhead is just 2 clocks plus the function body itself if it runs, 3 otherwise, and 4 bytes of flash. */
    //...
}

if (VPORTA.IN & (1 << 4)) { /* Same as above */
    //...

This file has been truncated. show original

After reading through that, it would seem that what AI told me below cannot be correct:

That is how it is done with the old 8-bit AVR MCU chips.

alto777 · May 4, 2025, 4:19pm

I'm not a macro wizard, but if you copy the entirety of the definition from the code window it looks like it just loops over all pins of the port you pass in as an argument.

The backslashes mean it's seen as one line, the do { } while (0) is a well-known macro hack so you can use the it like a single statement - you provide the semicolon.

ENABLE_PORT_PULLUPS(PORTA)                ;

The fist part is the expansion of the macro.

Google "C macro do while" for a more competent explanation.

You could do that same thing with a function call, that would be my preference.

And yeah, squinting at your AI's output makes mine look smarter.

Aside from its use of the constant PORT_PULLUPEN_bm instead of what your code does with 1, the method is identical. Set a bit in the pin control register, something the 328p has not.

a7

BitSeeker · May 4, 2025, 4:32pm

Hmm, I tried adding some bits and compiling but its telling me that PORT does not have DIR method, nor an IN method nor an OUT method nor a PINnCTRL method, but obviously the datasheets says it does, so I must be missing something....

/mnt/ntfs/Data/Data/Programming/Arduino/AR488/Testing/Mega4809/AR488/AR488_Layouts.cpp: In function 'void readyGpibDbus()':
/mnt/ntfs/Data/Data/Programming/Arduino/AR488/Testing/Mega4809/AR488/AR488_Layouts.cpp:1294:9: error: 'class PORTDClass' has no member named 'DIR'
   PORTD.DIR &= 0b00000000;
         ^~~
/mnt/ntfs/Data/Data/Programming/Arduino/AR488/Testing/Mega4809/AR488/AR488_Layouts.cpp:1298:14: error: 'class PORTDClass' has no member named 'PIN0CTRL'
       &PORTD.PIN0CTRL)[i] |= PORT_PULLUPEN_bm; \
              ^~~~~~~~
/mnt/ntfs/Data/Data/Programming/Arduino/AR488/Testing/Mega4809/AR488/AR488_Layouts.cpp: In function 'uint8_t readGpibDbus()':
/mnt/ntfs/Data/Data/Programming/Arduino/AR488/Testing/Mega4809/AR488/AR488_Layouts.cpp:1307:17: error: 'class PORTDClass' has no member named 'IN'
   return ~PORTD.IN;
                 ^~
/mnt/ntfs/Data/Data/Programming/Arduino/AR488/Testing/Mega4809/AR488/AR488_Layouts.cpp: In function 'void setGpibDbus(uint8_t)':
/mnt/ntfs/Data/Data/Programming/Arduino/AR488/Testing/Mega4809/AR488/AR488_Layouts.cpp:1315:9: error: 'class PORTDClass' has no member named 'OUT'
   PORTD.OUT = ~db;
         ^~~
exit status 1

Compilation error: 'class PORTDClass' has no member named 'DIR'

BitSeeker · May 4, 2025, 4:50pm

Ok, I think I have found it. Tools-> Registers Emulation-> None.

Yet the ATmega 328 register emulation does not provide PIND or DDRD, just PORTD which is why I got here in the first place. Curious.

PaulRB · May 4, 2025, 6:55pm

Oh, it isn't that complex. And, importantly, it's using Arduino standard functions and it's portable between many different Arduino boards with different chips.

Does the extra performance you get from direct port manipulation really justify moving away from those standards and portability? Only you know for sure, but please try to be honest with yourself about it. ATMEGA4809 will never be as important or long lived as ATMEGA328, they are the old generation and have already been outclassed by MCU based on ARM and that may be in turn replaced by RISC V and so on in ever quicker succession.

Topic		Replies	Views
Mixing PORT mode and bitMode Programming	11	185	June 15, 2026
Port und Pin an Funktion übergeben Deutsch	7	192	October 8, 2025
Nano R4 GPIO switching Nano R4	18	303	October 31, 2025
INPUT_PULLUP documentation is inconsistent Website and Forum	28	476	November 7, 2025
Programming ATtiny48 General Guidance	80	948	April 25, 2025

How do I set input pullups on the 4809 using registers?

Related topics