AVR Programming

From Free Pascal wiki
Jump to: navigation, search

The FPC commands for programming AVR microcontrollers such as ATMega or ATTiny are the same commands as always, but because of its character of an embedded system, i.e. no underlaying operating system and direct hardware access, there are several specific topics to know.

For building the FPC compiler see article AVR.


LCL is not available and FCL only partially.


Variables declaration etc. works in the same way as always. As there is no operating system which provides memory allocation, dynamic types like AnsiString could not be used.

Integer variables

The AVR is a 8 bit architecture, the native datatypes are Byte (= uInt8) with a range from 0 to 255 and Shortint (= Int8) ranging from -128 to 127. Calculations and operations with these two datatypes are always the fastest. A difference in comparison to e.g. target win32 is that the Integer type has 16 bit with range between -32768 and 32767 (instead of 32bit for win32). Using the integer type will increase the calculation time in comparison to the native types, but may be reasonable depending on the application. For loops with iteration of less than 256 cycles the iteration variable with type byte is appropriate.

For clarity reasons if not using Byte or Integer, the type names Int8, uInt8, Int16, uInt16, Int32, uInt32, Int64 and uInt64 may be prefered.

Floating point variables

The AVR architecture doesn’t have a floating point unit. Not sure if using soft float is possible by now.

String variables

Only strings with fixed length can be used, as AnsiString and WideString types are dynamically allocated.

Some Examples:

// passing a string as a procedure variable
procedure UARTSend(const AText: ShortString);
// using a variable with fixed length string
var myString: String[5];

Note: The string type ShortString is always 255 chars in length. If you need less characts, use a string with fixed length to not waste the scarce memory. Use ShortString only if you don't know how long your string may be at maximum length.

Accessing hardware configuration registers

All hardware configuration registers are defined in the controler specific configuration file which is located in the pascal sources under … \rtl\embedded\avr. The configuration file is named after the device, e.g. Atmega32.pp for the device Atmega32. Besides the configuration registers it also contains the predefined interrupt names. The file is included automatically for all units, so the register variables and constants are available in each unit. The device and thus the configuration file is defined by the command line parameter –wpdevicename (e.g. –wpatmega32).

The names of the hardware registers correspond to the register names defined in the device datasheet. The hardware registers can be accessed as easy as variables. E.g.

UDR:= 25;

writes the value 25 in the UARTDataRegister. Bit manipulation can effectively be done with shift operations (shl = „shift left“), e.g.

UCSRB:= (1 shl TXEN) or (1 shl RXEN) or (1 shl RXCIE); 
UCSRB:= (1 shl 3) or (1 shl 4) or (1 shl 7);   
UCSRB:= %10011000;                             //Bit 3,4 and 7 are set

All three lines are identical in the generated code: The bits TXEN, RXEN and RXCIE are set in the „USART Control and Status Register B“, thus the UART hardware is enabled for sending and receiving and for interrupts on receiving.

Register name aliases

For hardware abstraction purpose its common to define application specific names for hardware configuration registers, especially for port registers. This can be done by a variable definition using the absolute modifier. In the below example LEDPort is just an other name for the hardware register PortA, such definitions are neutral in memory consumption and performance.

  LEDPort: byte absolute PortA;
  LEDPin = 3;
  LEDPort:= LEDPort or (1 shl LEDPin); //LED=on

Bit arithmetics

Bit arithmetics are FreePascal standard, but as they are especially important for embedded target and quite rarly used in desktop applications they are mentioned here. Available operators for bit manipulation: Not, And, Or, Xor, Shl and Shr. Example:

ByteVal1:= (ByteVal1 and %11110000) or %00001000;

A leading ‚%’ indicates a binary number, a leading ‚$’ a hexadecimal number. For inline assembler ‚0b’ and ‚0x’ are the indicators respectively.


Again, classes are dynamically allocated, and thus are not available. But static classes (keyword object) can be used for structuring and reusing code.


Type TUART = object
  Procedure Init;
  Procedure Send(ByteVal:Byte);




When using hardware interrupts one has to configure and enable a interrupt source, enable interrupts globally and define an interrupt service routine (ISR) which is called by the CPU when the interrupt occurs. The configuration is done via the hardware registers. For globally enabling interrupts the assembler command SEI is available, it can be easily used with inline assembler:

asm SEI end;

Globally disabling interrupts is done with the assembler command CLI:

asm CLI end;

For definition of the ISRs, predefined names have to be used. Depending on the device different interrupts are available. The proper names can be found in the device specific unit (e.g. „Atmega32.pp“) and have to be used in the Alias modifier. A working ISR looks as follows (example: External Interrupt 0):

procedure ExternalInterrupt0_ISR; Alias: 'INT0_ISR'; Interrupt; Public;
  //Do some stuff…

The procedure name itself can be chosen randomly, important is the name behind Alias. The Alias modifier forces the compiler to use the name INT0_ISR for the procedure without name mangling. If this name fits to a predifined interrupt name, than the procedure will be automatically registered in the interrupt vector table. That the automatism is possible the procedure has to be globally visible, which is realised by the Public modifier. But this is only needed if the procedure is defined in a unit. Note: There is no procedure prototype to be defined in the INTERFACE section of the unit, but only the bare procedure in the IMPLEMENTATION section. If the procedure is placed in the program section instead of inside a unit, then the Public modifier is omitted. The Interrupt modifier makes sure that all registers and the status register are saved and recovered by the ISR and that interrupts are further enabled. Omitting the Interrupt modifier will result in malfunction and crashes of the program.

Inline assembler

In some cases performance can be highly increased by a few assembler statements. The inline assembler provides a very convenient way of mixing FreePascal and assembler. The ASM statement starts an assembler block which is finished with an END statement.

  // do some assembly

In the assembler block modified registers have to be published to the compiler in the end statement, so that the compiler can take care of pushing or avoiding registers. The resulting code is more effective than pushing registers manually inside the asm section. Modified but unpublished registers may leed to malfunction.

  ldi r20,35        // r20 modified
  ldi r21,12        // r21 modified
end['r20','r21'];   // publish modified registers to compiler

Constants and variables can be accessed within the inline assembler. The variables have to be distinguished between local variables (stack allocated), which are loaded and stored with the instructions LDD and STD (limited to 64 byte offset), and global variables (statically allocated), which are loaded and stored with the instructions LDS and STS. Variables passed by procedures are allocated on the stack and thus are handled like local variables. Its always usefull to check the assembler listing (.s-file extension).

When using jumps in the assembly then labels are used. They have to be declared in the label section of the procedure. Allowing this, the compiler directive {$goto on} is mandatory.

Note: The peripherial configuration registers defined in the controler specific file (e.g. ‚atmega32.pp’) are variables and not constants. Thus using I/O-instructions like in, out, sbis (and so on) with the register names doesn’t work. Also the addresses are shifted by 0x20, as the general registers are the bottom of the address space. Compare with the memory map in the controler datasheet. The bit definitions of the perpherial configuration registers nevertheless can be used in the inline assembler.

In the following an example for a delay function with inline assembler, which pauses the program for a certain time making the CPU busy.

unit Wait;     //wait.pas
{$mode objfpc}{$H+}
{$goto on}     //Important for using the label directive, the label is needed in the assembler section
  f_CPU = 4000000; //Hertz
Procedure Wait10m(Time: Byte);
// Waitingtime = Time * 10 Milliseconds
Procedure Wait10m(Time: Byte);
  Faktor = 15 * f_CPU div 1000000;
label                 // Labels, here for the loop, has to be declared explicitly
  outer, inner1, inner2;
  tmpByte: byte;      // In Inline assembler local variables are accesed by the instructions LDD and STD, global variables are accessed by LDS and STS.
  asm                 // asm states inline assembly block until the next END statement
    ldd r20,Time      // Variables can be accessed, here a local variable
      ldi r21, Faktor
      inner1:         // 640*Faktor= 640*15*1 = 9600 cycles/1MHz
        ldi r22,128
        inner2:       // 5*128 = 640 cycles
          nop         // 1 cycle
          nop         // 1 cycle
        dec r22       // 1 cycle
        brne inner2   // 2 cycles in case of branch //one loop in sum 5 cycles
      dec r21
      brne inner1 	
    dec r20
    brne outer
  end['r20','r21','r22']; //Used registers to be published to compiler
end;  //procedure
end.  //unit