Chapter 43: Hardware Registers and Bitfield Packing

[sanko]

In embedded systems, binary file formats, and network protocols, space is at a premium. Instead of using a whole int for a boolean flag, programmers often pack multiple small values into a single byte. For example, a single 8-bit value might store a 3-bit version number, a 1-bit status flag, and a 4-bit command code.

In C, this is handled via Bitfields. In standard Perl, you would have to use complex bitwise math (>>, &, |). In Affix, you can define these fields directly in your Struct and let the JIT engine handle the math for you.

The Recipe

We will build a parser for a hypothetical hardware "Control Register."

use v5.40;
use Affix qw[:all];

# Define the Packed Structure
# C equivalent:
# struct ControlRegister {
#     uint8_t enabled  : 1;
#     uint8_t mode     : 3;
#     uint8_t reserved : 2;
#     uint8_t error    : 1;
#     uint8_t interrupt: 1;
# };
typedef ControlRegister => Struct [
    enabled   => UInt8 | 1,    # 1 bit
    mode      => UInt8 | 3,    # 3 bits
    reserved  => UInt8 | 2,    # 2 bits
    error     => UInt8 | 1,    # 1 bit
    interrupt => UInt8 | 1     # 1 bit (Total = 8 bits / 1 byte)
];

# Simulate hardware input
# Raw byte: 0b10001011
#   Interrupt: 1
#   Error:     0
#   Reserved:  00
#   Mode:      101 (5)
#   Enabled:   1
my $ptr = Affix::malloc( sizeof UChar );
memcpy( $ptr, chr 0b10001011, sizeof UChar );    # Direct byte-level access

# 'Map' the structure onto the raw byte
# We use cast to interpret the scalar memory as our struct
my $reg = cast( $ptr, ControlRegister() );

# Access fields naturally
say 'Register Analysis:';
say ' Enabled:   ' . ( $reg->{enabled} ? 'YES' : 'NO' );
say ' Mode:      ' . $reg->{mode};
say ' Error:     ' . ( $reg->{error}     ? 'YES' : 'NO' );
say ' Interrupt: ' . ( $reg->{interrupt} ? 'YES' : 'NO' );

# Modify fields
say 'Disabling hardware and changing mode...';
$reg->{enabled} = 0;
$reg->{mode}    = 2;

# Inspect the raw byte again
my $new_val = unpack( 'C', $ptr );
say sprintf( 'New raw value: 0b%08b', $new_val );

How It Works

• 1. The | Operator
  When used inside a Struct definition, the | operator specifies the bit-width of a field.

  mode => UInt8 | 3

  This tells Affix: "The field mode is part of an unsigned 8-bit integer, but only occupies 3 bits."

• 2. Automatic Masking and Shifting
  When you write $reg->{mode} = 5, Affix does the following under the hood:

  1. Reads the current byte from memory.
  2. Clears the 3 bits belonging to mode using a bit-mask (0xF1).
  3. Shifts the value 5 to the correct bit-position.
  4. Bitwise-ORs the result and writes the byte back to memory.
     You get all the convenience of a high-level hash with the efficiency of hand-tuned C bitwise math.

• 3. Zero-Copy Views
  By using cast on a string (or a malloc'd pointer), we create a Live View. Modifying the hash keys modifies the underlying memory immediately. This is exactly how you would write a device driver that interacts with memory-mapped hardware registers.

Kitchen Reminders

• Endianness
  Bitfield ordering is platform-dependent. Most compilers (GCC, Clang, MSVC) pack bits from the least significant bit (LSB) to the most significant bit (MSB) on x86_64, but this can vary on ARM or PowerPC. Always check your hardware manual!

• Type Matching
  Ensure the base type (e.g., UInt8) matches the total width of the bits you are packing. If you pack 12 bits into a UInt8, Affix will correctly overflow into the next byte, but your C compiler might do something different.

• Reserved Fields
  Even if you don't use certain bits, you should include them in your struct definition (like the reserved field above) to ensure the subsequent fields are shifted to the correct bit offsets.