Motorola MPC533 Reference Manual page 1035

G.2.3 Instruction Class-Based Compression Algorithm
The code compression algorithm is based on creating optimal vocabularies of frequently
appearing RCPU RISC instructions or instruction halves and replacing these instructions
with pointers to the vocabularies. The system contains several sets of vocabularies for
different groups of instructions. These groups are referred to as classes.
Every instruction belongs to exactly one class. Compression of the instructions in a class
may be in one of the following modes. Refer to Figure G-1.
1. Compression of the whole instruction into one vocabulary pointer
2. Compression of each half of the instruction into a different vocabulary
3. Compression of one of the instruction's halves into a vocabulary pointer and bypass
of the other half. A bypassed field is one for which non-compressed data (16-bit
halfword or 32-bit word) is placed in the compressed code. After compression is
defined, the non-compressed data field is defined in the class.
4. Bypass of the whole instruction. No compression is permitted.
Uncompressed Instruction
1.
2.
3.
4.
Legend
Figure G-1. Instruction Compression Alternatives
A 4-bit class identifier is added to the beginning of each compressed instruction to enable
class identification during decompression. Compressed and bypass field lengths may vary.
(A fully bypassed instruction, including its 4-bit class identifier, is 36 bits.)
MOTOROLA
PRELIMINARY—SUBJECT TO CHANGE WITHOUT NOTICE
Uncompressed or Bypassed Code
Compressed Code
Class Identifier
Appendix G. MPC534 Compression Features
Class-Based Compression Model Main Principles
Compressed Instruction
1.
2.
3.
OR
4.
G-3