Motorola MPC533 Reference Manual page 1034

Class-Based Compression Model Main Principles
G.2 Class-Based Compression Model Main Principles
The operational model used by the is explained in the sections below.
G.2.1 Compression Model Features
• Implemented for MPC53x architecture
• Up to 50% instruction code size reduction
• No need for address translation tables
• No changes in the CPU architecture
• A compressor tool performs compression off-line using instruction class-based
algorithms optimized for the MPC53x instruction set
• Decompression is done at run-time by special hardware
• Optimized for cache-less systems:
— Highly effective in system solutions for a low-cache hit ratio environment and
for systems with fast embedded program memory
— Deterministic program execution
— No performance penalty during sequential program flow execution
— Minimal performance penalty due to change of program flow execution
• Switches between compressed and non-compressed user application sections is
possible. (A compressed subroutine can call a non-compressed one and be called
from non-compressed portions of the user application)
• Adaptive vocabularies, generated for a particular application
• Compressed address space is up to 1 Gbyte
• Branch displacement from its target:
— Conditional branch displacement is up to 4 Kbytes
— Unconditional branch displacement is up to 4 Mbytes
Branch displacement is hardware limited. The compiler can
enlarge the branch scope by creating branch chains.
G.2.2 Model Limitations
No address arithmetic is allowed because the address map changes during compression and
no software tool can identify address arithmetic structures in the code.
G-2
PRELIMINARY—SUBJECT TO CHANGE WITHOUT NOTICE
NOTE
MPC533 Reference Manual MOTOROLA