1.2.1. Instruction compression

Microprocessor architectures traditionally had the same width for instructions and data. Therefore, 32-bit architectures had higher performance manipulating 32-bit data and could address a large address space much more efficiently than 16-bit architectures.

16-bit architectures typically had higher code density than 32-bit architectures, and greater than half the performance.

Thumb implements a 16-bit instruction set on a 32-bit architecture to provide:

Copyright © 2001 ARM Limited. All rights reserved.ARM DDI 0234B
Non-Confidential