Comparison of CORTEX-M3 and ARM7
In March 2005, ARM announced the latest ARMV7 architecture and defined three major series:
the "A" series is designed for cutting-edge virtual memory-based operating systems and user applications. Mainly for the growing running of consumer electronics and wireless products including Linux, Windows CE and Symbian;
The "R" series is for real-time systems. Mainly for systems that need to run real-time operating systems for control applications, including cycling electronics, networking and imaging systems.
the "M" series offers optimizations for microcontrollers and low-cost applications. An embedded application that is sensitive to development costs while increasing performance requirements, such as microcontrollers, bike body control systems and a variety of large appliances.
Compare items |
ARM7 |
Cortex-m3 |
Architecture |
ARMV4T (von Neumann) instruction and data bus are common, bottlenecks occur |
ARMV7-M (Harvard) instruction and data bus separate, no bottleneck |
Instruction Set |
32-bit ARM instruction + 16-bit thumb instruction A state switch is required between two sets of instructions |
THUMB/THUMB-2 instruction set 16-bit and 32-bit instructions can be directly mixed without state switching |
Pipeline |
3-Stage Pipeline if there is a transfer need to refresh the pipeline, heavy losses |
3-Stage Pipeline + branch prediction The pipeline does not need to be refreshed when the transfer occurs, almost no loss |
Performance |
0.95dmips/mhz (ARM mode) |
1.25dmips/mhz |
Power |
0.28mw/mhz |
0.19mw/mhz |
Low power mode |
No |
Built-in sleep mode |
Area |
0.62MM2 (kernel only) |
0.86MM2 (core + peripherals) |
Interrupt |
Normal interrupt IRQ and fast interrupt Fiq too few, many peripherals have to be reused interrupt |
Non-shielded interrupt nmi+1-240 physical interrupts each peripheral can have an exclusive interrupt, high efficiency |
Interrupt delay |
24-42 clock Cycles, slow |
12 clock cycles, as fast as 6 |
Interrupt pressure Stack |
Software manual stacking, long code and low efficiency |
Hardware Auto-stack, no code and high efficiency |
Memory protection |
No |
8-Segment Memory Protection Unit (MPU) |
Kernel Registers |
Registers are divided into multiple groups, complex structure, more than the nuclear area |
Register not grouped (except SP), simple structure |
Working mode |
7 modes of operation, more complex |
Only two types of threading mode and processing mode, simple |
Multiplication Law Directive |
Multi-cycle multiplication instruction, no division instruction |
Single-cycle multiplication instruction, 2-12-period Division Instruction |
-bit operation |
No access peripheral register requires "read-change-write" 3 steps. |
Advanced Bit-band-bit operation technology for direct access to a value of peripheral registers |
System beat Timing |
No |
Built-in System beat timer facilitates OS porting |
The difference between ARM7 and Cortex M3