Next: Static 2superscalar DLX
Up: Multiple instructions/cycle
Previous: Multiple instructions/cycle
In superscalar,
- Avoids wasted instruction space
- Can be compatible with older implementations
- But actual issue rate can be improved
by compile-time scheduling to create
hazard-free instruction packages
In VLIW,
- Instruction set exposes internal architecture
In both cases,
- Dynamic scheduling (eg Tomasulo) can be used.
- Compiler scheduling helps
Paul H J Kelly
Mon Dec 1 20:07:28 GMT 1997