Most of texts say VLIW, which is lack of hardware to detect hazards and schedule the instructions. But I haven't see any evaluation about how much VLIW can save area(h/w complexity) and power compared to performance-matching superscalar.
When VLIW came out, the advancement of technology node size was slowing down, but not anymore. I want to know it's still valid to say VLIW can be considered to be efficient enough in power/area compared to equivalent superscalar when enough ILP is given.
No comments:
Post a Comment