Mostly missless memory in the Mill CPU: Avoiding the pain of cache misses in a atatically-scheduled architecture
The Mill is a new CPU architecture designed for very high single-thread performance within a very small power envelope. It achieves DSP-like power/performance on general purpose codes, without reprogramming. The Mill is a wide-issue, statically scheduled design with exposed pipeline. High-end Mills can decode, issue, and execute over thirty MIMD operations per cycle, sustained. The pipeline is very short, with a mispredict penalty of only four cycles.
It is well known that exposed-pipe static scheduling yields near-perfect code with minimal power -- except when there is a miss in the cache. In a conventional VLIW, a miss stalls the whole machine, whereas an out-of-order architecture can sometimes find other useful operations to execute while waiting on the memory hierarchy. The Mill uses a novel load instruction that tolerates load misses as well as hardware out-of-order approaches can do, while avoiding the need for expensive load buffers and completely avoiding false aliasing. In addition, store misses are impossible on a Mill, and a large fraction of the memory traffic of a conventional processor can be omitted entirely.
The presentation will cover these and other technical aspects of the memory hierarchy in the Mill design. Videos and other material about other aspects of the Mill can be found at ootbcomp.com/docs.
Room B03
Speaker: Ivan Godard, Out-of-the-Box Computing
Wednesday, 10/16/13
Contact:
Website: Click to VisitCost:
FreeSave this Event:
iCalendarGoogle Calendar
Yahoo! Calendar
Windows Live Calendar
Gates Computer Science Building
NEC Auditorium (B3)
Stanford, CA 94305
Website: Click to Visit
