- Timestamp:
- Sep 8, 2022, 5:11:19 PM (21 months ago)
- Branches:
- ADT, ast-experimental, master, pthread-emulation
- Children:
- 264f6c9
- Parents:
- e4855f6
- File:
-
- 1 edited
Legend:
- Unmodified
- Added
- Removed
-
doc/theses/thierry_delisle_PhD/thesis/text/conclusion.tex
re4855f6 r918e0137 9 9 However, I discovered two significant challenges. 10 10 11 First, modern symmetric multiprocessing CPU have significant performance penalties for communication (often cache related).11 First, modern symmetric multiprocessing CPU have significant performance penalties for communicationm, often cache related. 12 12 A SQMS scheduler (see Section~\ref{sched}), with its \proc-shared ready-queue, has perfect load-balancing but poor affinity resulting in high communication across \procs. 13 13 A MQMS scheduler, with its \proc-specific ready-queues, has poor load-balancing but perfect affinity often resulting in significantly reduced communication. 14 14 However, implementing fairness for an MQMS scheduler is difficult, since fairness requires \procs to be aware of each other's ready-queue progress, \ie communicated knowledge. 15 For balanced workloads with little or no data sharing (embarrassingly parallel), an MQMS scheduler, \eg a state-of-the-art work-stealing scheduler, is near optimal.15 For balanced workloads with little or no data sharing, \ie embarrassingly parallel, an MQMS scheduler is near optimal, \eg a state-of-the-art work-stealing scheduler. 16 16 For these kinds of fair workloads, adding fairness must be low-cost to hide the communication costs needed for global ready-queue progress or performance suffers. 17 17 While I was aware of these realities, I underestimated how little performance margin there is for communication.
Note: See TracChangeset
for help on using the changeset viewer.