[Gems-users] Three questions about transactional memory


Date: Fri, 15 Oct 2010 18:37:09 +0200
From: eakpinar <eakpinar@xxxxxx>
Subject: [Gems-users] Three questions about transactional memory
Hello all

I was looking into implementation of LogTM-SE and was wondering if you
could help me out with a few questions. 

- I run STAMP benchmarks. With 1 processor single threaded runs (1 SMT
thread), I get matches and empty checks for signatures in stats. (Even
with perfect signatures). I cannot account for the matching signatures
checks. With 1 processor runs, there should not be any other non-empty
signature (except for the main running thread's). So there should not be
any matching checks. About empty checks, I suspect execution gets
interrupted by non-tx.nal stuff (like page faults or similar OS
activity) and thus the process gets descheduled. When it gets scheduled
again, it checks again the summary signature (which is empty) and hence
the number of empty checks.
- I am trying to understand how non-tx.nal accesses are handled during
transactions. For instance, what happens when there is a page fault
during a transaction? Is conflict detection (and thus signature
checking) carried out? Or are non-tx.nal accesses allowed to happen
without any signature checking? (Thus stability and correctness are
maintained by the coherence protocol)
- Commits are listed in stats with XIDs (transaction IDs I assume).
Regardless of the number of processors, there always seem to be 5 XIDs.
Any
idea how XIDs are assigned?

I'd appreciate any response. Thanks in advance.

Note : I am using gems-2.1.1 with simics-3.0.31

Ege

WARNING / LEGAL TEXT: This message is intended only for the use of the
individual or entity to which it is addressed and may contain
information which is privileged, confidential, proprietary, or exempt
from disclosure under applicable law. If you are not the intended
recipient or the person responsible for delivering the message to the
intended recipient, you are strictly prohibited from disclosing,
distributing, copying, or in any way using this message. If you have
received this communication in error, please notify the sender and
destroy and delete any copies you may have received.

http://www.bsc.es/disclaimer.htm
[← Prev in Thread] Current Thread [Next in Thread→]
  • [Gems-users] Three questions about transactional memory, eakpinar <=