Subject RE: [IBO] Catastrophic Failures on Hyperthreading machines
Author Alan McDonald
> Hi all,
> Let's start a new thread regarding the issue (multithreaded IBO) I
> had with IBO / Firebird. It appears that the access violation and
> crashes seen in both the firebird API as well as IBO is due to the
> enabling of Hyperthreading.
> We've tried a few HT machines and they all exhibit the same failure -
> - complete crash of application. Once I disabled HT, the failure
> goes away. Tested it on the Centrino variants as well and it passed
> flawlessly.
> So from this I can conclude -- either IBO or Firebird is not ready
> to be ran on HT machines yet (read: IBO/Firebird crashes on HT
> machines. I'm not saying it doesn't utilize HT). Anyone else seeing
> this problem?

this is well known. Search the firebird newsgroup for hyperthreading.

> This is quite sad considering most systems are equiped with HT now.
> Being not able to utilize HT's advantages is one thing, crashing out
> on HT machine is another...
> In fact, whether running multithreaded IBO or not is not an issue.
> On a HT machine, IBO exhibits anomalies as well even if it's running
> on a single thread -- CommitRetaining will sometimes clear the just-
> posted record while in the non-HT mode, it faithfully retained the
> record in the dataset. In fact, in HT mode, there have been numerous
> times when IBO ran into Access Violation just as it's IB_Component
> was set to connected = true.
> The only thing left for me to do now would be to find out which is
> the culprit -- IBO, or Firebird's API (fbclient / gds32).dll. I'd
> really appreciate it if someone would be kind enough to share
> his/her experience in IBO/Firebird regarding its stability in HT
> environment.
> Thank you all in advance.

the server is not equipped for HT - disable it if you want to mount FB on a
such a machine