Messages in this thread | | | Date | Wed, 21 Dec 2011 14:22:33 +0100 | From | Ingo Molnar <> | Subject | Re: [RFC 4/5] x86, perf: implements lwp-perf-integration (rc1) |
| |
* Avi Kivity <avi@redhat.com> wrote:
> On 12/21/2011 02:34 PM, Ingo Molnar wrote: > > > I think it can all be supported in a consistent way (see my > > previous mails) - but the feature as presented today just > > does not look useful enough to me if only supports that > > niche self-monitoring usecase. > > I hate to re-enter this thread, but this "niche use case" is > exactly what LWP is designed for. [...]
It's not the only usecase that it can be used in, and that is what matters to me.
> [...] And once the JVM is adapted to exploit LWP, its use will > dwarf all of the uses of perf put together (except the NMI > watchdog). You're only causing the developers needless pain > by forcing them to fit this red peg into a green hole.
I disagree - i think LWP has been seriously over-sold and seriously under-designed. Anyway, i'm willing to be convinced that it's worth to be merged upstream, if it brings tangible benefits to the usecases i mentioned.
Thanks,
Ingo
| |