[9] RFR(S): 8068945: Use RBP register as proper frame pointer in JIT compiled code on x86

[Vitaly Davidovich]

>
> IMO, if this thing comes in enabled by default, there should be a shadow
> high-priority bug saying "Figure out if we need to to disable $flag", so
> it does not slip from the release radar.


More like "Turn off $flag before release" :) It seems uncontroversial to
say this can only cause regressions, and not improve performance (and if it
somehow does, then perhaps it's a bug elsewhere).  Moreover, the use case
is for performance profiling/tuning, i.e. not an everyday feature you'd
care about.

On Wed, Apr 22, 2015 at 5:16 PM, Aleksey Shipilev <
aleksey.shipilev at oracle.com> wrote:

> On 04/23/2015 12:09 AM, Vladimir Kozlov wrote:
> > You can always find a test which will regress with any changes :)
>
> Yes, and the job for performance guys is to say which tests are really
> important, and which can be disregarded ;)
>
> > We need to enable the flag by default for now to test new code path.
> > As I said we can switch it off later - it may be not final value which
> > will be shipped.
>
> I fully dig the idea of running the tests with a flag set on.
>
> IMO, if this thing comes in enabled by default, there should be a shadow
> high-priority bug saying "Figure out if we need to to disable $flag", so
> it does not slip from the release radar.
>
> -Aleksey.
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/attachments/20150422/d4b1b60b/attachment.html>