You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@harmony.apache.org by Vladimir Ivanov <iv...@gmail.com> on 2007/03/01 14:34:21 UTC
[drlvm][tesing]run modes for DRLVM tests
Hello everybody,
I think that we also should run DRLVM tests in the default (without
any switches) mode. Issue 3276 was created to track it. Note, this
issue also update default test run modes from "int, jet, opt" to "int,
def, opt".
Thanks, Vladimir
Re: [drlvm][tesing]run modes for DRLVM tests
Posted by Egor Pasko <eg...@gmail.com>.
On the 0x28D day of Apache Harmony Vladimir Ivanov wrote:
> On 3/2/07, Pavel Ozhdikhin <pa...@gmail.com> wrote:
> > >
> > > Yes, I'll update the CC to run tests in 4 modes (int+jet+jit+opt).
> > > Should we add the server mode to CC testing? If yes the exclude lists
> > > for this mode should be added.
> >
> >
> >
> > It would be nice because the server mode is the most powerful in terms of
> > performance and includes more JIT optimizations than any other mode. Let's
> > CC tests the following modes: int, jet, default, opt, server. I think this
> > won't make a single CC run much longer. Will you take care of the initial
> > exclude-list for this mode?
>
> Yes, I prepare update.
> thanks, Vladimir
OK, reasonable, no objections from me
--
Egor Pasko
Re: [drlvm][tesing]run modes for DRLVM tests
Posted by Vladimir Ivanov <iv...@gmail.com>.
On 3/2/07, Pavel Ozhdikhin <pa...@gmail.com> wrote:
> >
> > Yes, I'll update the CC to run tests in 4 modes (int+jet+jit+opt).
> > Should we add the server mode to CC testing? If yes the exclude lists
> > for this mode should be added.
>
>
>
> It would be nice because the server mode is the most powerful in terms of
> performance and includes more JIT optimizations than any other mode. Let's
> CC tests the following modes: int, jet, default, opt, server. I think this
> won't make a single CC run much longer. Will you take care of the initial
> exclude-list for this mode?
Yes, I prepare update.
thanks, Vladimir
>
> Thank you,
> Pavel
>
Re: [drlvm][tesing]run modes for DRLVM tests
Posted by Pavel Ozhdikhin <pa...@gmail.com>.
>
> Yes, I'll update the CC to run tests in 4 modes (int+jet+jit+opt).
> Should we add the server mode to CC testing? If yes the exclude lists
> for this mode should be added.
It would be nice because the server mode is the most powerful in terms of
performance and includes more JIT optimizations than any other mode. Let's
CC tests the following modes: int, jet, default, opt, server. I think this
won't make a single CC run much longer. Will you take care of the initial
exclude-list for this mode?
Thank you,
Pavel
Re: [drlvm][tesing]run modes for DRLVM tests
Posted by Vladimir Ivanov <iv...@gmail.com>.
On 3/2/07, Pavel Ozhdikhin <pa...@gmail.com> wrote:
> We have to make sure the default mode is 100% stable. It's just strange that
> we did not test it in pre-commit checks till now. BTW in the default mode
> all methods first compiled by JET so testing default mode may be considered
> as testing JET + recompilation mechanics. I vote for changing JET to default
> in pre-commit tests.
>
> To ensure coverage of other modes (JET, server) we may add them to the CC
> testing. We should keep balance between the coverage of different modes by
> pre-commit tests and the time needed to run them.
Yes, I'll update the CC to run tests in 4 modes (int+jet+jit+opt).
Should we add the server mode to CC testing? If yes the exclude lists
for this mode should be added.
thanks, Vladimir
>
> Thanks,
> Pavel
>
>
> On 3/2/07, Alexey Varlamov <al...@gmail.com> wrote:
> >
> > 2007/3/2, Vladimir Ivanov <iv...@gmail.com>:
> > > On 01 Mar 2007 18:58:47 +0300, Egor Pasko <eg...@gmail.com> wrote:
> > > > On the 0x28C day of Apache Harmony Vladimir Ivanov wrote:
> > > > > Hello everybody,
> > > > > I think that we also should run DRLVM tests in the default (without
> > > > > any switches) mode.
> > > >
> > > > why? has there been any test that failed in default mode while passing
> > > > in jet, opt, server modes?
> >
> > OTOH the jet mode did not detect regressions for a long, either. The
> > question of default settings is usually a subtle matter, but indeed
> > testing of default mode by default seems reasonable ;)
> >
> > >
> > > Seems we should check the default behavior of VM. And yes, this mode
> > > has some specific bugs, for example, Harmony-2692.
> > >
> > > Thanks, Vladimir
> > >
> > > >
> > > > > Issue 3276 was created to track it. Note, this
> > > > > issue also update default test run modes from "int, jet, opt" to
> > "int,
> > > > > def, opt".
> > > > >
> > > > > Thanks, Vladimir
> > > > >
> > > >
> > > > --
> > > > Egor Pasko
> > > >
> > > >
> > >
> >
>
Re: [drlvm][tesing]run modes for DRLVM tests
Posted by Pavel Ozhdikhin <pa...@gmail.com>.
We have to make sure the default mode is 100% stable. It's just strange that
we did not test it in pre-commit checks till now. BTW in the default mode
all methods first compiled by JET so testing default mode may be considered
as testing JET + recompilation mechanics. I vote for changing JET to default
in pre-commit tests.
To ensure coverage of other modes (JET, server) we may add them to the CC
testing. We should keep balance between the coverage of different modes by
pre-commit tests and the time needed to run them.
Thanks,
Pavel
On 3/2/07, Alexey Varlamov <al...@gmail.com> wrote:
>
> 2007/3/2, Vladimir Ivanov <iv...@gmail.com>:
> > On 01 Mar 2007 18:58:47 +0300, Egor Pasko <eg...@gmail.com> wrote:
> > > On the 0x28C day of Apache Harmony Vladimir Ivanov wrote:
> > > > Hello everybody,
> > > > I think that we also should run DRLVM tests in the default (without
> > > > any switches) mode.
> > >
> > > why? has there been any test that failed in default mode while passing
> > > in jet, opt, server modes?
>
> OTOH the jet mode did not detect regressions for a long, either. The
> question of default settings is usually a subtle matter, but indeed
> testing of default mode by default seems reasonable ;)
>
> >
> > Seems we should check the default behavior of VM. And yes, this mode
> > has some specific bugs, for example, Harmony-2692.
> >
> > Thanks, Vladimir
> >
> > >
> > > > Issue 3276 was created to track it. Note, this
> > > > issue also update default test run modes from "int, jet, opt" to
> "int,
> > > > def, opt".
> > > >
> > > > Thanks, Vladimir
> > > >
> > >
> > > --
> > > Egor Pasko
> > >
> > >
> >
>
Re: [drlvm][tesing]run modes for DRLVM tests
Posted by Alexey Varlamov <al...@gmail.com>.
Advancing the idea to test most demanded configs, we could include
"server" mode to testing - say instead of opt. How about that?
2007/3/2, Alexey Varlamov <al...@gmail.com>:
> 2007/3/2, Vladimir Ivanov <iv...@gmail.com>:
> > On 01 Mar 2007 18:58:47 +0300, Egor Pasko <eg...@gmail.com> wrote:
> > > On the 0x28C day of Apache Harmony Vladimir Ivanov wrote:
> > > > Hello everybody,
> > > > I think that we also should run DRLVM tests in the default (without
> > > > any switches) mode.
> > >
> > > why? has there been any test that failed in default mode while passing
> > > in jet, opt, server modes?
>
> OTOH the jet mode did not detect regressions for a long, either. The
> question of default settings is usually a subtle matter, but indeed
> testing of default mode by default seems reasonable ;)
>
> >
> > Seems we should check the default behavior of VM. And yes, this mode
> > has some specific bugs, for example, Harmony-2692.
> >
> > Thanks, Vladimir
> >
> > >
> > > > Issue 3276 was created to track it. Note, this
> > > > issue also update default test run modes from "int, jet, opt" to "int,
> > > > def, opt".
> > > >
> > > > Thanks, Vladimir
> > > >
> > >
> > > --
> > > Egor Pasko
> > >
> > >
> >
>
Re: [drlvm][tesing]run modes for DRLVM tests
Posted by Alexey Varlamov <al...@gmail.com>.
2007/3/2, Vladimir Ivanov <iv...@gmail.com>:
> On 01 Mar 2007 18:58:47 +0300, Egor Pasko <eg...@gmail.com> wrote:
> > On the 0x28C day of Apache Harmony Vladimir Ivanov wrote:
> > > Hello everybody,
> > > I think that we also should run DRLVM tests in the default (without
> > > any switches) mode.
> >
> > why? has there been any test that failed in default mode while passing
> > in jet, opt, server modes?
OTOH the jet mode did not detect regressions for a long, either. The
question of default settings is usually a subtle matter, but indeed
testing of default mode by default seems reasonable ;)
>
> Seems we should check the default behavior of VM. And yes, this mode
> has some specific bugs, for example, Harmony-2692.
>
> Thanks, Vladimir
>
> >
> > > Issue 3276 was created to track it. Note, this
> > > issue also update default test run modes from "int, jet, opt" to "int,
> > > def, opt".
> > >
> > > Thanks, Vladimir
> > >
> >
> > --
> > Egor Pasko
> >
> >
>
Re: [drlvm][tesing]run modes for DRLVM tests
Posted by Vladimir Ivanov <iv...@gmail.com>.
On 01 Mar 2007 18:58:47 +0300, Egor Pasko <eg...@gmail.com> wrote:
> On the 0x28C day of Apache Harmony Vladimir Ivanov wrote:
> > Hello everybody,
> > I think that we also should run DRLVM tests in the default (without
> > any switches) mode.
>
> why? has there been any test that failed in default mode while passing
> in jet, opt, server modes?
Seems we should check the default behavior of VM. And yes, this mode
has some specific bugs, for example, Harmony-2692.
Thanks, Vladimir
>
> > Issue 3276 was created to track it. Note, this
> > issue also update default test run modes from "int, jet, opt" to "int,
> > def, opt".
> >
> > Thanks, Vladimir
> >
>
> --
> Egor Pasko
>
>
Re: [drlvm][tesing]run modes for DRLVM tests
Posted by Egor Pasko <eg...@gmail.com>.
On the 0x28C day of Apache Harmony Vladimir Ivanov wrote:
> Hello everybody,
> I think that we also should run DRLVM tests in the default (without
> any switches) mode.
why? has there been any test that failed in default mode while passing
in jet, opt, server modes?
> Issue 3276 was created to track it. Note, this
> issue also update default test run modes from "int, jet, opt" to "int,
> def, opt".
>
> Thanks, Vladimir
>
--
Egor Pasko
Re: [drlvm][tesing]run modes for DRLVM tests
Posted by "Geir Magnusson Jr." <ge...@pobox.com>.
+1
On Mar 1, 2007, at 5:34 AM, Vladimir Ivanov wrote:
> Hello everybody,
> I think that we also should run DRLVM tests in the default (without
> any switches) mode. Issue 3276 was created to track it. Note, this
> issue also update default test run modes from "int, jet, opt" to "int,
> def, opt".
>
> Thanks, Vladimir