Implement kToBeExecutedOnceCodeAge.

Implement kToBeExecutedOnceCodeAge.

An initial 'code age' state that will turn into a 'pre-aging' code age only after it was executed the first time.

BUG=470930
LOG=Y

Committed: https://crrev.com/6b905c3a160e55d84606d866f0cd90f929e39ac4
Cr-Commit-Position: refs/heads/master@{#28162}

[vogelheim, 2015-04-28 16:54:00]

vogelheim@chromium.org changed reviewers:
+ rmcilroy@chromium.org, yangguo@chromium.org

[vogelheim, 2015-04-28 16:54:00]

- Since we're not sure whether this is a good idea in the first place, please do
suggest alternatives if it isn't.

- I removed the 'effective age' function. It assumed the 'virtual' states could
be mapped to a proper state for IsOld and MakeOlder, but this doesn't work for
the new state. I don't think the code suffered as a result.

- Non-x64 platform code is missing, but I assume that'll be one-liners as well.

- Please let me know if you're not the right reviewers.

[vogelheim, 2015-04-29 09:26:43]

On 2015/04/28 16:54:00, vogelheim wrote:
> - Since we're not sure whether this is a good idea in the first place, please
do
> suggest alternatives if it isn't.
> 
> - I removed the 'effective age' function. It assumed the 'virtual' states
could
> be mapped to a proper state for IsOld and MakeOlder, but this doesn't work for
> the new state. I don't think the code suffered as a result.
> 
> - Non-x64 platform code is missing, but I assume that'll be one-liners as
well.
> 
> - Please let me know if you're not the right reviewers.

Meeh. The "removed the 'effective age' function" part breaks the previous
pre-aging logic. I'll fix that first.

[rmcilroy, 2015-04-29 10:22:15]

On 2015/04/29 09:26:43, vogelheim wrote:
> On 2015/04/28 16:54:00, vogelheim wrote:
> > - Since we're not sure whether this is a good idea in the first place,
please
> do
> > suggest alternatives if it isn't.
> > 
> > - I removed the 'effective age' function. It assumed the 'virtual' states
> could
> > be mapped to a proper state for IsOld and MakeOlder, but this doesn't work
for
> > the new state. I don't think the code suffered as a result.
> > 
> > - Non-x64 platform code is missing, but I assume that'll be one-liners as
> well.
> > 
> > - Please let me know if you're not the right reviewers.
> 
> Meeh. The "removed the 'effective age' function" part breaks the previous
> pre-aging logic. I'll fix that first.

I had a suspicion that that was going to be the case and was just trying to work
it out :). Thanks for spotting it and let me know if I can help.

Also I like danno's suggestion of transition in this to be old after running.
Could you have a go at doing this? The stub would need to change to be more
similar to the  kNotExecuted one.

[vogelheim, 2015-04-29 14:31:11]

On 2015/04/29 10:22:15, rmcilroy wrote:
> On 2015/04/29 09:26:43, vogelheim wrote:
> > On 2015/04/28 16:54:00, vogelheim wrote:
> > > - Since we're not sure whether this is a good idea in the first place,
> please
> > do
> > > suggest alternatives if it isn't.
> > > 
> > > - I removed the 'effective age' function. It assumed the 'virtual' states
> > could
> > > be mapped to a proper state for IsOld and MakeOlder, but this doesn't work
> for
> > > the new state. I don't think the code suffered as a result.
> > > 
> > > - Non-x64 platform code is missing, but I assume that'll be one-liners as
> > well.
> > > 
> > > - Please let me know if you're not the right reviewers.
> > 
> > Meeh. The "removed the 'effective age' function" part breaks the previous
> > pre-aging logic. I'll fix that first.
> 
> I had a suspicion that that was going to be the case and was just trying to
work
> it out :). Thanks for spotting it and let me know if I can help.

Should be fixed now.

> Also I like danno's suggestion of transition in this to be old after running.
> Could you have a go at doing this? The stub would need to change to be more
> similar to the  kNotExecuted one.

I'll do that next.

(Not sure if you want to review already. I expect the execution-transition
change to be limited to the builtins-*.cc files.)

[rmcilroy, 2015-04-29 14:39:04]

On 2015/04/29 14:31:11, vogelheim wrote:
> On 2015/04/29 10:22:15, rmcilroy wrote:
> > On 2015/04/29 09:26:43, vogelheim wrote:
> > > On 2015/04/28 16:54:00, vogelheim wrote:
> > > > - Since we're not sure whether this is a good idea in the first place,
> > please
> > > do
> > > > suggest alternatives if it isn't.
> > > > 
> > > > - I removed the 'effective age' function. It assumed the 'virtual'
states
> > > could
> > > > be mapped to a proper state for IsOld and MakeOlder, but this doesn't
work
> > for
> > > > the new state. I don't think the code suffered as a result.
> > > > 
> > > > - Non-x64 platform code is missing, but I assume that'll be one-liners
as
> > > well.
> > > > 
> > > > - Please let me know if you're not the right reviewers.
> > > 
> > > Meeh. The "removed the 'effective age' function" part breaks the previous
> > > pre-aging logic. I'll fix that first.
> > 
> > I had a suspicion that that was going to be the case and was just trying to
> work
> > it out :). Thanks for spotting it and let me know if I can help.
> 
> Should be fixed now.
> 
> > Also I like danno's suggestion of transition in this to be old after
running.
> > Could you have a go at doing this? The stub would need to change to be more
> > similar to the  kNotExecuted one.
> 
> I'll do that next.
> 
> (Not sure if you want to review already. I expect the execution-transition
> change to be limited to the builtins-*.cc files.)

[rmcilroy, 2015-04-29 14:39:30]

On 2015/04/29 14:39:04, rmcilroy wrote:
> On 2015/04/29 14:31:11, vogelheim wrote:
> > On 2015/04/29 10:22:15, rmcilroy wrote:
> > > On 2015/04/29 09:26:43, vogelheim wrote:
> > > > On 2015/04/28 16:54:00, vogelheim wrote:
> > > > > - Since we're not sure whether this is a good idea in the first place,
> > > please
> > > > do
> > > > > suggest alternatives if it isn't.
> > > > > 
> > > > > - I removed the 'effective age' function. It assumed the 'virtual'
> states
> > > > could
> > > > > be mapped to a proper state for IsOld and MakeOlder, but this doesn't
> work
> > > for
> > > > > the new state. I don't think the code suffered as a result.
> > > > > 
> > > > > - Non-x64 platform code is missing, but I assume that'll be one-liners
> as
> > > > well.
> > > > > 
> > > > > - Please let me know if you're not the right reviewers.
> > > > 
> > > > Meeh. The "removed the 'effective age' function" part breaks the
previous
> > > > pre-aging logic. I'll fix that first.
> > > 
> > > I had a suspicion that that was going to be the case and was just trying
to
> > work
> > > it out :). Thanks for spotting it and let me know if I can help.
> > 
> > Should be fixed now.
> > 
> > > Also I like danno's suggestion of transition in this to be old after
> running.
> > > Could you have a go at doing this? The stub would need to change to be
more
> > > similar to the  kNotExecuted one.
> > 
> > I'll do that next.
> > 
> > (Not sure if you want to review already. I expect the execution-transition
> > change to be limited to the builtins-*.cc files.)

Generally looks fine but I'll review properly once you've done the kNotExecuted
change.

[vogelheim, 2015-04-29 16:43:29]

On 2015/04/29 14:39:30, rmcilroy wrote:
> On 2015/04/29 14:39:04, rmcilroy wrote:
> > On 2015/04/29 14:31:11, vogelheim wrote:
> > > On 2015/04/29 10:22:15, rmcilroy wrote:
> > > > On 2015/04/29 09:26:43, vogelheim wrote:
> > > > > On 2015/04/28 16:54:00, vogelheim wrote:
> > > > > > - Since we're not sure whether this is a good idea in the first
place,
> > > > please
> > > > > do
> > > > > > suggest alternatives if it isn't.
> > > > > > 
> > > > > > - I removed the 'effective age' function. It assumed the 'virtual'
> > states
> > > > > could
> > > > > > be mapped to a proper state for IsOld and MakeOlder, but this
doesn't
> > work
> > > > for
> > > > > > the new state. I don't think the code suffered as a result.
> > > > > > 
> > > > > > - Non-x64 platform code is missing, but I assume that'll be
one-liners
> > as
> > > > > well.
> > > > > > 
> > > > > > - Please let me know if you're not the right reviewers.
> > > > > 
> > > > > Meeh. The "removed the 'effective age' function" part breaks the
> previous
> > > > > pre-aging logic. I'll fix that first.
> > > > 
> > > > I had a suspicion that that was going to be the case and was just trying
> to
> > > work
> > > > it out :). Thanks for spotting it and let me know if I can help.
> > > 
> > > Should be fixed now.
> > > 
> > > > Also I like danno's suggestion of transition in this to be old after
> > running.
> > > > Could you have a go at doing this? The stub would need to change to be
> more
> > > > similar to the  kNotExecuted one.
> > > 
> > > I'll do that next.
> > > 
> > > (Not sure if you want to review already. I expect the execution-transition
> > > change to be limited to the builtins-*.cc files.)
> 
> Generally looks fine but I'll review properly once you've done the
kNotExecuted
> change.

That would be right this minute! :)


This turned out to be really easy, since the code-side transition for
kToBeExecutedOnceCodeAge is now the same as for kNotExecutedCodeAge. So both now
use the same pre-aging logic, but one enters this state based entirely on GCs,
while the new state will always wait for a first execution before the
"pre-aging" logic applies.

[rmcilroy, 2015-04-30 10:42:04]

Looks great, thanks!

lgtm.

https://codereview.chromium.org/1107233004/diff/40001/src/objects.cc
File src/objects.cc (right):

https://codereview.chromium.org/1107233004/diff/40001/src/objects.cc#newcode1...
src/objects.cc:11335: // IsOldAge defines the collection criteria for a Code
objectsgit diff.
remove "git diff"  (typing into the wrong window? ;)).

Also add fullstop.

[vogelheim, 2015-04-30 11:01:47]

https://codereview.chromium.org/1107233004/diff/40001/src/objects.cc
File src/objects.cc (right):

https://codereview.chromium.org/1107233004/diff/40001/src/objects.cc#newcode1...
src/objects.cc:11335: // IsOldAge defines the collection criteria for a Code
objectsgit diff.
On 2015/04/30 10:42:03, rmcilroy wrote:
> remove "git diff"  (typing into the wrong window? ;)).

Yes. :)

> Also add fullstop.

Both done.

[vogelheim, 2015-04-30 11:45:11]

The CQ bit was checked by vogelheim@chromium.org

[vogelheim, 2015-04-30 11:45:11]

The patchset sent to the CQ was uploaded after l-g-t-m from
rmcilroy@chromium.org
Link to the patchset: https://codereview.chromium.org/1107233004/#ps60001
(title: "Fix comment.")

[commit-bot: I haz the power, 2015-04-30 11:45:34]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1107233004/60001

[Yang, 2015-04-30 11:51:29]

lgtm

[commit-bot: I haz the power, 2015-04-30 12:31:42]

Committed patchset #4 (id:60001)

[commit-bot: I haz the power, 2015-04-30 12:32:03]

Patchset 4 (id:??) landed as
https://crrev.com/6b905c3a160e55d84606d866f0cd90f929e39ac4
Cr-Commit-Position: refs/heads/master@{#28162}