VectorICs: ia32 store ics need a virtual register.

VectorICs: ia32 store ics need a virtual register.

Some pretty hacky code was used to carry out the tail-call
handler dispatch on ia32 vector stores due to a lack
of free registers. It really tanks performance. A better
approach is to use a virtual register on the isolate.

BUG=

Committed: https://crrev.com/b26e98f19b5e77df50cb4a259f099ee6afa335e3
Cr-Commit-Position: refs/heads/master@{#30718}

[mvstanton, 2015-09-14 13:23:45]

mvstanton@chromium.org changed reviewers:
+ jkummerow@chromium.org

[mvstanton, 2015-09-14 13:23:45]

Hi Jakob,
I've reintroduced the semi-comical virtual register we discussed before. The
performance profile is much, much closer to what is required for vector stores.
Only ia32 changes are required. Thanks for the look,
--Michael

ps - I'll turn off the flag before checkin.

[Jakob Kummerow, 2015-09-14 14:02:42]

jkummerow@chromium.org changed reviewers:
+ vogelheim@chromium.org

[Jakob Kummerow, 2015-09-14 14:02:42]

+vogelheim for src/snapshot/ OWNERS approval.

LGTM with nits (comments like proper capitalization).

https://codereview.chromium.org/1336313002/diff/1/src/ia32/code-stubs-ia32.cc
File src/ia32/code-stubs-ia32.cc (right):

https://codereview.chromium.org/1336313002/diff/1/src/ia32/code-stubs-ia32.cc...
src/ia32/code-stubs-ia32.cc:4599: __ pop(handler);  // pop "value".
nit: s/pop/Pop/, again below.

https://codereview.chromium.org/1336313002/diff/1/src/ic/ia32/stub-cache-ia32.cc
File src/ic/ia32/stub-cache-ia32.cc (right):

https://codereview.chromium.org/1336313002/diff/1/src/ic/ia32/stub-cache-ia32...
src/ic/ia32/stub-cache-ia32.cc:73: __ pop(extra);  // pop "slot".
nit: s/pop/Pop/, again below

[vogelheim, 2015-09-14 14:21:58]

lgtm for src/snapshot

https://codereview.chromium.org/1336313002/diff/20001/src/snapshot/serialize.cc
File src/snapshot/serialize.cc (right):

https://codereview.chromium.org/1336313002/diff/20001/src/snapshot/serialize....
src/snapshot/serialize.cc:144:
Add(ExternalReference::vector_store_virtual_register(isolate).address(),
super nitpick: Maybe move a few lines up, to the end of the "regular"
references? (Rather than between the debug_* and #ifdef-ed ones?)

[mvstanton, 2015-09-14 14:25:19]

The CQ bit was checked by mvstanton@chromium.org

[mvstanton, 2015-09-14 14:25:19]

The patchset sent to the CQ was uploaded after l-g-t-m from
jkummerow@chromium.org, vogelheim@chromium.org
Link to the patchset: https://codereview.chromium.org/1336313002/#ps40001
(title: "One more comments.")

[commit-bot: I haz the power, 2015-09-14 14:25:27]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1336313002/40001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1336313002/40001

[mvstanton, 2015-09-14 14:25:36]

thanks guys!
--Michael

https://codereview.chromium.org/1336313002/diff/1/src/ia32/code-stubs-ia32.cc
File src/ia32/code-stubs-ia32.cc (right):

https://codereview.chromium.org/1336313002/diff/1/src/ia32/code-stubs-ia32.cc...
src/ia32/code-stubs-ia32.cc:4599: __ pop(handler);  // pop "value".
On 2015/09/14 14:02:42, Jakob wrote:
> nit: s/pop/Pop/, again below.

Done.

https://codereview.chromium.org/1336313002/diff/1/src/ic/ia32/stub-cache-ia32.cc
File src/ic/ia32/stub-cache-ia32.cc (right):

https://codereview.chromium.org/1336313002/diff/1/src/ic/ia32/stub-cache-ia32...
src/ic/ia32/stub-cache-ia32.cc:73: __ pop(extra);  // pop "slot".
On 2015/09/14 14:02:42, Jakob wrote:
> nit: s/pop/Pop/, again below

Done.

https://codereview.chromium.org/1336313002/diff/20001/src/snapshot/serialize.cc
File src/snapshot/serialize.cc (right):

https://codereview.chromium.org/1336313002/diff/20001/src/snapshot/serialize....
src/snapshot/serialize.cc:144:
Add(ExternalReference::vector_store_virtual_register(isolate).address(),
On 2015/09/14 14:21:58, vogelheim wrote:
> super nitpick: Maybe move a few lines up, to the end of the "regular"
> references? (Rather than between the debug_* and #ifdef-ed ones?)

Done.

[commit-bot: I haz the power, 2015-09-14 14:49:28]

Committed patchset #3 (id:40001)

[commit-bot: I haz the power, 2015-09-14 14:49:49]

Patchset 3 (id:??) landed as
https://crrev.com/b26e98f19b5e77df50cb4a259f099ee6afa335e3
Cr-Commit-Position: refs/heads/master@{#30718}

[Michael Achenbach, 2015-09-14 18:58:32]

A revert of this CL (patchset #3 id:40001) has been created in
https://codereview.chromium.org/1340123002/ by machenbach@chromium.org.

The reason for reverting is: [Sheriff] Breaks GC stress on mac:
http://build.chromium.org/p/client.v8/builders/V8%20Mac%20GC%20Stress/builds/....