Issue 22833002: Fix regressions triggered by map invalidation during graph creation.

Issue 22833002: Fix regressions triggered by map invalidation during graph creation. (Closed)

Created:
7 years, 4 months ago by Yang

Modified:
7 years, 4 months ago

Reviewers:
titzer, Jakob Kummerow

CC:
v8-dev

Base URL:
https://v8.googlecode.com/svn/branches/bleeding_edge

Visibility:
Public.

More Reviews

Description

Fix regressions triggered by map invalidation during graph creation. R=jkummerow@chromium.org BUG=

Patch Set 1 #

Total comments: 5

Created: 7 years, 4 months ago

Download [raw] [tar.bz2]

	Unified diffs	Side-by-side diffs	Stats (+62 lines, -25 lines)			Patch
M	src/assert-scope.h	View	2 chunks	+9 lines, -0 lines	0 comments	Download
M	src/compiler.h	View	3 chunks	+10 lines, -4 lines	0 comments	Download
M	src/compiler.cc	View	4 chunks	+10 lines, -0 lines	3 comments	Download
	src/objects.cc	View	1 chunk	+1 line, -0 lines	0 comments	Download
A +	test/mjsunit/regress/regress-map-invalidation-1.js	View	1 chunk	+14 lines, -15 lines	2 comments	Download
A +	test/mjsunit/regress/regress-map-invalidation-2.js	View	1 chunk	+18 lines, -6 lines	0 comments	Download

Messages

Total messages: 2 (0 generated)

Expand Messages | Collapse Messages

titzer

https://codereview.chromium.org/22833002/diff/1/src/compiler.cc File src/compiler.cc (right): https://codereview.chromium.org/22833002/diff/1/src/compiler.cc#newcode963 src/compiler.cc:963: } else { We need to info->AbortOptimization() here if ...

7 years, 4 months ago (2013-08-12 13:28:58 UTC) #1

Yang

7 years, 4 months ago (2013-08-12 13:49:22 UTC) #2

Message was sent while issue was closed.

https://codereview.chromium.org/22833002/diff/1/src/compiler.cc
File src/compiler.cc (right):

https://codereview.chromium.org/22833002/diff/1/src/compiler.cc#newcode963
src/compiler.cc:963: } else {
On 2013/08/12 13:28:58, titzer wrote:
> We need to info->AbortOptimization() here if the mode was not BASE or NONOPT.
> Uggh, what a mess.

Since I introduced a new field, we don't have to do that anymore. The mode_
field is no longer overloaded with that map dependency value.

I added a check in InstallFullCode that we are indeed looking at unoptimized
code.

https://codereview.chromium.org/22833002/diff/1/test/mjsunit/regress/regress-...
File test/mjsunit/regress/regress-map-invalidation-1.js (right):

https://codereview.chromium.org/22833002/diff/1/test/mjsunit/regress/regress-...
test/mjsunit/regress/regress-map-invalidation-1.js:39: for (var i = 0; i <
100000; i++) {
On 2013/08/12 13:28:58, titzer wrote:
> Are you trying to force an OSR here?

yup. we could of course also use %OptimizeFunctionOnNextCall(f, "osr") if you
prefer it that way.

Expand Messages | Collapse Messages