ARM: Fix deoptimizer for two-word call sites.

ARM: Fix deoptimizer for two-word call sites.

R=svenpanne@chromium.org,kmillikin@chromium.org
BUG=v8:1584
Committed: http://code.google.com/p/v8/source/detail?r=8868

[Sven Panne, 2011-08-05 13:47:15]

LGTM

[Kevin Millikin (Chromium), 2011-08-05 15:10:15]

http://codereview.chromium.org/7585006/diff/1/src/arm/deoptimizer-arm.cc
File src/arm/deoptimizer-arm.cc (right):

http://codereview.chromium.org/7585006/diff/1/src/arm/deoptimizer-arm.cc#newc...
src/arm/deoptimizer-arm.cc:96: #ifdef DEBUG
There must be a better fix than this.  The patcher is designed to treat an
existing code object like a macro assembler buffer, and overwrite exactly N
bytes.

The assert is that we overwrote too few bytes.  This fix just disables the
assert at this site.

Is it possible to change patch_size() to give the correct answer in this case?

[Michael Starzinger, 2011-08-08 09:31:25]

Added new patch set.

http://codereview.chromium.org/7585006/diff/1/src/arm/deoptimizer-arm.cc
File src/arm/deoptimizer-arm.cc (right):

http://codereview.chromium.org/7585006/diff/1/src/arm/deoptimizer-arm.cc#newc...
src/arm/deoptimizer-arm.cc:96: #ifdef DEBUG
On 2011/08/05 15:10:16, Kevin Millikin wrote:
> There must be a better fix than this.  The patcher is designed to treat an
> existing code object like a macro assembler buffer, and overwrite exactly N
> bytes.
> 
> The assert is that we overwrote too few bytes.  This fix just disables the
> assert at this site.
> 
> Is it possible to change patch_size() to give the correct answer in this case?

I think making patch_size() giving the correct answer is not possible, because
the size depends on the call target, and that target is not known when the
compiler generates the original (i.e. to-be-patched) code. So it returns the
maximum size (i.e. 3 words) of a call site.

But what we can do, is get the correct size before instantiating a CodePatcher
for each safe-point. See the recent patch set for my proposal.

[Kevin Millikin (Chromium), 2011-08-09 12:46:26]

LGTM, thanks.