Fix a crash when generating forward jumps to labels at very high assembly offsets

Fix a crash when generating forward jumps to labels at very high assembly offsets
    
The first jump to a specific label was marked as jump to absolute
position -4. This value was stored in the assembly as a branch to a
offset (-4 - (instruction offset + 8)). The offset is only 24 bit
long on ARM. Thus instruction offsets higher than 2^23 - 12 would overflow
the offset.
    
Fix by denoting the first jump to a label by storing the jump
instruction location as the target. This will result in offset of -8,
which of course always fits in the branch instruction.

BUG=v8:2736
TEST=cctest/test-assembler-arm/17
R=bmeurer@chromium.org, svenpanne@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=15997

[Kimmo Kinnunen, 2013-06-18 05:51:45]

There might be a problem with the testcase, it fails in stress 5/5.

============ Stress 5/5 ============
#
# Fatal error in Malloced operator new
# Allocation failed - process out of memory

[Yang, 2013-06-25 12:09:46]

https://codereview.chromium.org/17116006/diff/1/test/mjsunit/big-code-segment.js
File test/mjsunit/big-code-segment.js (right):

https://codereview.chromium.org/17116006/diff/1/test/mjsunit/big-code-segment...
test/mjsunit/big-code-segment.js:1: // Copyright 2009 the V8 project authors.
All rights reserved.
Update year to 2013 since this is a new file.

https://codereview.chromium.org/17116006/diff/1/test/mjsunit/big-code-segment...
test/mjsunit/big-code-segment.js:31: var hunk = "for (var i = 0; i < 32; ++i) {
if (v & (1 << i)) ones++; else zeros++; }";
We got a 80 character limit per line, please reformat your code accordingly!

[Kimmo Kinnunen, 2013-07-15 12:59:23]

The testcase is fairly slow and uses all the memory.
Ideas how to improve the testcase so that it wouldn't ruin the test runs would
be appreciated.

[Kimmo Kinnunen, 2013-07-16 11:49:48]

Turned the .js testcase into a cctest

[Benedikt Meurer, 2013-07-31 11:35:55]

LGTM, with nit.

https://codereview.chromium.org/17116006/diff/13001/test/cctest/test-assemble...
File test/cctest/test-assembler-arm.cc (right):

https://codereview.chromium.org/17116006/diff/13001/test/cctest/test-assemble...
test/cctest/test-assembler-arm.cc:1422: for (size_t i = 0; i < 8388608; ++i) { 
// 2^23
nit: use 1 << 23 instead of 8388608.

[Sven Panne, 2013-07-31 13:39:06]

LGTM if the AUTHORS are changed as proposed.

https://codereview.chromium.org/17116006/diff/13001/AUTHORS
File AUTHORS (right):

https://codereview.chromium.org/17116006/diff/13001/AUTHORS#newcode37
AUTHORS:37: Kimmo Kinnunen <kkinnunen@nvidia.com>
If I understand our email discussion correctly, a NVIDIA line should be added
near the top instead of the individual author entry.

[Kimmo Kinnunen, 2013-08-01 06:06:53]

https://codereview.chromium.org/17116006/diff/13001/AUTHORS
File AUTHORS (right):

https://codereview.chromium.org/17116006/diff/13001/AUTHORS#newcode37
AUTHORS:37: Kimmo Kinnunen <kkinnunen@nvidia.com>
On 2013/07/31 13:39:06, Sven Panne wrote:
> If I understand our email discussion correctly, a NVIDIA line should be added
> near the top instead of the individual author entry.

Done.

https://codereview.chromium.org/17116006/diff/13001/test/cctest/test-assemble...
File test/cctest/test-assembler-arm.cc (right):

https://codereview.chromium.org/17116006/diff/13001/test/cctest/test-assemble...
test/cctest/test-assembler-arm.cc:1422: for (size_t i = 0; i < 8388608; ++i) { 
// 2^23
On 2013/07/31 11:35:55, Benedikt Meurer wrote:
> nit: use 1 << 23 instead of 8388608.

Done.

[Benedikt Meurer, 2013-08-01 08:13:32]

Committed patchset #5 manually as r15997 (presubmit successful).