Change RelocInfo to write 64-bit data field on x64 architecture.

Change RelocInfo to write 64-bit data field on x64 architecture.

Committed: http://code.google.com/p/v8/source/detail?r=2081

[William Hesse, 2009-05-28 13:16:09]

Tested on ia32, with --debug-code and --print-code.

http://codereview.chromium.org/115860/diff/10/12
File src/assembler.cc (right):

http://codereview.chromium.org/115860/diff/10/12#newcode186
Line 186: data_delta = data_delta >> kBitsPerByte;
ArithmeticShiftRight is defined for ints, not intptr_t.  We cannot overload it
on integer type, so we either need to
make multiple versions of it, or else use >> on negative ints without
encapsulating them.

[Kevin Millikin (Chromium), 2009-05-29 09:43:37]

LGTM.

http://codereview.chromium.org/115860/diff/10/12
File src/assembler.cc (right):

http://codereview.chromium.org/115860/diff/10/12#newcode186
Line 186: data_delta = data_delta >> kBitsPerByte;
On 2009/05/28 13:16:09, William Hesse wrote:
> ArithmeticShiftRight is defined for ints, not intptr_t.  We cannot overload it
> on integer type, so we either need to
> make multiple versions of it, or else use >> on negative ints without
> encapsulating them.

This is fine, but there should be a comment here.  There should also be a test
somewhere or a STATIC_ASSERT that would fail if an implementation did the wrong
thing.

http://codereview.chromium.org/115860/diff/10/12#newcode300
Line 300: rinfo_.data_ += signed_b >> kPositionTypeTagBits;
Possibly also a comment here.

[William Hesse, 2009-05-29 11:16:08]

Added test to test-utils.cc

http://codereview.chromium.org/115860/diff/10/12
File src/assembler.cc (right):

http://codereview.chromium.org/115860/diff/10/12#newcode186
Line 186: data_delta = data_delta >> kBitsPerByte;
On 2009/05/29 09:43:37, Kevin Millikin wrote:
> On 2009/05/28 13:16:09, William Hesse wrote:
> > ArithmeticShiftRight is defined for ints, not intptr_t.  We cannot overload
it
> > on integer type, so we either need to
> > make multiple versions of it, or else use >> on negative ints without
> > encapsulating them.
> 
> This is fine, but there should be a comment here.  There should also be a test
> somewhere or a STATIC_ASSERT that would fail if an implementation did the
wrong
> thing.

Added test to test-utils.cc.
Added comment.

http://codereview.chromium.org/115860/diff/10/12#newcode300
Line 300: rinfo_.data_ += signed_b >> kPositionTypeTagBits;
On 2009/05/29 09:43:37, Kevin Millikin wrote:
> Possibly also a comment here.

Done.