Removing redundant stub for runtime native calls.

Removing redundant stub for runtime native calls.

Committed: http://code.google.com/p/v8/source/detail?r=3745

[Mads Ager (chromium), 2010-01-27 08:56:13]

I believe you are right that the RuntimeStub is now obsolete.  Please get rid of
it completely. There is one other use in the ia32 macro-assembler and it is used
in the x64 and arm macro-assemblers as well.  Once you get rid of all the uses,
please remove the RuntimeStub from codegen.h and codegen.cc as well.

[SeRya, 2010-01-28 19:55:56]

RuntimeStub removed. It used to be used by the debugger to identify the
'debugger' statement (added a test covering this case). Now CEntryDebugBreak is
used for this purpose.

[Mads Ager (chromium), 2010-01-29 09:21:34]

Søren, could you double check the debugger part of this change?

If the debugger part of the change looks good to Søren, the rest looks good to
me.

[Søren Thygesen Gjesse, 2010-01-29 10:02:12]

LGTM

The debugger part looks OK.

http://codereview.chromium.org/543207/diff/9007/8015
File src/codegen.cc (right):

http://codereview.chromium.org/543207/diff/9007/8015#newcode498
src/codegen.cc:498: masm->TailCallRuntime(ExternalReference(f),
I think these parameters can fit on one line.

http://codereview.chromium.org/543207/diff/9007/8012
File src/codegen.h (right):

http://codereview.chromium.org/543207/diff/9007/8012#newcode388
src/codegen.h:388: // Markd the debugger statemet to be recognized bu debugger
(by the MajorKey)
Markd -> mark

http://codereview.chromium.org/543207/diff/9007/8012#newcode389
src/codegen.h:389: class CEntryDebugBreakStub : public CodeStub {
This is not in itself a C-entry stub. Please rename it to DebugBreakStub and its
major key to DebugBreak.

As it is only used for the debugger statement calling it DebuggerStatement would
be more precise and will not cause confusion with other ways to enter the
debugger.

Maybe introducing new relocation info type and just using a normal runtime call
could be considered.

http://codereview.chromium.org/543207/diff/9007/8012#newcode397
src/codegen.h:397: int MinorKey() { return 1; }
Use a minor key of 0 as there is only one instance of this stub.

http://codereview.chromium.org/543207/diff/9007/8016
File src/debug.cc (right):

http://codereview.chromium.org/543207/diff/9007/8016#newcode80
src/debug.cc:80: debug_break_stub_ = CEntryDebugBreakStub().GetCode();
As the check is now on the major key the member debug_break_stub_ should not be
needed any more.

http://codereview.chromium.org/543207/diff/9007/8019
File src/x64/codegen-x64.cc (right):

http://codereview.chromium.org/543207/diff/9007/8019#newcode6983
src/x64/codegen-x64.cc:6983: return (result_size_ < 2) ? 0 : result_size_ * 2;
Does this special assignment of minor key on 64-bit Windows make sense any more?

[SeRya, 2010-01-29 11:46:57]

http://codereview.chromium.org/543207/diff/9007/8015
File src/codegen.cc (right):

http://codereview.chromium.org/543207/diff/9007/8015#newcode498
src/codegen.cc:498: masm->TailCallRuntime(ExternalReference(f),
On 2010/01/29 10:02:12, Søren Gjesse wrote:
> I think these parameters can fit on one line.

Done.

http://codereview.chromium.org/543207/diff/9007/8012
File src/codegen.h (right):

http://codereview.chromium.org/543207/diff/9007/8012#newcode388
src/codegen.h:388: // Markd the debugger statemet to be recognized bu debugger
(by the MajorKey)
On 2010/01/29 10:02:12, Søren Gjesse wrote:
> Markd -> mark

Done.

http://codereview.chromium.org/543207/diff/9007/8012#newcode389
src/codegen.h:389: class CEntryDebugBreakStub : public CodeStub {
On 2010/01/29 10:02:12, Søren Gjesse wrote:
> This is not in itself a C-entry stub. Please rename it to DebugBreakStub and
its
> major key to DebugBreak.
> 
> As it is only used for the debugger statement calling it DebuggerStatement
would
> be more precise and will not cause confusion with other ways to enter the
> debugger.
> 
> Maybe introducing new relocation info type and just using a normal runtime
call
> could be considered.

Done.

http://codereview.chromium.org/543207/diff/9007/8012#newcode397
src/codegen.h:397: int MinorKey() { return 1; }
On 2010/01/29 10:02:12, Søren Gjesse wrote:
> Use a minor key of 0 as there is only one instance of this stub.

Done.

http://codereview.chromium.org/543207/diff/9007/8016
File src/debug.cc (right):

http://codereview.chromium.org/543207/diff/9007/8016#newcode80
src/debug.cc:80: debug_break_stub_ = CEntryDebugBreakStub().GetCode();
On 2010/01/29 10:02:12, Søren Gjesse wrote:
> As the check is now on the major key the member debug_break_stub_ should not
be
> needed any more.

Done.

http://codereview.chromium.org/543207/diff/9007/8019
File src/x64/codegen-x64.cc (right):

http://codereview.chromium.org/543207/diff/9007/8019#newcode6983
src/x64/codegen-x64.cc:6983: return (result_size_ < 2) ? 0 : result_size_ * 2;
On 2010/01/29 10:02:12, Søren Gjesse wrote:
> Does this special assignment of minor key on 64-bit Windows make sense any
more?

Probably so (since generated code for WIN64 depends on result_size_). But this
line could be simplified since values could not conflict with a descendant class
anymore.