x64 support for d-to-i (truncated)

x64 support for d-to-i (truncated)

R=yangguo@chromium.org

Committed: http://code.google.com/p/v8/source/detail?r=15761

[Yang, 2013-07-18 14:54:38]

LGTM with comments.

https://codereview.chromium.org/18333012/diff/5001/src/x64/code-stubs-x64.cc
File src/x64/code-stubs-x64.cc (right):

https://codereview.chromium.org/18333012/diff/5001/src/x64/code-stubs-x64.cc#...
src/x64/code-stubs-x64.cc:634: // to calculate the result if ecx is the
requested return register.
comments do not apply to x64 registers

https://codereview.chromium.org/18333012/diff/5001/src/x64/code-stubs-x64.cc#...
src/x64/code-stubs-x64.cc:671: __ negl(result_reg);
We don't do zero-extension while moving anywhere. Could it happen that the upper
32-bit of the result just carries over?

https://codereview.chromium.org/18333012/diff/5001/test/cctest/test-code-stub...
File test/cctest/test-code-stubs-x64.cc (right):

https://codereview.chromium.org/18333012/diff/5001/test/cctest/test-code-stub...
test/cctest/test-code-stubs-x64.cc:137: 
Why not make it something like
Register source_registers[7] = {rsp, rax, rbx, rcx, rdx, rsi, rdi};
Register dest_registers[6] = {rax, rbx, rcx, rdx, rsi, rdi};

for (int s = 0; s < 7; s++) {
  for (int d = 0; d < 6; d++) {
    RunAllTruncationTests(
        MakeConvertDToIFuncTrampoline(isolate,
                                      source_registers[s],
                                      dest_registers[d]));
  }
}

[danno, 2013-07-19 08:21:49]

Committed patchset #4 manually as r15761 (presubmit successful).

[danno, 2013-07-19 08:56:50]

Feedback addressed, landed

https://codereview.chromium.org/18333012/diff/5001/src/x64/code-stubs-x64.cc
File src/x64/code-stubs-x64.cc (right):

https://codereview.chromium.org/18333012/diff/5001/src/x64/code-stubs-x64.cc#...
src/x64/code-stubs-x64.cc:634: // to calculate the result if ecx is the
requested return register.
On 2013/07/18 14:54:38, Yang wrote:
> comments do not apply to x64 registers

Done.

https://codereview.chromium.org/18333012/diff/5001/src/x64/code-stubs-x64.cc#...
src/x64/code-stubs-x64.cc:671: __ negl(result_reg);
On 2013/07/18 14:54:38, Yang wrote:
> We don't do zero-extension while moving anywhere. Could it happen that the
upper
> 32-bit of the result just carries over?

As discussed, on x64 we ignore the top 32-bits for most operations and
explicitly clear them before we need them if we do.