Implementation of a few more assembly instructions on x64

Implementation of a few more assembly instructions on x64

Committed: http://code.google.com/p/v8/source/detail?r=2043

[William Hesse, 2009-05-22 15:15:43]

A few instruction implementations, and comment changes.
There is supposed to be a 15-byte limit on instruction length, that applies also
to x64, with a GP fault caused
by violations.  The 17 byte figure is thus wrong.

[Kevin Millikin (Chromium), 2009-05-25 11:45:04]

http://codereview.chromium.org/113767/diff/1/3
File src/x64/assembler-x64.h (right):

http://codereview.chromium.org/113767/diff/1/3#newcode478
Line 478: // - Function names correspond one-to-one to ia32 instruction
mnemonics.
ia32?

http://codereview.chromium.org/113767/diff/1/3#newcode479
Line 479: // - Unless specified otherwise, instructions operate on 64bit
operands.
64-bit.

http://codereview.chromium.org/113767/diff/1/3#newcode486
Line 486: // NOTE ON INTERFACE: In this interface, we always use a Register when
Do you even need this NOTE?  It sounds like it just says: the IA32 interface was
ugly and this one is better.

[William Hesse, 2009-05-25 12:24:59]

http://codereview.chromium.org/113767/diff/1/3
File src/x64/assembler-x64.h (right):

http://codereview.chromium.org/113767/diff/1/3#newcode478
Line 478: // - Function names correspond one-to-one to ia32 instruction
mnemonics.
On 2009/05/25 11:45:04, Kevin Millikin wrote:
> ia32?

Done.

http://codereview.chromium.org/113767/diff/1/3#newcode479
Line 479: // - Unless specified otherwise, instructions operate on 64bit
operands.
On 2009/05/25 11:45:04, Kevin Millikin wrote:
> 64-bit.

Done.

http://codereview.chromium.org/113767/diff/1/3#newcode484
Line 484: // - Naming conflicts with C++ keywords are resolved via a trailing
'_'.
Made more explicit, dashes are removed, agrees with gcc assembler on suffixes.

http://codereview.chromium.org/113767/diff/1/3#newcode486
Line 486: // NOTE ON INTERFACE: In this interface, we always use a Register when
On 2009/05/25 11:45:04, Kevin Millikin wrote:
> Do you even need this NOTE?  It sounds like it just says: the IA32 interface
was
> ugly and this one is better.

Done.

[Lasse Reichstein, 2009-05-25 12:59:21]

LGTM.

http://codereview.chromium.org/113767/diff/1/2
File src/x64/assembler-x64.cc (right):

http://codereview.chromium.org/113767/diff/1/2#newcode37
Line 37: Register rsi = { 7 };
Will we have both eax and rax as registers? We will likely use both,
meaningfully, so the code is only really readable if they both exist. They might
just be aliases, though, but it would be nice to actually test that you use 32
bit registers with 32 bit opcodes.

http://codereview.chromium.org/113767/diff/1/3
File src/x64/assembler-x64.h (right):

http://codereview.chromium.org/113767/diff/1/3#newcode553
Line 553: void add(const Operand& dst, const Immediate& x);
For completeness, we will need the following combinarions:

Register, register
Register, Operand
Register, Immediate
Operand, register
Operand, Immediate

And we will need them for all four sizes (byte, word, dword, qword).
Are there any more cases that I have forgotten?

This seems like a lot (on the other hand, we have so far gotten by with only
creating the ones we actually use, so it might not be a problem).