[turbofan] add new x64 addressing modes

[turbofan] add new x64 addressing modes

R=bmeurer@chromium.org, titzer@chromium.org

BUG=

Committed: https://code.google.com/p/v8/source/detail?r=24267

[dcarney, 2014-09-25 14:54:26]

Patchset #1 (id:1) has been deleted

[dcarney, 2014-09-26 10:47:15]

dcarney@chromium.org changed reviewers:
+ titzer@chromium.org

[dcarney, 2014-09-26 10:47:45]

ptal - check out the approach, i'm adding unit tests now

[titzer, 2014-09-26 11:04:12]

https://codereview.chromium.org/605693002/diff/80001/src/compiler/x64/instruc...
File src/compiler/x64/instruction-selector-x64.cc (right):

https://codereview.chromium.org/605693002/diff/80001/src/compiler/x64/instruc...
src/compiler/x64/instruction-selector-x64.cc:77: if (opcode() !=
IrOpcode::kInt32Mul) return;
I think you can use an Int32BinopMatcher here. It will actually put the constant
on the right for you too!

https://codereview.chromium.org/605693002/diff/80001/src/compiler/x64/instruc...
src/compiler/x64/instruction-selector-x64.cc:105: //  [x + K]
This case is redundant since N can be 1.

https://codereview.chromium.org/605693002/diff/80001/src/compiler/x64/instruc...
src/compiler/x64/instruction-selector-x64.cc:123: if (opcode() ==
IrOpcode::kInt32Add) {
Here also.

https://codereview.chromium.org/605693002/diff/80001/src/compiler/x64/instruc...
src/compiler/x64/instruction-selector-x64.cc:145: class AddressingModeMatcher {
Is it possible to inline IndexAndDisplacement matcher into this class and make
that match routine a method here?

[dcarney, 2014-09-26 12:49:51]

https://codereview.chromium.org/605693002/diff/80001/src/compiler/x64/instruc...
File src/compiler/x64/instruction-selector-x64.cc (right):

https://codereview.chromium.org/605693002/diff/80001/src/compiler/x64/instruc...
src/compiler/x64/instruction-selector-x64.cc:77: if (opcode() !=
IrOpcode::kInt32Mul) return;
On 2014/09/26 11:04:11, titzer wrote:
> I think you can use an Int32BinopMatcher here. It will actually put the
constant
> on the right for you too!

Done.

https://codereview.chromium.org/605693002/diff/80001/src/compiler/x64/instruc...
src/compiler/x64/instruction-selector-x64.cc:105: //  [x + K]
On 2014/09/26 11:04:12, titzer wrote:
> This case is redundant since N can be 1.

yeah. but it's matched differently.  so i list it

https://codereview.chromium.org/605693002/diff/80001/src/compiler/x64/instruc...
src/compiler/x64/instruction-selector-x64.cc:123: if (opcode() ==
IrOpcode::kInt32Add) {
On 2014/09/26 11:04:12, titzer wrote:
> Here also.

Done.

https://codereview.chromium.org/605693002/diff/80001/src/compiler/x64/instruc...
src/compiler/x64/instruction-selector-x64.cc:145: class AddressingModeMatcher {
On 2014/09/26 11:04:11, titzer wrote:
> Is it possible to inline IndexAndDisplacement matcher into this class and make
> that match routine a method here?

yeah. but i want the 2 above matcher classes to be the arch independent bits,
and AddressingModeMatcher to be the copy and pasted bit

[Benedikt Meurer, 2014-09-29 07:44:51]

bmeurer@chromium.org changed reviewers:
+ bmeurer@chromium.org

[Benedikt Meurer, 2014-09-29 07:44:51]

https://codereview.chromium.org/605693002/diff/120001/src/compiler/x64/code-g...
File src/compiler/x64/code-generator-x64.cc (right):

https://codereview.chromium.org/605693002/diff/120001/src/compiler/x64/code-g...
src/compiler/x64/code-generator-x64.cc:148: static inline int NextOffset(int*
offset) {
Nit: remove inline.

https://codereview.chromium.org/605693002/diff/120001/src/compiler/x64/code-g...
src/compiler/x64/code-generator-x64.cc:154: static inline ScaleFactor
ScaleFor(AddressingMode one, AddressingMode mode) {
Nit: remove inline.

https://codereview.chromium.org/605693002/diff/120001/src/compiler/x64/instru...
File src/compiler/x64/instruction-selector-x64.cc (right):

https://codereview.chromium.org/605693002/diff/120001/src/compiler/x64/instru...
src/compiler/x64/instruction-selector-x64.cc:65: bool matches() { return left_
!= NULL; }
Nit: Matches

https://codereview.chromium.org/605693002/diff/120001/src/compiler/x64/instru...
src/compiler/x64/instruction-selector-x64.cc:66: int power() {
Nit: Power

https://codereview.chromium.org/605693002/diff/120001/src/compiler/x64/instru...
src/compiler/x64/instruction-selector-x64.cc:70: Node* left() {
Nit: Left

[dcarney, 2014-09-29 07:56:14]

https://codereview.chromium.org/605693002/diff/120001/src/compiler/x64/code-g...
File src/compiler/x64/code-generator-x64.cc (right):

https://codereview.chromium.org/605693002/diff/120001/src/compiler/x64/code-g...
src/compiler/x64/code-generator-x64.cc:148: static inline int NextOffset(int*
offset) {
On 2014/09/29 07:44:51, Benedikt Meurer wrote:
> Nit: remove inline.

Done.

https://codereview.chromium.org/605693002/diff/120001/src/compiler/x64/code-g...
src/compiler/x64/code-generator-x64.cc:154: static inline ScaleFactor
ScaleFor(AddressingMode one, AddressingMode mode) {
On 2014/09/29 07:44:51, Benedikt Meurer wrote:
> Nit: remove inline.

Done.

https://codereview.chromium.org/605693002/diff/120001/src/compiler/x64/instru...
File src/compiler/x64/instruction-selector-x64.cc (right):

https://codereview.chromium.org/605693002/diff/120001/src/compiler/x64/instru...
src/compiler/x64/instruction-selector-x64.cc:65: bool matches() { return left_
!= NULL; }
On 2014/09/29 07:44:51, Benedikt Meurer wrote:
> Nit: Matches

Done.

https://codereview.chromium.org/605693002/diff/120001/src/compiler/x64/instru...
src/compiler/x64/instruction-selector-x64.cc:66: int power() {
On 2014/09/29 07:44:51, Benedikt Meurer wrote:
> Nit: Power

Done.

https://codereview.chromium.org/605693002/diff/120001/src/compiler/x64/instru...
src/compiler/x64/instruction-selector-x64.cc:70: Node* left() {
On 2014/09/29 07:44:51, Benedikt Meurer wrote:
> Nit: Left

Done.

[Benedikt Meurer, 2014-09-29 07:56:28]

lgtm

[dcarney, 2014-09-29 08:11:14]

Committed patchset #8 (id:160001) manually as 24267 (presubmit successful).