[turbofan] x64 lea multiplication matching

[turbofan] x64 lea multiplication matching

R=bmeurer@chromium.org

BUG=

Committed: https://code.google.com/p/v8/source/detail?r=24317

[dcarney, 2014-09-29 13:13:52]

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

[dcarney, 2014-09-29 13:14:00]

ptal

[Benedikt Meurer, 2014-09-30 04:55:05]

https://codereview.chromium.org/615483003/diff/20001/src/compiler/node-matche...
File src/compiler/node-matchers.h (right):

https://codereview.chromium.org/615483003/diff/20001/src/compiler/node-matche...
src/compiler/node-matchers.h:215: bool Matches() { return left_ != NULL; }
Nit: const

https://codereview.chromium.org/615483003/diff/20001/src/compiler/node-matche...
src/compiler/node-matchers.h:216: int Power() {
Nit: const

https://codereview.chromium.org/615483003/diff/20001/src/compiler/node-matche...
src/compiler/node-matchers.h:220: Node* Left() {
Nit: const

https://codereview.chromium.org/615483003/diff/20001/src/compiler/node-matche...
src/compiler/node-matchers.h:225: int32_t Displacement() {
Nit: const

https://codereview.chromium.org/615483003/diff/20001/src/compiler/node-matche...
src/compiler/node-matchers.h:231: void Match();
Hm, those Match method(s) could be inlined into the constructor, and the
constructor should reside in the .cc file.

https://codereview.chromium.org/615483003/diff/20001/src/compiler/x64/instruc...
File src/compiler/x64/instruction-codes-x64.h (right):

https://codereview.chromium.org/615483003/diff/20001/src/compiler/x64/instruc...
src/compiler/x64/instruction-codes-x64.h:97: V(MX1)  /* [%r1 + %r1*1    ] */    
 \
Why do we need these additional modes? MXn is just a special case of MRn, so
it's basically redundant or am I missing something?

https://codereview.chromium.org/615483003/diff/20001/src/compiler/x64/instruc...
File src/compiler/x64/instruction-selector-x64-unittest.cc (right):

https://codereview.chromium.org/615483003/diff/20001/src/compiler/x64/instruc...
src/compiler/x64/instruction-selector-x64-unittest.cc:312: static const
MultParam kMultParams[] = {{-1, false, kMode_None},
No need for static here, it's already inside an anonymous namespace.

[dcarney, 2014-09-30 08:03:22]

https://codereview.chromium.org/615483003/diff/20001/src/compiler/node-matche...
File src/compiler/node-matchers.h (right):

https://codereview.chromium.org/615483003/diff/20001/src/compiler/node-matche...
src/compiler/node-matchers.h:215: bool Matches() { return left_ != NULL; }
On 2014/09/30 04:55:04, Benedikt Meurer wrote:
> Nit: const

Done.

https://codereview.chromium.org/615483003/diff/20001/src/compiler/node-matche...
src/compiler/node-matchers.h:216: int Power() {
On 2014/09/30 04:55:04, Benedikt Meurer wrote:
> Nit: const

Done.

https://codereview.chromium.org/615483003/diff/20001/src/compiler/node-matche...
src/compiler/node-matchers.h:220: Node* Left() {
On 2014/09/30 04:55:04, Benedikt Meurer wrote:
> Nit: const

Done.

https://codereview.chromium.org/615483003/diff/20001/src/compiler/node-matche...
src/compiler/node-matchers.h:225: int32_t Displacement() {
On 2014/09/30 04:55:04, Benedikt Meurer wrote:
> Nit: const

Done.

https://codereview.chromium.org/615483003/diff/20001/src/compiler/node-matche...
src/compiler/node-matchers.h:231: void Match();
On 2014/09/30 04:55:04, Benedikt Meurer wrote:
> Hm, those Match method(s) could be inlined into the constructor, and the
> constructor should reside in the .cc file.

Done.

https://codereview.chromium.org/615483003/diff/20001/src/compiler/x64/instruc...
File src/compiler/x64/instruction-codes-x64.h (right):

https://codereview.chromium.org/615483003/diff/20001/src/compiler/x64/instruc...
src/compiler/x64/instruction-codes-x64.h:97: V(MX1)  /* [%r1 + %r1*1    ] */    
 \
On 2014/09/30 04:55:04, Benedikt Meurer wrote:
> Why do we need these additional modes? MXn is just a special case of MRn, so
> it's basically redundant or am I missing something?

The number of operands to generated instructions are different.  I wasn't sure
if we had a better way to encode this.

[Benedikt Meurer, 2014-09-30 09:14:21]

https://codereview.chromium.org/615483003/diff/20001/src/compiler/x64/instruc...
File src/compiler/x64/instruction-codes-x64.h (right):

https://codereview.chromium.org/615483003/diff/20001/src/compiler/x64/instruc...
src/compiler/x64/instruction-codes-x64.h:97: V(MX1)  /* [%r1 + %r1*1    ] */    
 \
Well, just specify the same operand twice, that does the trick.

[dcarney, 2014-09-30 09:26:39]

On 2014/09/30 09:14:21, Benedikt Meurer wrote:
>
https://codereview.chromium.org/615483003/diff/20001/src/compiler/x64/instruc...
> File src/compiler/x64/instruction-codes-x64.h (right):
> 
>
https://codereview.chromium.org/615483003/diff/20001/src/compiler/x64/instruc...
> src/compiler/x64/instruction-codes-x64.h:97: V(MX1)  /* [%r1 + %r1*1    ] */  
 
>  \
> Well, just specify the same operand twice, that does the trick.

errr, right.  Done.

[Benedikt Meurer, 2014-09-30 09:28:31]

LGTM

[dcarney, 2014-09-30 09:46:43]

Committed patchset #5 (id:80001) manually as 24317 (presubmit successful).