Improvements in lithium code generation. Recognizing if some operands are constants, we can often s…

Improvements in lithium code generation. Recognizing if some operands are constants, we can often save on registers and instructions.

BUG=

Committed: https://code.google.com/p/v8/source/detail?r=14364

[mvstanton, 2013-04-18 09:29:23]

Hi, many of these changes showed up mistakenly as "unrelated" in a previous
review :). I've addressed the comments you offered there, nonetheless.
All the best,
--Michael

[danno, 2013-04-18 15:37:54]

lgtm with comments

https://codereview.chromium.org/14022011/diff/2001/src/ia32/lithium-codegen-i...
File src/ia32/lithium-codegen-ia32.cc (right):

https://codereview.chromium.org/14022011/diff/2001/src/ia32/lithium-codegen-i...
src/ia32/lithium-codegen-ia32.cc:4485: __ mov(operand, Immediate(const_value));
Shouldn't you store as a SMI?

https://codereview.chromium.org/14022011/diff/2001/src/ia32/lithium-ia32.cc
File src/ia32/lithium-ia32.cc (right):

https://codereview.chromium.org/14022011/diff/2001/src/ia32/lithium-ia32.cc#n...
src/ia32/lithium-ia32.cc:2422: val = UseRegister(instr->value());
Can't you just move this to the else of the above clause?

[mvstanton, 2013-04-19 16:53:28]

Thanks for the look and SMI-convert catch!

https://codereview.chromium.org/14022011/diff/2001/src/ia32/lithium-codegen-i...
File src/ia32/lithium-codegen-ia32.cc (right):

https://codereview.chromium.org/14022011/diff/2001/src/ia32/lithium-codegen-i...
src/ia32/lithium-codegen-ia32.cc:4485: __ mov(operand, Immediate(const_value));
On 2013/04/18 15:37:54, danno wrote:
> Shouldn't you store as a SMI?

Good catch, thanks. I'm wondering why a test didn't see this.

https://codereview.chromium.org/14022011/diff/2001/src/ia32/lithium-ia32.cc
File src/ia32/lithium-ia32.cc (right):

https://codereview.chromium.org/14022011/diff/2001/src/ia32/lithium-ia32.cc#n...
src/ia32/lithium-ia32.cc:2422: val = UseRegister(instr->value());
On 2013/04/18 15:37:54, danno wrote:
> Can't you just move this to the else of the above clause?

I reworked this to be less awkward, I think.

[mvstanton, 2013-04-22 07:47:33]

Committed patchset #3 manually as r14364 (presubmit successful).