Pass correct flags for use_lto==1 with Clang.

Pass correct flags for use_lto==1 with Clang.

BUG=453195
R=thakis@chromium.org
CC=fdegans@chromium.org

Committed: https://crrev.com/836b0aaeae4a2984c9f0c10c69fbe9627b15a542
Cr-Commit-Position: refs/heads/master@{#313770}

[Nico, 2015-01-29 18:48:01]

one question about avx below, otherwise lgtm.

https://codereview.chromium.org/884923002/diff/1/build/common.gypi
File build/common.gypi (right):

https://codereview.chromium.org/884923002/diff/1/build/common.gypi#newcode2989
build/common.gypi:2989: 'ldflags+': ['-Wl,-plugin-opt,mcpu=corei7-avx'],
nit: i think you don't need the +, gyp merges lists by default

Isn't avx fairly new? I think we require cpus with SSE2, but not avx.

https://codereview.chromium.org/884923002/diff/1/build/common.gypi#newcode5821
build/common.gypi:5821: '-flto',
i'd be nice if gcc's and clang's flag here were compatible :-/

[pcc, 2015-01-29 19:11:13]

https://codereview.chromium.org/884923002/diff/1/build/common.gypi
File build/common.gypi (right):

https://codereview.chromium.org/884923002/diff/1/build/common.gypi#newcode2989
build/common.gypi:2989: 'ldflags+': ['-Wl,-plugin-opt,mcpu=corei7-avx'],
On 2015/01/29 18:48:01, Nico wrote:
> nit: i think you don't need the +, gyp merges lists by default

Done.

> Isn't avx fairly new? I think we require cpus with SSE2, but not avx.

I'm not sure if there is a way to avoid this at the moment, as I've found that
LLVM LTO currently compiles all functions in the program for the same
sub-target. As I recall, LLVM has a concept of function-specific codegen
attributes to support this exact scenario, but for some reason it doesn't help
here. Eventually once this is fixed we should be able to remove this flag.

[pcc, 2015-01-29 19:12:08]

The CQ bit was checked by pcc@google.com

[commit-bot: I haz the power, 2015-01-29 19:13:20]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/884923002/20001

[Nico, 2015-01-29 19:14:15]

ok, but add a comment that says that this flag needs to go away or be tweaked
eventually (you can do that in a follow-up that you tbr against me if you don't
want to uncheck cq)

[commit-bot: I haz the power, 2015-01-29 20:09:13]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-01-29 20:11:36]

Patchset 2 (id:??) landed as
https://crrev.com/836b0aaeae4a2984c9f0c10c69fbe9627b15a542
Cr-Commit-Position: refs/heads/master@{#313770}