write to cpu_flags once

write to cpu_flags once

To make init cpu flags thread safe, there can only be one write to the variable.

R=richard.winterton@intel.com, harryjin@google.com
BUG=libyuv:508
Committed: https://chromium.googlesource.com/libyuv/libyuv/+/9be6d21ae726f66bc96502c2498866d41dcce629

[richard.winterton, 2015-10-20 21:02:23]

On 2015/10/20 17:37:29, fbarchard1 wrote:

So I am not sure how this is used for Linux 32 bit but if it is then I am
assuming the EBX is saved for PIC if used.  Other than that I think the code
looks right assuming all the constants are correct which they were before and
look right still.  Compared some code to what Max published: 
https://software.intel.com/en-us/articles/how-to-detect-new-instruction-suppo...

[fbarchard1, 2015-10-20 21:52:20]

https://codereview.chromium.org/1412793006/diff/1/source/cpu_id.cc
File source/cpu_id.cc (right):

https://codereview.chromium.org/1412793006/diff/1/source/cpu_id.cc#newcode210
source/cpu_id.cc:210: cpu_info = ((cpu_info1[3] & 0x04000000) ? kCpuHasSSE2 : 0)
|
indent wrong.

https://codereview.chromium.org/1412793006/diff/1/source/cpu_id.cc#newcode223
source/cpu_id.cc:223: kCpuHasAVX;
ditto indent.

[fbarchard1, 2015-10-20 22:10:13]

fbarchard@google.com changed reviewers:
+ harryjin@chromium.org, richard.winterton@intel.com
- harryjin@google.com

[fbarchard1, 2015-10-20 22:10:14]

ebx needs to be preserved in the assembly block with gcc with fpic.  It needs to
be preserved at the function level for all x86/x64, but the compiler can figure
that out as long as its not fpic.

https://codereview.chromium.org/1412793006/diff/1/source/cpu_id.cc
File source/cpu_id.cc (right):

https://codereview.chromium.org/1412793006/diff/1/source/cpu_id.cc#newcode210
source/cpu_id.cc:210: cpu_info = ((cpu_info1[3] & 0x04000000) ? kCpuHasSSE2 : 0)
|
On 2015/10/20 21:52:20, fbarchard1 wrote:
> indent wrong.

Done.

https://codereview.chromium.org/1412793006/diff/1/source/cpu_id.cc#newcode223
source/cpu_id.cc:223: kCpuHasAVX;
On 2015/10/20 21:52:20, fbarchard1 wrote:
> ditto indent.

Done.

[richard.winterton, 2015-10-20 22:57:23]

On 2015/10/20 22:10:14, fbarchard1 wrote:
> ebx needs to be preserved in the assembly block with gcc with fpic.  It needs
to
> be preserved at the function level for all x86/x64, but the compiler can
figure
> that out as long as its not fpic.
> 
> https://codereview.chromium.org/1412793006/diff/1/source/cpu_id.cc
> File source/cpu_id.cc (right):
> 
> https://codereview.chromium.org/1412793006/diff/1/source/cpu_id.cc#newcode210
> source/cpu_id.cc:210: cpu_info = ((cpu_info1[3] & 0x04000000) ? kCpuHasSSE2 :
0)
> |
> On 2015/10/20 21:52:20, fbarchard1 wrote:
> > indent wrong.
> 
> Done.
> 
> https://codereview.chromium.org/1412793006/diff/1/source/cpu_id.cc#newcode223
> source/cpu_id.cc:223: kCpuHasAVX;
> On 2015/10/20 21:52:20, fbarchard1 wrote:
> > ditto indent.
> 
> Done.

lgtm

[fbarchard1, 2015-10-20 23:24:04]

Committed patchset #3 (id:40001) manually as
9be6d21ae726f66bc96502c2498866d41dcce629 (presubmit successful).