to enable chromium x86 android build which doesn't support SSSE3

to enable chromium x86 android build which doesn't support SSSE3

[Wei James, 2012-04-26 07:31:57]

In case SSSE3 not supported, reference SSSE3 function will results in undefined
error when building. 

this patch is to enable chromium x86 android build, which doesn't support SSSE3
but support SSE2.

[TomH, 2012-04-26 10:17:46]

I don't understand the purpose of this patch.
If there's an Android platform that supports SSE2 but not SSSE3, hasSSE2()
should be returning true and hasSSSE3() should be returning false.

If your #ifdef is truly necessary, that argues our feature detection code is
broken, and we should be fixing that rather than patching like this.

What actual platform requires this?

[Wei James, 2012-04-26 11:55:04]

On 2012/04/26 10:17:46, TomH wrote:
> I don't understand the purpose of this patch.
> If there's an Android platform that supports SSE2 but not SSSE3, hasSSE2()
> should be returning true and hasSSSE3() should be returning false.
> 
> If your #ifdef is truly necessary, that argues our feature detection code is
> broken, and we should be fixing that rather than patching like this.
> 
> What actual platform requires this?

tom, thanks a lot for your review. 

the problem is happened when compiling time, not when running. 
As no SSSE3 functions defined, so although hasSSSE3() will return false, the
issue happens when compiling for S32_alpha_D32_filter_DX_SSSE3 etc.  

as for the platform, with current Intel Android NDK:
$
/usr/local/google/android-ndk-r7/toolchains/x86-4.4.3/prebuilt/linux-x86/bin/i686-android-linux-gcc
-dM -E - < /dev/null |grep SSE
#define __SSE2_MATH__ 1
#define __SSE_MATH__ 1
#define __SSE2__ 1
#define __SSE__ 1
#define __SSE3__ 1
 
we can see that SSSE3 not supported while SSE2 supported.

[Wei James, 2012-04-26 11:57:03]

this patch is to work together with
https://chromiumcodereview.appspot.com/10198027/ to fix
http://code.google.com/p/chromium/issues/detail?id=125160

[TomH, 2012-04-26 13:01:57]

Oops, you're right. This is what we get for lumping together multiple
optimization types in the same file.

LGTM.
I'm out of the office today to give a conference presentation, but should be
able to land this patch tomorrow.

[Wei James, 2012-04-26 13:34:01]

On 2012/04/26 13:01:57, TomH wrote:
> Oops, you're right. This is what we get for lumping together multiple
> optimization types in the same file.
> 
> LGTM.
> I'm out of the office today to give a conference presentation, but should be
> able to land this patch tomorrow.

that's great. thanks a lot for the review and help to land it. thanks

[TomH, 2012-04-27 12:53:21]

Committed in r3770.
Please close the issue when you get a chance.

[Wei James, 2012-04-27 13:39:15]

On 2012/04/27 12:53:21, TomH wrote:
> Committed in r3770.
> Please close the issue when you get a chance.

thanks a lot.