On Wednesday 18 August 2010 16:13:01 Arjan van de Ven wrote:
> On 8/18/2010 6:30 AM, Rémi Denis-Courmont wrote:
> > The only portable manner to establish NEON support is the __ARM_NEON__
> > predefined constant, just like __SSE3__ implies SSE3 support. Obviously,
> > that simply reflects the compiler settings. If you want to make the
> > decision at run- time, you're going to need operating system-specific
> > code.
> 
> it's more complex, at least for __SSE3__
> you won't get __SSE3__ set unless you tell the compiler you support this
> instruction, and at that point, the compiler
> will automatically use the instructions as well.

That's exactly the same as the Neon support.

I wish there was a way to use the intrinsics specifically for a portion of the 
code.

I've been experimenting with using SIMD for string operations to benefit of 
higher bandwidth. But strings are MUCH shorter than images, so the benefit of 
using SIMD is only in the order of 15-30%. That benefit is quickly offset by 
the 
runtime-dispatching techniques because the code can't be inlined.

-- 
Thiago Macieira - thiago (AT) macieira.info - thiago (AT) kde.org
  Senior Product Manager - Nokia, Qt Development Frameworks
      PGP/GPG: 0x6EF45358; fingerprint:
      E067 918B B660 DBD1 105C  966C 33F5 F005 6EF4 5358

Attachment: signature.asc
Description: This is a digitally signed message part.

_______________________________________________
MeeGo-dev mailing list
[email protected]
http://lists.meego.com/listinfo/meego-dev

Reply via email to