[FFmpeg-devel] [PATCH 3/3] Use DSPContext.vector_fmul() and DSPContext.vector_fmul_reverse() in floating-point version of apply_window(). 46% faster in function apply_window().

[Justin Ruggles]

On 01/05/2011 04:06 PM, Loren Merritt wrote:

> On Tue, 4 Jan 2011, Justin Ruggles wrote:
> 
>> Currently we have vector_fmul() for: C, neon, vfp, altivec, 3dnow, sse
>>
>> I implemented vector_fmul_copy() for C, altivec, 3dnow, and sse to use 2
>> src and 1 dst. The Altivec version of vector_fmul_copy() has not been
>> tested, but I implemented it in the hope that someone else will test and
>> review it.  Here are some benchmarks on my Athlon64. benchmark numbers
>> are in dezicycles.
>>
>> I also tried to rewrite the current C version in SSE.  It was faster
>> than the fmul_copy+fmul_reverse since it basically merges the 2 loops,
>> but it was slower than vector_fmul_copy(512).  I left that out of the
>> patch.  If anyone is interested I can send it...
> 
> I predict that all of the vector_fmul_* mentioned here are memory-bound on 
> intel and arithmetic-bound on amd.
> 
> Is there any reason to keep both the 2-arg and 3-arg version of 
> vector_fmul?


I tested using vector_fmul_copy with same value for src0 and dst and it
ended up being slower.  I thought it was weird, so I kept both versions.
 Maybe I did something wrong in my tests though...

Also, I'll try benchmarking these on my laptop (Intel Atom 330, 64-bit
Ubuntu).

-Justin