[FFmpeg-devel] [PATCH 0/3] showcqt x86 optimization using intrinsic

[Muhammad Faiz]

On Thu, Mar 10, 2016 at 5:18 PM, Hendrik Leppkes <h.leppkes at gmail.com> wrote:
> On Thu, Mar 10, 2016 at 11:08 AM, Muhammad Faiz <mfcc64 at gmail.com> wrote:
>> On Thu, Mar 10, 2016 at 5:01 PM, Hendrik Leppkes <h.leppkes at gmail.com> wrote:
>>> On Thu, Mar 10, 2016 at 10:53 AM, Muhammad Faiz <mfcc64 at gmail.com> wrote:
>>>> I use intrinsic because writing asm using nasm or inline asm
>>>> is difficult task for me.
>>>>
>>>
>>> We have intentionally avoided intrinsics so far, because the result is
>>> extremely compiler dependent, and I don't think we should just start
>>
>> What does the result means. Code correctness or performance?
>>
>
> Performance and code complexity, not to mention readability and
> possibly portability.

For performance, I tolerate if intrinsic is slower than asm/inline asm,
or vary between compiler.

> For a little extra effort, you get a solution thats consistent in
> performance, generally easier to read since its a style everyone here
> is used to, and proven to work on every build configuration we have.
>

In my opinion, intrinsic is more readable.

> IMHO its totally worth learning how to do it, over just shoehorning
> some intrinsics in "because its easier".
>

It will need time. But probably I will try it.

Thank's