[FFmpeg-devel] [PATCH] tests/fate/filter-audio: convert atempo test to oneoff

[James Almer]

On 7/5/2024 2:18 PM, Martin Storsjö wrote:
> On Fri, 5 Jul 2024, James Almer wrote:
> 
>> On 7/5/2024 2:38 AM, Anton Khirnov wrote:
>>> Quoting James Almer (2024-07-04 22:45:28)
>>>> On 7/4/2024 4:04 PM, Anton Khirnov wrote:
>>>>> Filter output is not bitexact.
>>>>> ---
>>>>> Reference file at https://up.khirnov.net/7r.pcm, please put it in
>>>>> filter-reference/atempo.pcm
>>>>
>>>> How did you create it? x86_32 uses x87 floats which are a lot more
>>>> precise than sse ones, for example, so it's best to create a ref file
>>>> using such a build.
>>>
>>> Does it matter when the result is s16 anyway?
>>
>> Eh, who knows. Just in case i generated it on x86_32 with -cpuflags 0, 
>> and uploaded it. Confirm it's fine on your end too, otherwise I'll 
>> replace it with your file.
> 
> The sample you uploaded seems to work fine for me, on aarch64 with 
> clang, where the test was failing before.
> 
> FWIW, re x86_32 and x87 - some compilers default to SSE2 math even for 
> x86_32 targets these days, so depending on how you build, you may still 
> get similar behaviour as to x86_64.

I used gcc 14 and -march=i686 at configure time, so it should not have 
sse enabled.