[FFmpeg-devel] [PATCH v5] gcc: Don't disable '-ftree-vectorize' if gcc version higher than 13.

[Martin Storsjö]

On Sun, 15 Jun 2025, Michael Niedermayer wrote:

> Hi Martin
>
> On Fri, Jun 13, 2025 at 11:42:11AM +0300, Martin Storsjö wrote:
>> On Fri, 13 Jun 2025, Jiawei wrote:
>>
>>> This patch modifies the FFmpeg build system to allow GCC to use the
>>> `-ftree-vectorize` flag when the compiler version is 13 or newer.
>>> Enabling this flag can improve performance through better loop analysis
>>> and auto-vectorization (SIMD) opportunities in modern GCC versions.
>>>
>>> The explicit `-fno-tree-vectorize` flag was originally added in commit
>>> 973859f5230e (2009). A previous attempt to enable `-ftree-vectorize`
>>> was made in commit cb8646af24bd (2016), but it was reverted in
>>> fd6dbc53855f. The reason for the revert was that the inline x86 CABAC
>>> assembly code caused compiler errors ??? the compiler would run out of
>>> available registers during vectorization, making it unable to compile
>>> some functions.
>>>
>>> In commit 182663a58a7a (2023), the problematic CABAC function was made
>>> non-inline. This significantly reduces the risk of register exhaustion
>>> caused by inlining large assembly blocks, making the vectorizer safer
>>> to enable for other functions.
>>>
>>> Given improvements in GCC's vectorizer and the mitigation of the
>>> original issue, we now re-enable `-ftree-vectorize` for GCC version 13
>>> and above.
>>>
>>> Version log:
>>>  Only allow GCC versions >= 13 to use -ftree-vectorize.
>>>  Modify the commit description.
>>>  Add historical context and rationale for the change.
>>>
>>> Disscussion see:
>>> https://patchwork.ffmpeg.org/project/ffmpeg/patch/20250521061750.54882-1-jiawei@iscas.ac.cn/
>>>
>>>
>>> Signed-off-by: Jiawei <jiawei at iscas.ac.cn>
>>> ---
>>> configure | 6 +++++-
>>> 1 file changed, 5 insertions(+), 1 deletion(-)
>>
>> Thanks; I'm ok with this version. Whoever applies it can touch up the commit
>> message to remove the ???? word (maybe just replace it with an "as"), it
>> sounds reasonably right like that.
>
> the patch seems not to be on the mailing list
> I have v4 in my inbox but not v5, so if the patch is ok, maybe you are
> the only one who can apply it
>
> Maybe Jiawei is not subscribed to the ML so his mails would go to the
> moderation que (i didnt check)

That's possible; I don't see v3 or any of the newer versions on the 
mailing list archive.

Content-wise, the patch is the same since v2 I think, it's just been work 
on the commit message.

But I guess I can repost a version that I would be ok with getting 
applied.

// Martin