diff mbox series

[FFmpeg-devel] h274: remove optimization pragma

Message ID MhylPZz--3-2@lynne.ee
State Accepted
Commit 033105a73901cf9ecfa6d410e96d7f347dc69c71
Headers show
Series [FFmpeg-devel] h274: remove optimization pragma | expand

Checks

Context Check Description
andriy/make_x86 success Make finished
andriy/make_fate_x86 success Make fate finished
andriy/make_ppc success Make finished
andriy/make_fate_ppc success Make fate finished

Commit Message

Lynne Aug. 25, 2021, 7:24 p.m. UTC 
This results in warnings on compilers which don't support it, 
objections were raised during the review process about it but went unnoticed,
and the speed benefit is highly compiler and version specific, and
also not very critical.

We generally hand-write assembly to optimize loops like that, rather
than use compiler magic, and for 40% best case scenario, it's simply
not worth it.

Plus, tree vectorization is still problematic with GCC and disabled by default
for a good reason, so enabling it locally is sketchy.

Patch attached.
Subject: [PATCH] h274: remove optimization pragma

This results in warnings on compilers which don't support it,
objections were raised during the review process about it but went unnoticed,
and the speed benefit is highly compiler and version specific, and
also not very critical.

We generally hand-write assembly to optimize loops like that, rather
than use compiler magic, and for 40% best case scenario, it's simply
not worth it.

Plus, tree vectorization is still problematic with GCC and disabled by default
for a good reason, so enabling it locally is sketchy.
---
 libavcodec/h274.c | 4 ----
 1 file changed, 4 deletions(-)

Comments

Jan Poonthong Aug. 26, 2021, 3:13 a.m. UTC | #1 
I didn't really understand what you meant. So I should install nasm and run
./configure or just ./configure --disable-x86asm?

On Thu, Aug 26, 2021 at 2:24 AM Lynne <dev@lynne.ee> wrote:

> This results in warnings on compilers which don't support it,
> objections were raised during the review process about it but went
> unnoticed,
> and the speed benefit is highly compiler and version specific, and
> also not very critical.
>
> We generally hand-write assembly to optimize loops like that, rather
> than use compiler magic, and for 40% best case scenario, it's simply
> not worth it.
>
> Plus, tree vectorization is still problematic with GCC and disabled by
> default
> for a good reason, so enabling it locally is sketchy.
>
> Patch attached.
>
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel@ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
>
Paul B Mahol Aug. 26, 2021, 7:11 a.m. UTC | #2 
LGTM
diff mbox series

Patch

diff --git a/libavcodec/h274.c b/libavcodec/h274.c
index 0efc00ca1d..5e2cf150ea 100644
--- a/libavcodec/h274.c
+++ b/libavcodec/h274.c
@@ -30,10 +30,6 @@ 
 
 #include "h274.h"
 
-// The code in this file has a lot of loops that vectorize very well, this is
-// about a 40% speedup for no obvious downside.
-#pragma GCC optimize("tree-vectorize")
-
 static const int8_t Gaussian_LUT[2048+256];
 static const uint32_t Seed_LUT[256];
 static const int8_t R64T[64][64];