From: mkver via ffmpeg-devel <ffmpeg-devel@ffmpeg.org>
To: ffmpeg-devel@ffmpeg.org
Cc: mkver <code@ffmpeg.org>
Subject: [FFmpeg-devel] [PATCH] avcodec/x86/hpeldsp: Don't use saturated addition when unnecessary (PR #20791)
Date: Thu, 30 Oct 2025 11:02:06 -0000
Message-ID: <176182212706.81.6416319141458007086@7d278768979e> (raw)
PR #20791 opened by mkver
URL: https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/20791
Patch URL: https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/20791.patch
>From 50f2e0e7ba41e4aedf36244d63c42a1381fc0336 Mon Sep 17 00:00:00 2001
From: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Date: Thu, 30 Oct 2025 10:27:00 +0100
Subject: [PATCH 1/3] avcodec/x86/hpeldsp: Actually use constants in registers
Forgotten in 36f92206bb90d6f0268749bd6fe6aa57974442db.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
---
libavcodec/x86/hpeldsp.asm | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/libavcodec/x86/hpeldsp.asm b/libavcodec/x86/hpeldsp.asm
index 2587e3c315..0974286b0d 100644
--- a/libavcodec/x86/hpeldsp.asm
+++ b/libavcodec/x86/hpeldsp.asm
@@ -428,7 +428,7 @@ cglobal %1%3_pixels8_xy2, 4,5,5
psrlw m2, 2
%else
paddusw m2, m0
- pmulhrsw m2, [pw_8192]
+ pmulhrsw m2, m3
%endif
%ifidn %1, avg
movh m1, [r0+r4]
@@ -450,7 +450,7 @@ cglobal %1%3_pixels8_xy2, 4,5,5
psrlw m0, 2
%else
paddusw m0, m2
- pmulhrsw m0, [pw_8192]
+ pmulhrsw m0, m3
%endif
%ifidn %1, avg
movh m1, [r0+r4]
--
2.49.1
>From a84ea10f93fbb66530eaa5ebb6f0275203d18356 Mon Sep 17 00:00:00 2001
From: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Date: Thu, 30 Oct 2025 10:44:41 +0100
Subject: [PATCH 2/3] avcodec/x86/hpeldsp: Don't use saturated addition when
unnecessary
The numbers here are small (sums of values unpacked from bytes).
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
---
libavcodec/x86/hpeldsp.asm | 48 +++++++++++++++++++-------------------
1 file changed, 24 insertions(+), 24 deletions(-)
diff --git a/libavcodec/x86/hpeldsp.asm b/libavcodec/x86/hpeldsp.asm
index 0974286b0d..c92c70f5ad 100644
--- a/libavcodec/x86/hpeldsp.asm
+++ b/libavcodec/x86/hpeldsp.asm
@@ -423,11 +423,11 @@ cglobal %1%3_pixels8_xy2, 4,5,5
punpcklbw m0, m1
pmaddubsw m0, m4
%ifidn %3, _no_rnd
- paddusw m2, m3
- paddusw m2, m0
+ paddw m2, m3
+ paddw m2, m0
psrlw m2, 2
%else
- paddusw m2, m0
+ paddw m2, m0
pmulhrsw m2, m3
%endif
%ifidn %1, avg
@@ -445,11 +445,11 @@ cglobal %1%3_pixels8_xy2, 4,5,5
punpcklbw m2, m1
pmaddubsw m2, m4
%ifidn %3, _no_rnd
- paddusw m0, m3
- paddusw m0, m2
+ paddw m0, m3
+ paddw m0, m2
psrlw m0, 2
%else
- paddusw m0, m2
+ paddw m0, m2
pmulhrsw m0, m3
%endif
%ifidn %1, avg
@@ -485,8 +485,8 @@ cglobal %1%3_pixels16_xy2, 4,5,8
punpcklbw m4, m7
punpckhbw m1, m7
punpckhbw m5, m7
- paddusw m4, m0
- paddusw m5, m1
+ paddw m4, m0
+ paddw m5, m1
xor r4, r4
add r1, r2
.loop:
@@ -498,12 +498,12 @@ cglobal %1%3_pixels16_xy2, 4,5,8
punpcklbw m2, m7
punpckhbw m1, m7
punpckhbw m3, m7
- paddusw m0, m2
- paddusw m1, m3
- paddusw m4, m6
- paddusw m5, m6
- paddusw m4, m0
- paddusw m5, m1
+ paddw m0, m2
+ paddw m1, m3
+ paddw m4, m6
+ paddw m5, m6
+ paddw m4, m0
+ paddw m5, m1
psrlw m4, 2
psrlw m5, 2
%ifidn %1, avg
@@ -524,12 +524,12 @@ cglobal %1%3_pixels16_xy2, 4,5,8
punpcklbw m4, m7
punpckhbw m3, m7
punpckhbw m5, m7
- paddusw m4, m2
- paddusw m5, m3
- paddusw m0, m6
- paddusw m1, m6
- paddusw m0, m4
- paddusw m1, m5
+ paddw m4, m2
+ paddw m5, m3
+ paddw m0, m6
+ paddw m1, m6
+ paddw m0, m4
+ paddw m1, m5
psrlw m0, 2
psrlw m1, 2
%ifidn %1, avg
@@ -567,8 +567,8 @@ cglobal %1_pixels16_xy2, 4,5,%2
movu m3, [r1+r4+1]
pmaddubsw m2, m5
pmaddubsw m3, m5
- paddusw m0, m2
- paddusw m1, m3
+ paddw m0, m2
+ paddw m1, m3
pmulhrsw m0, [pw_8192]
pmulhrsw m1, [pw_8192]
%ifidn %1, avg
@@ -587,8 +587,8 @@ cglobal %1_pixels16_xy2, 4,5,%2
movu m1, [r1+r4+1]
pmaddubsw m0, m5
pmaddubsw m1, m5
- paddusw m2, m0
- paddusw m3, m1
+ paddw m2, m0
+ paddw m3, m1
pmulhrsw m2, [pw_8192]
pmulhrsw m3, [pw_8192]
%ifidn %1, avg
--
2.49.1
>From 88f4641db2d488308f04b70cba9f285d30da6eb5 Mon Sep 17 00:00:00 2001
From: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Date: Thu, 30 Oct 2025 11:07:43 +0100
Subject: [PATCH 3/3] avcodec/x86/hpeldsp: Don't use PAVGB macro
It was only needed for MMX and there are no MMX functions here any more.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
---
libavcodec/x86/hpeldsp.asm | 84 +++++++++++++++++++-------------------
1 file changed, 42 insertions(+), 42 deletions(-)
diff --git a/libavcodec/x86/hpeldsp.asm b/libavcodec/x86/hpeldsp.asm
index c92c70f5ad..cbdf0e460d 100644
--- a/libavcodec/x86/hpeldsp.asm
+++ b/libavcodec/x86/hpeldsp.asm
@@ -54,8 +54,8 @@ cglobal put_pixels8_x2, 4,5
pavgb m0, m2
pavgb m1, m3
%else
- PAVGB m0, [r1]
- PAVGB m1, [r1+r2]
+ pavgb m0, [r1]
+ pavgb m1, [r1+r2]
%endif
mova [r0], m0
mova [r0+r2], m1
@@ -69,8 +69,8 @@ cglobal put_pixels8_x2, 4,5
pavgb m0, m2
pavgb m1, m3
%else
- PAVGB m0, [r1]
- PAVGB m1, [r1+r2]
+ pavgb m0, [r1]
+ pavgb m1, [r1+r2]
%endif
add r1, r4
mova [r0], m0
@@ -103,8 +103,8 @@ cglobal put_no_rnd_pixels8_x2, 4,5
add r1, r4
psubusb m0, m6
psubusb m2, m6
- PAVGB m0, m1
- PAVGB m2, m3
+ pavgb m0, m1
+ pavgb m2, m3
mova [r0], m0
mova [r0+r2], m2
mova m0, [r1]
@@ -115,8 +115,8 @@ cglobal put_no_rnd_pixels8_x2, 4,5
add r1, r4
psubusb m0, m6
psubusb m2, m6
- PAVGB m0, m1
- PAVGB m2, m3
+ pavgb m0, m1
+ pavgb m2, m3
mova [r0], m0
mova [r0+r2], m2
add r0, r4
@@ -143,8 +143,8 @@ cglobal %1_no_rnd_pixels8_x2_exact, 4,5
pxor m2, m4
pxor m1, m4
pxor m3, m4
- PAVGB m0, m1
- PAVGB m2, m3
+ pavgb m0, m1
+ pavgb m2, m3
pxor m0, m4
pxor m2, m4
%ifidn %1, avg
@@ -161,8 +161,8 @@ cglobal %1_no_rnd_pixels8_x2_exact, 4,5
pxor m1, m4
pxor m2, m4
pxor m3, m4
- PAVGB m0, m1
- PAVGB m2, m3
+ pavgb m0, m1
+ pavgb m2, m3
pxor m0, m4
pxor m2, m4
%ifidn %1, avg
@@ -198,16 +198,16 @@ cglobal put_pixels8_y2, 4,5
movu m1, [r1+r2]
movu m2, [r1+r4]
add r1, r4
- PAVGB m0, m1
- PAVGB m1, m2
+ pavgb m0, m1
+ pavgb m1, m2
mova [r0+r2], m0
mova [r0+r4], m1
movu m1, [r1+r2]
movu m0, [r1+r4]
add r0, r4
add r1, r4
- PAVGB m2, m1
- PAVGB m1, m0
+ pavgb m2, m1
+ pavgb m1, m0
mova [r0+r2], m2
mova [r0+r4], m1
add r0, r4
@@ -235,8 +235,8 @@ cglobal put_no_rnd_pixels8_y2, 4,5
mova m2, [r1+r4]
add r1, r4
psubusb m1, m6
- PAVGB m0, m1
- PAVGB m1, m2
+ pavgb m0, m1
+ pavgb m1, m2
mova [r0+r2], m0
mova [r0+r4], m1
mova m1, [r1+r2]
@@ -244,8 +244,8 @@ cglobal put_no_rnd_pixels8_y2, 4,5
add r0, r4
add r1, r4
psubusb m1, m6
- PAVGB m2, m1
- PAVGB m1, m0
+ pavgb m2, m1
+ pavgb m1, m0
mova [r0+r2], m2
mova [r0+r4], m1
add r0, r4
@@ -271,8 +271,8 @@ cglobal %1_no_rnd_pixels8_y2_exact, 4,5
movu m2, [r1+r2]
pxor m1, m3
pxor m2, m3
- PAVGB m0, m1
- PAVGB m1, m2
+ pavgb m0, m1
+ pavgb m1, m2
pxor m0, m3
pxor m1, m3
%ifidn %1, avg
@@ -285,8 +285,8 @@ cglobal %1_no_rnd_pixels8_y2_exact, 4,5
movu m0, [r1+r4]
pxor m1, m3
pxor m0, m3
- PAVGB m2, m1
- PAVGB m1, m0
+ pavgb m2, m1
+ pavgb m1, m0
pxor m2, m3
pxor m1, m3
%ifidn %1, avg
@@ -325,11 +325,11 @@ cglobal avg_pixels8_x2, 4,5
pavgb m0, m1
pavgb m2, m3
%else
- PAVGB m0, [r1+1], m3, m5
- PAVGB m2, [r1+r2+1], m4, m5
+ pavgb m0, [r1+1]
+ pavgb m2, [r1+r2+1]
%endif
- PAVGB m0, [r0], m3, m5
- PAVGB m2, [r0+r2], m4, m5
+ pavgb m0, [r0]
+ pavgb m2, [r0+r2]
add r1, r4
mova [r0], m0
mova [r0+r2], m2
@@ -341,13 +341,13 @@ cglobal avg_pixels8_x2, 4,5
pavgb m0, m1
pavgb m2, m3
%else
- PAVGB m0, [r1+1], m3, m5
- PAVGB m2, [r1+r2+1], m4, m5
+ pavgb m0, [r1+1]
+ pavgb m2, [r1+r2+1]
%endif
add r0, r4
add r1, r4
- PAVGB m0, [r0], m3, m5
- PAVGB m2, [r0+r2], m4, m5
+ pavgb m0, [r0]
+ pavgb m2, [r0+r2]
mova [r0], m0
mova [r0+r2], m2
add r0, r4
@@ -377,20 +377,20 @@ cglobal avg_pixels8_y2, 4,5
movu m1, [r1+r2]
movu m2, [r1+r4]
add r1, r4
- PAVGB m0, m1
- PAVGB m1, m2
- PAVGB m0, [r0+r2]
- PAVGB m1, [r0+r4]
+ pavgb m0, m1
+ pavgb m1, m2
+ pavgb m0, [r0+r2]
+ pavgb m1, [r0+r4]
mova [r0+r2], m0
mova [r0+r4], m1
movu m1, [r1+r2]
movu m0, [r1+r4]
- PAVGB m2, m1
- PAVGB m1, m0
+ pavgb m2, m1
+ pavgb m1, m0
add r0, r4
add r1, r4
- PAVGB m2, [r0+r2]
- PAVGB m1, [r0+r4]
+ pavgb m2, [r0+r2]
+ pavgb m1, [r0+r4]
mova [r0+r2], m2
mova [r0+r4], m1
add r0, r4
@@ -509,7 +509,7 @@ cglobal %1%3_pixels16_xy2, 4,5,8
%ifidn %1, avg
mova m3, [r0+r4]
packuswb m4, m5
- PAVGB m4, m3
+ pavgb m4, m3
%else
packuswb m4, m5
%endif
@@ -535,7 +535,7 @@ cglobal %1%3_pixels16_xy2, 4,5,8
%ifidn %1, avg
mova m3, [r0+r4]
packuswb m0, m1
- PAVGB m0, m3
+ pavgb m0, m3
%else
packuswb m0, m1
%endif
--
2.49.1
_______________________________________________
ffmpeg-devel mailing list -- ffmpeg-devel@ffmpeg.org
To unsubscribe send an email to ffmpeg-devel-leave@ffmpeg.org
reply other threads:[~2025-10-30 11:02 UTC|newest]
Thread overview: [no followups] expand[flat|nested] mbox.gz Atom feed
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=176182212706.81.6416319141458007086@7d278768979e \
--to=ffmpeg-devel@ffmpeg.org \
--cc=code@ffmpeg.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Git Inbox Mirror of the ffmpeg-devel mailing list - see https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
This inbox may be cloned and mirrored by anyone:
git clone --mirror https://master.gitmailbox.com/ffmpegdev/0 ffmpegdev/git/0.git
# If you have public-inbox 1.1+ installed, you may
# initialize and index your mirror using the following commands:
public-inbox-init -V2 ffmpegdev ffmpegdev/ https://master.gitmailbox.com/ffmpegdev \
ffmpegdev@gitmailbox.com
public-inbox-index ffmpegdev
Example config snippet for mirrors.
AGPL code for this site: git clone https://public-inbox.org/public-inbox.git