From: James Almer <jamrial@gmail.com>
To: ffmpeg-devel@ffmpeg.org
Subject: [FFmpeg-devel] [PATCH] swscale/swscale_unscaled: account for semi planar formats with data in the msb
Date: Fri, 14 Mar 2025 11:52:33 -0300
Message-ID: <20250314145233.49307-1-jamrial@gmail.com> (raw)
Fixes fate failures introduced by recent tests that exercise the faulty code.
Signed-off-by: James Almer <jamrial@gmail.com>
---
libswscale/swscale_unscaled.c | 74 ++++++++++++++++++-----------------
tests/ref/pixfmt/p410-nv24 | 2 +-
tests/ref/pixfmt/p410-p412be | 2 +-
tests/ref/pixfmt/p410-p412le | 2 +-
tests/ref/pixfmt/p410-p416be | 2 +-
tests/ref/pixfmt/p410-p416le | 2 +-
tests/ref/pixfmt/p412-nv24 | 2 +-
tests/ref/pixfmt/p412-p410be | 2 +-
tests/ref/pixfmt/p412-p410le | 2 +-
tests/ref/pixfmt/p412-p416be | 2 +-
tests/ref/pixfmt/p412-p416le | 2 +-
11 files changed, 48 insertions(+), 46 deletions(-)
diff --git a/libswscale/swscale_unscaled.c b/libswscale/swscale_unscaled.c
index 1ab36b9568..92d5386567 100644
--- a/libswscale/swscale_unscaled.c
+++ b/libswscale/swscale_unscaled.c
@@ -2146,17 +2146,17 @@ static int packedCopyWrapper(SwsInternal *c, const uint8_t *const src[],
if (c->opts.dither == SWS_DITHER_NONE) {\
for (i = 0; i < height; i++) {\
for (j = 0; j < length-7; j+=8) {\
- tmp = (bswap(src[j+0]) + bias)>>shift; dst[j+0] = dbswap(tmp - (tmp>>dst_depth));\
- tmp = (bswap(src[j+1]) + bias)>>shift; dst[j+1] = dbswap(tmp - (tmp>>dst_depth));\
- tmp = (bswap(src[j+2]) + bias)>>shift; dst[j+2] = dbswap(tmp - (tmp>>dst_depth));\
- tmp = (bswap(src[j+3]) + bias)>>shift; dst[j+3] = dbswap(tmp - (tmp>>dst_depth));\
- tmp = (bswap(src[j+4]) + bias)>>shift; dst[j+4] = dbswap(tmp - (tmp>>dst_depth));\
- tmp = (bswap(src[j+5]) + bias)>>shift; dst[j+5] = dbswap(tmp - (tmp>>dst_depth));\
- tmp = (bswap(src[j+6]) + bias)>>shift; dst[j+6] = dbswap(tmp - (tmp>>dst_depth));\
- tmp = (bswap(src[j+7]) + bias)>>shift; dst[j+7] = dbswap(tmp - (tmp>>dst_depth));\
+ tmp = ((bswap(src[j+0]) >> src_shift) + bias)>>shift; dst[j+0] = dbswap((tmp - (tmp>>dst_depth)) << dst_shift);\
+ tmp = ((bswap(src[j+1]) >> src_shift) + bias)>>shift; dst[j+1] = dbswap((tmp - (tmp>>dst_depth)) << dst_shift);\
+ tmp = ((bswap(src[j+2]) >> src_shift) + bias)>>shift; dst[j+2] = dbswap((tmp - (tmp>>dst_depth)) << dst_shift);\
+ tmp = ((bswap(src[j+3]) >> src_shift) + bias)>>shift; dst[j+3] = dbswap((tmp - (tmp>>dst_depth)) << dst_shift);\
+ tmp = ((bswap(src[j+4]) >> src_shift) + bias)>>shift; dst[j+4] = dbswap((tmp - (tmp>>dst_depth)) << dst_shift);\
+ tmp = ((bswap(src[j+5]) >> src_shift) + bias)>>shift; dst[j+5] = dbswap((tmp - (tmp>>dst_depth)) << dst_shift);\
+ tmp = ((bswap(src[j+6]) >> src_shift) + bias)>>shift; dst[j+6] = dbswap((tmp - (tmp>>dst_depth)) << dst_shift);\
+ tmp = ((bswap(src[j+7]) >> src_shift) + bias)>>shift; dst[j+7] = dbswap((tmp - (tmp>>dst_depth)) << dst_shift);\
}\
for (; j < length; j++) {\
- tmp = (bswap(src[j]) + bias)>>shift; dst[j] = dbswap(tmp - (tmp>>dst_depth));\
+ tmp = (bswap(src[j]) + bias)>>shift; dst[j] = dbswap(tmp - (tmp>>dst_depth) << dst_shift);\
}\
dst += dstStride;\
src += srcStride;\
@@ -2165,17 +2165,17 @@ static int packedCopyWrapper(SwsInternal *c, const uint8_t *const src[],
for (i = 0; i < height; i++) {\
const uint8_t *dither= dithers[shift-1][i&7];\
for (j = 0; j < length-7; j+=8) {\
- tmp = (bswap(src[j+0]) + dither[0])>>shift; dst[j+0] = dbswap(tmp - (tmp>>dst_depth));\
- tmp = (bswap(src[j+1]) + dither[1])>>shift; dst[j+1] = dbswap(tmp - (tmp>>dst_depth));\
- tmp = (bswap(src[j+2]) + dither[2])>>shift; dst[j+2] = dbswap(tmp - (tmp>>dst_depth));\
- tmp = (bswap(src[j+3]) + dither[3])>>shift; dst[j+3] = dbswap(tmp - (tmp>>dst_depth));\
- tmp = (bswap(src[j+4]) + dither[4])>>shift; dst[j+4] = dbswap(tmp - (tmp>>dst_depth));\
- tmp = (bswap(src[j+5]) + dither[5])>>shift; dst[j+5] = dbswap(tmp - (tmp>>dst_depth));\
- tmp = (bswap(src[j+6]) + dither[6])>>shift; dst[j+6] = dbswap(tmp - (tmp>>dst_depth));\
- tmp = (bswap(src[j+7]) + dither[7])>>shift; dst[j+7] = dbswap(tmp - (tmp>>dst_depth));\
+ tmp = ((bswap(src[j+0]) >> src_shift) + dither[0])>>shift; dst[j+0] = dbswap((tmp - (tmp>>dst_depth)) << dst_shift);\
+ tmp = ((bswap(src[j+1]) >> src_shift) + dither[1])>>shift; dst[j+1] = dbswap((tmp - (tmp>>dst_depth)) << dst_shift);\
+ tmp = ((bswap(src[j+2]) >> src_shift) + dither[2])>>shift; dst[j+2] = dbswap((tmp - (tmp>>dst_depth)) << dst_shift);\
+ tmp = ((bswap(src[j+3]) >> src_shift) + dither[3])>>shift; dst[j+3] = dbswap((tmp - (tmp>>dst_depth)) << dst_shift);\
+ tmp = ((bswap(src[j+4]) >> src_shift) + dither[4])>>shift; dst[j+4] = dbswap((tmp - (tmp>>dst_depth)) << dst_shift);\
+ tmp = ((bswap(src[j+5]) >> src_shift) + dither[5])>>shift; dst[j+5] = dbswap((tmp - (tmp>>dst_depth)) << dst_shift);\
+ tmp = ((bswap(src[j+6]) >> src_shift) + dither[6])>>shift; dst[j+6] = dbswap((tmp - (tmp>>dst_depth)) << dst_shift);\
+ tmp = ((bswap(src[j+7]) >> src_shift) + dither[7])>>shift; dst[j+7] = dbswap((tmp - (tmp>>dst_depth)) << dst_shift);\
}\
for (; j < length; j++) {\
- tmp = (bswap(src[j]) + dither[j&7])>>shift; dst[j] = dbswap(tmp - (tmp>>dst_depth));\
+ tmp = (bswap(src[j]) + dither[j&7])>>shift; dst[j] = dbswap(tmp - (tmp>>dst_depth) << dst_shift);\
}\
dst += dstStride;\
src += srcStride;\
@@ -2184,14 +2184,14 @@ static int packedCopyWrapper(SwsInternal *c, const uint8_t *const src[],
for (i = 0; i < height; i++) {\
const uint8_t *dither= dithers[shift-1][i&7];\
for (j = 0; j < length-7; j+=8) {\
- tmp = bswap(src[j+0]); dst[j+0] = dbswap((tmp - (tmp>>dst_depth) + dither[0])>>shift);\
- tmp = bswap(src[j+1]); dst[j+1] = dbswap((tmp - (tmp>>dst_depth) + dither[1])>>shift);\
- tmp = bswap(src[j+2]); dst[j+2] = dbswap((tmp - (tmp>>dst_depth) + dither[2])>>shift);\
- tmp = bswap(src[j+3]); dst[j+3] = dbswap((tmp - (tmp>>dst_depth) + dither[3])>>shift);\
- tmp = bswap(src[j+4]); dst[j+4] = dbswap((tmp - (tmp>>dst_depth) + dither[4])>>shift);\
- tmp = bswap(src[j+5]); dst[j+5] = dbswap((tmp - (tmp>>dst_depth) + dither[5])>>shift);\
- tmp = bswap(src[j+6]); dst[j+6] = dbswap((tmp - (tmp>>dst_depth) + dither[6])>>shift);\
- tmp = bswap(src[j+7]); dst[j+7] = dbswap((tmp - (tmp>>dst_depth) + dither[7])>>shift);\
+ tmp = bswap(src[j+0]) >> src_shift; dst[j+0] = dbswap(((tmp - (tmp>>dst_depth) + dither[0])>>shift) << dst_shift);\
+ tmp = bswap(src[j+1]) >> src_shift; dst[j+1] = dbswap(((tmp - (tmp>>dst_depth) + dither[1])>>shift) << dst_shift);\
+ tmp = bswap(src[j+2]) >> src_shift; dst[j+2] = dbswap(((tmp - (tmp>>dst_depth) + dither[2])>>shift) << dst_shift);\
+ tmp = bswap(src[j+3]) >> src_shift; dst[j+3] = dbswap(((tmp - (tmp>>dst_depth) + dither[3])>>shift) << dst_shift);\
+ tmp = bswap(src[j+4]) >> src_shift; dst[j+4] = dbswap(((tmp - (tmp>>dst_depth) + dither[4])>>shift) << dst_shift);\
+ tmp = bswap(src[j+5]) >> src_shift; dst[j+5] = dbswap(((tmp - (tmp>>dst_depth) + dither[5])>>shift) << dst_shift);\
+ tmp = bswap(src[j+6]) >> src_shift; dst[j+6] = dbswap(((tmp - (tmp>>dst_depth) + dither[6])>>shift) << dst_shift);\
+ tmp = bswap(src[j+7]) >> src_shift; dst[j+7] = dbswap(((tmp - (tmp>>dst_depth) + dither[7])>>shift) << dst_shift);\
}\
for (; j < length; j++) {\
tmp = bswap(src[j]); dst[j] = dbswap((tmp - (tmp>>dst_depth) + dither[j&7])>>shift);\
@@ -2235,6 +2235,8 @@ static int planarCopyWrapper(SwsInternal *c, const uint8_t *const src[],
) {
const int src_depth = desc_src->comp[plane].depth;
const int dst_depth = desc_dst->comp[plane].depth;
+ const int src_shift = desc_src->comp[plane].shift;
+ const int dst_shift = desc_dst->comp[plane].shift;
const uint16_t *srcPtr2 = (const uint16_t *) srcPtr;
uint16_t *dstPtr2 = (uint16_t*)dstPtr;
@@ -2250,11 +2252,11 @@ static int planarCopyWrapper(SwsInternal *c, const uint8_t *const src[],
#define COPY816(w)\
if (shiftonly) {\
for (j = 0; j < length; j++)\
- w(&dstPtr2[j], srcPtr[j]<<(dst_depth-8));\
+ w(&dstPtr2[j], (srcPtr[j]<<(dst_depth-8)) << dst_shift);\
} else {\
for (j = 0; j < length; j++)\
- w(&dstPtr2[j], (srcPtr[j]<<(dst_depth-8)) |\
- (srcPtr[j]>>(2*8-dst_depth)));\
+ w(&dstPtr2[j], ((srcPtr[j]<<(dst_depth-8)) |\
+ (srcPtr[j]>>(2*8-dst_depth))) << dst_shift);\
}
if(isBE(c->opts.dst_format)){
COPY816(AV_WB16)
@@ -2274,13 +2276,13 @@ static int planarCopyWrapper(SwsInternal *c, const uint8_t *const src[],
#if HAVE_FAST_64BIT
#define FAST_COPY_UP(shift) \
for (; j < length - 3; j += 4) { \
- uint64_t v = AV_RN64A(srcPtr2 + j); \
+ uint64_t v = AV_RN64A(srcPtr2 + j) >> src_shift; \
AV_WN64A(dstPtr2 + j, v << shift); \
}
#else
#define FAST_COPY_UP(shift) \
for (; j < length - 1; j += 2) { \
- uint32_t v = AV_RN32A(srcPtr2 + j); \
+ uint32_t v = AV_RN32A(srcPtr2 + j) >> src_shift; \
AV_WN32A(dstPtr2 + j, v << shift); \
}
#endif
@@ -2293,14 +2295,14 @@ static int planarCopyWrapper(SwsInternal *c, const uint8_t *const src[],
#define COPY_UP(r,w) \
if(shiftonly){\
for (; j < length; j++){ \
- unsigned int v= r(&srcPtr2[j]);\
- w(&dstPtr2[j], v<<(dst_depth-src_depth));\
+ unsigned int v= r(&srcPtr2[j]) >> src_shift;\
+ w(&dstPtr2[j], (v << (dst_depth-src_depth)) << dst_shift);\
}\
}else{\
for (; j < length; j++){ \
- unsigned int v= r(&srcPtr2[j]);\
- w(&dstPtr2[j], (v<<(dst_depth-src_depth)) | \
- (v>>(2*src_depth-dst_depth)));\
+ unsigned int v= r(&srcPtr2[j]) >> src_shift;\
+ w(&dstPtr2[j], ((v<<(dst_depth-src_depth)) | \
+ (v>>(2*src_depth-dst_depth))) << dst_shift);\
}\
}
if(isBE(c->opts.src_format)){
diff --git a/tests/ref/pixfmt/p410-nv24 b/tests/ref/pixfmt/p410-nv24
index a13bc773ae..8996a24afb 100644
--- a/tests/ref/pixfmt/p410-nv24
+++ b/tests/ref/pixfmt/p410-nv24
@@ -1,2 +1,2 @@
-65d33af5a5828eb16b42573e9720b370 *tests/data/pixfmt/p410-nv24.yuv
+bd4ba14d2828d10001aa6111adda31a8 *tests/data/pixfmt/p410-nv24.yuv
15206400 tests/data/pixfmt/p410-nv24.yuv
diff --git a/tests/ref/pixfmt/p410-p412be b/tests/ref/pixfmt/p410-p412be
index 3e033d71a0..bfcb2aa7e1 100644
--- a/tests/ref/pixfmt/p410-p412be
+++ b/tests/ref/pixfmt/p410-p412be
@@ -1,2 +1,2 @@
-1b2996e48f9f2f9c08b47a7af2d1ad3a *tests/data/pixfmt/p410-p412be.yuv
+60b454bfd6f48b64cc24de3041d0a01d *tests/data/pixfmt/p410-p412be.yuv
15206400 tests/data/pixfmt/p410-p412be.yuv
diff --git a/tests/ref/pixfmt/p410-p412le b/tests/ref/pixfmt/p410-p412le
index a80edad6c3..42823bf4d7 100644
--- a/tests/ref/pixfmt/p410-p412le
+++ b/tests/ref/pixfmt/p410-p412le
@@ -1,2 +1,2 @@
-1b2996e48f9f2f9c08b47a7af2d1ad3a *tests/data/pixfmt/p410-p412le.yuv
+60b454bfd6f48b64cc24de3041d0a01d *tests/data/pixfmt/p410-p412le.yuv
15206400 tests/data/pixfmt/p410-p412le.yuv
diff --git a/tests/ref/pixfmt/p410-p416be b/tests/ref/pixfmt/p410-p416be
index 0c944745fd..67d224753c 100644
--- a/tests/ref/pixfmt/p410-p416be
+++ b/tests/ref/pixfmt/p410-p416be
@@ -1,2 +1,2 @@
-0a3bffd9200664676aa69684570a0053 *tests/data/pixfmt/p410-p416be.yuv
+60b454bfd6f48b64cc24de3041d0a01d *tests/data/pixfmt/p410-p416be.yuv
15206400 tests/data/pixfmt/p410-p416be.yuv
diff --git a/tests/ref/pixfmt/p410-p416le b/tests/ref/pixfmt/p410-p416le
index 8eaa12cccd..f7be585a49 100644
--- a/tests/ref/pixfmt/p410-p416le
+++ b/tests/ref/pixfmt/p410-p416le
@@ -1,2 +1,2 @@
-cffbc4dedbfc209aec8bb4921c19fbf2 *tests/data/pixfmt/p410-p416le.yuv
+60b454bfd6f48b64cc24de3041d0a01d *tests/data/pixfmt/p410-p416le.yuv
15206400 tests/data/pixfmt/p410-p416le.yuv
diff --git a/tests/ref/pixfmt/p412-nv24 b/tests/ref/pixfmt/p412-nv24
index 074f07eb36..f7cf14c2e0 100644
--- a/tests/ref/pixfmt/p412-nv24
+++ b/tests/ref/pixfmt/p412-nv24
@@ -1,2 +1,2 @@
-5400f2bfef1b3dc0fcea1fd8ef91e2c8 *tests/data/pixfmt/p412-nv24.yuv
+bd4ba14d2828d10001aa6111adda31a8 *tests/data/pixfmt/p412-nv24.yuv
15206400 tests/data/pixfmt/p412-nv24.yuv
diff --git a/tests/ref/pixfmt/p412-p410be b/tests/ref/pixfmt/p412-p410be
index 245d2f4c43..72df60c699 100644
--- a/tests/ref/pixfmt/p412-p410be
+++ b/tests/ref/pixfmt/p412-p410be
@@ -1,2 +1,2 @@
-1e268eb904e8d173d1585474d12e6f10 *tests/data/pixfmt/p412-p410be.yuv
+51e761c143c3fce8e9921d8cceb0a433 *tests/data/pixfmt/p412-p410be.yuv
15206400 tests/data/pixfmt/p412-p410be.yuv
diff --git a/tests/ref/pixfmt/p412-p410le b/tests/ref/pixfmt/p412-p410le
index 04aa4eda77..4bd47bcf21 100644
--- a/tests/ref/pixfmt/p412-p410le
+++ b/tests/ref/pixfmt/p412-p410le
@@ -1,2 +1,2 @@
-1e268eb904e8d173d1585474d12e6f10 *tests/data/pixfmt/p412-p410le.yuv
+51e761c143c3fce8e9921d8cceb0a433 *tests/data/pixfmt/p412-p410le.yuv
15206400 tests/data/pixfmt/p412-p410le.yuv
diff --git a/tests/ref/pixfmt/p412-p416be b/tests/ref/pixfmt/p412-p416be
index ec2b35b705..1fe9935d2f 100644
--- a/tests/ref/pixfmt/p412-p416be
+++ b/tests/ref/pixfmt/p412-p416be
@@ -1,2 +1,2 @@
-598b43d980ff2a5e24ee4003befcb875 *tests/data/pixfmt/p412-p416be.yuv
+9b5fc804354cf299fcc4c3f7bd2781ad *tests/data/pixfmt/p412-p416be.yuv
15206400 tests/data/pixfmt/p412-p416be.yuv
diff --git a/tests/ref/pixfmt/p412-p416le b/tests/ref/pixfmt/p412-p416le
index 60b159fac0..07fdcfcc26 100644
--- a/tests/ref/pixfmt/p412-p416le
+++ b/tests/ref/pixfmt/p412-p416le
@@ -1,2 +1,2 @@
-598b43d980ff2a5e24ee4003befcb875 *tests/data/pixfmt/p412-p416le.yuv
+9b5fc804354cf299fcc4c3f7bd2781ad *tests/data/pixfmt/p412-p416le.yuv
15206400 tests/data/pixfmt/p412-p416le.yuv
--
2.48.1
_______________________________________________
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".
next reply other threads:[~2025-03-14 14:53 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-03-14 14:52 James Almer [this message]
2025-03-15 19:31 ` James Almer
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=20250314145233.49307-1-jamrial@gmail.com \
--to=jamrial@gmail.com \
--cc=ffmpeg-devel@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