sysutils/ffmpeg
1
0
Fork 0
mirror of https://git.ffmpeg.org/ffmpeg.git synced 2026-10-24 02:09:03 +03:00
Code Activity

avcodec/arm|aarch64/h264dsp: Fix edge cases for weight

Browse Source 
Add/sub with saturation is used only on log2_denom > 1 path, but with
log2_denom == 1 (block[x]*weight + offset << 1) can fall below S16_MIN.
Use saturation in all paths for safety.

This fixes the weight checkasm test for arm and aarch64

Signed-off-by: Zuxy Meng <zuxy.meng@gmail.com>
This commit is contained in:
Zuxy Meng
2026-09-04 21:09:51 -07:00
parent 460cb05218
commit c80bfd962f
3 changed files with 4 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
libavcodec/aarch64/h264dsp_neon.S
View File

@@ -810,9 +810,9 @@ function ff_weight_h264_pixels_\w\()_neon, export=1
dup v18.8h, w6
cmp w4, #0
b.lt 10f
weight_\w add
weight_\w sqadd
10: neg w4, w4
weight_\w sub
weight_\w sqsub
endfunc
.endm

4
libavcodec/arm/h264dsp_neon.S
View File

@@ -549,9 +549,9 @@ function ff_weight_h264_pixels_\w\()_neon, export=1
vdup.16 q9, lr
cmp r12, #0
blt 10f
weight_\w vadd.s16
weight_\w vqadd.s16
10: rsb r12, r12, #0
weight_\w vsub.s16
weight_\w vqsub.s16
endfunc
.endm

7
tests/checkasm/h264dsp.c
View File

@@ -500,10 +500,6 @@ static void check_loop_filter_intra(void)
}
}
// neon fails at edge cases
#define H264_CHECK_WEIGHT (!ARCH_ARM && !ARCH_AARCH64)
#if H264_CHECK_WEIGHT
static void check_weight(void)
{
LOCAL_ALIGNED_16(uint8_t, dst, [32 * 32 * 2]);
@@ -547,7 +543,6 @@ static void check_weight(void)
}
}
}
#endif
// only arch that can pass test
#define H264_CHECK_BIWEIGHT ARCH_X86
@@ -634,10 +629,8 @@ void checkasm_check_h264dsp(void)
check_loop_filter_intra();
report("loop_filter_intra");
#if H264_CHECK_WEIGHT
check_weight();
report("weight");
#endif
#if H264_CHECK_BIWEIGHT
check_biweight();