支持FLV H265的patch编译遇到了问题 #1
Comments
|
make -j8 |
wangsan
pushed a commit
to wangsan/FFmpeg
that referenced
this issue
Nov 26, 2022
The code originally pre-multiply by 2 the steps, causing the running sum
of the h factors to drift away due to the lack of precision. It quickly
causes an inaccuracy > 0.01.
I tried diverse approaches such as multiply by 2.0 (instead of adding
the value itself) without success.
I'm unable to bench the impact of this change, feel free to compare.
This commit fixes the incoming aacpsdsp tests.
Following is an alternative simplified function (matching the incoming
AArch64 code) that may be used:
function ff_ps_stereo_interpolate_neon, export=1
vld1.32 {q0}, [r2]
vld1.32 {q1}, [r3]
ldr r12, [sp]
vmov.f32 q8, q0
vmov.f32 q9, q1
vzip.32 q8, q0
vzip.32 q9, q1
1:
vld1.32 {d4}, [r0,:64]
vld1.32 {d6}, [r1,:64]
vadd.f32 q8, q8, q9
vadd.f32 q0, q0, q1
vmov.f32 d5, d4
vmov.f32 d7, d6
vmul.f32 q2, q2, q8
vmla.f32 q2, q3, q0
vst1.32 {d4}, [r0,:64]!
vst1.32 {d5}, [r1,:64]!
subs r12, r12, ksvc#1
bgt 1b
bx lr
endfunc
wangsan
pushed a commit
to wangsan/FFmpeg
that referenced
this issue
Nov 26, 2022
default_ref[] is unconditionally initialized in h264_initialise_ref_list()
(called from ff_h264_build_ref_list(), called from h264_slice_init()).
This fixes the following tsan warning when running fate-h264:
WARNING: ThreadSanitizer: data race (pid=31070)
Write of size 8 at 0x7bbc000082a8 by thread T1 (mutexes: write M1628):
#0 memcpy /work/release-test/final/llvm.src/projects/compiler-rt/lib/tsan/../sanitizer_common/sanitizer_common_interceptors.inc:655:5
(ffmpeg+0x10de9d)
ksvc#1 h264_initialise_ref_list ffmpeg/libavcodec/h264_refs.c:214:29 (ffmpeg+0x1186b3f)
ksvc#2 ff_h264_build_ref_list ffmpeg/libavcodec/h264_refs.c:306 (ffmpeg+0x1186b3f)
ksvc#3 h264_slice_init ffmpeg/libavcodec/h264_slice.c:1900:11 (ffmpeg+0x1191149)
[..]
Previous read of size 8 at 0x7bbc000082a8 by main thread (mutexes:
write M1630):
#0 memcpy /work/release-test/final/llvm.src/projects/compiler-rt/lib/tsan/../sanitizer_common/sanitizer_common_interceptors.inc:655:5
(ffmpeg+0x10de9d)
ksvc#1 ff_h264_update_thread_context ffmpeg/libavcodec/h264_slice.c:411:5 (ffmpeg+0x118b7dc)
Signed-off-by: Wan-Teh Chang <[email protected]>
Signed-off-by: Ronald S. Bultje <[email protected]>
wangsan
pushed a commit
to wangsan/FFmpeg
that referenced
this issue
Nov 26, 2022
…-hls_segment_filename previously, specifying -hls_segment_filename meant s->base_output_dirname was never set, causing a segfault: (lldb) bt * thread ksvc#1, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x0) * frame #0: 0x00007fffbf7d82d0 libsystem_platform.dylib`_platform_strcmp + 80 frame ksvc#1: 0x00000001005f4dcf libavformat.57.dylib`io_open_default(s=0x000000010481cc00, pb=0x0000000104806020, url=0x0000000000000000, flags=2, options=0x0000000000000000) at options.c:107 frame ksvc#2: 0x0000000100528968 libavformat.57.dylib`hls_mux_init(s=0x000000010481cc00) at hlsenc.c:595 frame ksvc#3: 0x00000001005273cb libavformat.57.dylib`hls_write_header(s=0x000000010481cc00) at hlsenc.c:1518 frame ksvc#4: 0x00000001005c08d0 libavformat.57.dylib`write_header_internal(s=0x000000010481cc00) at mux.c:486 frame ksvc#5: 0x00000001005c0774 libavformat.57.dylib`avformat_write_header(s=0x000000010481cc00, options=0x00000001029026e8) at mux.c:539
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
https://github.com/ksvc/FFmpeg/wiki/hevcpush
zexudeMBP:FFmpeg zexu$ make j8
make: *** No rule to make target `j8'. Stop.
平台：MAC OS
The text was updated successfully, but these errors were encountered: