michaelni changed the topic of #ffmpeg-devel to: Welcome to the FFmpeg development channel | Questions about using FFmpeg or developing with libav* libs should be asked in #ffmpeg | This channel is publicly logged | FFmpeg 7.0.1 has been released! | Please read ffmpeg.org/developer.html#Code-of-conduct
psykose has quit [Remote host closed the connection]
psykose has joined #ffmpeg-devel
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
psykose_ has joined #ffmpeg-devel
psykose has quit [Ping timeout: 264 seconds]
psykose_ is now known as psykose
System_Error has quit [Remote host closed the connection]
<cone-063>
ffmpeg Tomas Härdin master:818a487849b9: lavu/common.h: Fix UB in av_clipl_int32_c()
<cone-063>
ffmpeg Tomas Härdin master:60ab40be7086: lavu/common.h: Fix UB in av_clip_intp2_c()
<cone-063>
ffmpeg Tomas Härdin master:3b9e457647a3: lavu/common.h: Fix UB in av_clip_uintp2_c()
<cone-063>
ffmpeg Tomas Härdin master:be2cabce327b: lavu/intmath.h: Fix UB in ff_ctz_c() and ff_ctzll_c()
LaserEyess has quit [Quit: fugg]
LaserEyess has joined #ffmpeg-devel
LaserEyess has quit [Changing host]
LaserEyess has joined #ffmpeg-devel
LaserEyess has quit [Quit: fugg]
LaserEyess has joined #ffmpeg-devel
Krowl has quit [Read error: Connection reset by peer]
ccawley2011 has quit [Ping timeout: 264 seconds]
Krowl has joined #ffmpeg-devel
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
ccawley2011 has joined #ffmpeg-devel
<mkver>
courmisch: How big is the penalty for accesses not in the +-2KiB range?
<courmisch>
you'll need to materialise the offset and then add it, as with most RISC ISAs including Arm
<courmisch>
same as Armv8 really, except maybe the range is slightly different because Armv8 has scaled offsets?
<courmisch>
I don't think it's a performance problem. It's more of a maintainability problem if anybody changing MpegEncContext needs to edit assembler
<jamrial>
i agree. in an utopian world MpegEncContext would be untangled
<mkver>
jamrial: I'll come to this. Soon.
<ramiro>
mkver: do you plan on splitting the per-thread variables out of the main context?
<mkver>
ramiro: Actually, splitting the stuff only used by the main encoder thread into another struct first. That is the lowest hanging fruit.
<courmisch>
Linux has some fugly Makefile trickery to evaluate the offsetof()s into a .h file that can be included in .S
<courmisch>
I am *not* going to try to replicate that
<ramiro>
mkver: also the casting of pointers to bigger structures that contain MpegEncContext as first element (for example in ff_h261_encode_picture_header()). that's particularly nasty
<mkver>
Yeah, getters are better for that.
<courmisch>
is it? that's effectively what C++ virtual functions do, or what the Linux-originated container_of() macro does in many C projects
<courmisch>
I guess CHERI people don't like it though
<mkver>
courmisch: I mean static inline getters that do the cast, so that you don't have to do it at random places in the codebase and to preserve as much type-safety as possible.
<ramiro>
mkver: nice. thank you for all that work.
<courmisch>
this is going nowhere. I'm goign to have to ask the TC.
cone-063 has quit [Quit: transmission timeout]
Krowl has quit [Read error: Connection reset by peer]
j45_ has joined #ffmpeg-devel
j45 has quit [Ping timeout: 240 seconds]
j45_ is now known as j45
j45 has quit [Changing host]
j45 has joined #ffmpeg-devel
AbleBacon has joined #ffmpeg-devel
<courmisch>
is the ARMv5TE stuff for Nokia 770? I thought that was an ARMv6 device, but I guess not?
ngaullier has quit [Ping timeout: 272 seconds]
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
Sean_McG has joined #ffmpeg-devel
* Sean_McG
waves
<Sean_McG>
so I finally got a bit of time to look into if my Mac mini G4 can be used -- last time I tried (many months ago), it let out the magic smoke and would not boot again. Now for some reason I am getting the boot chime, but sadly no network activity. I don't have a display with DVI input so I can't see where it is stopping, I wonder if it's still possible to find one used and cheap.
<s55>
Sean_McG, If it's working after magic smoke, you probably blew a RIFA capacitor. While not strictly mandatory, I wouldn't run with a blown RIFA in there.
<Sean_McG>
I don't think they used RIFA's by the time this was built
<s55>
Possibly one of the other caps then. Would explain why it's sort of booting if one of the rails is poorly.
AbleBacon has quit [Read error: Connection reset by peer]
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
AbleBacon has joined #ffmpeg-devel
odrling has quit [Remote host closed the connection]
System_Error has quit [Remote host closed the connection]
mkver has quit [Ping timeout: 264 seconds]
System_Error has joined #ffmpeg-devel
cone-535 has joined #ffmpeg-devel
<cone-535>
ffmpeg Michael Niedermayer release/4.3:473663d53685: avfilter/vf_thumbnail_cuda: Set ret before checking it
<cone-535>
ffmpeg Michael Niedermayer release/4.3:fe924faa2f85: avfilter/signature_lookup: Fix 2 differences to the refernce SW
<cone-535>
ffmpeg Michael Niedermayer release/4.3:c8a5242a5457: doc/examples/vaapi_transcode: Simplify loop
<cone-535>
ffmpeg Michael Niedermayer release/4.3:1bff9db6918c: avfilter/signature_lookup: Dont copy uninitialized stuff around
<cone-535>
ffmpeg Michael Niedermayer release/4.3:72c789e02824: avcodec/ac3_parser: Check init_get_bits8() for failure
<cone-535>
ffmpeg Michael Niedermayer release/4.3:121fae5591a0: avcodec/hevcdec: Check ref frame
<cone-535>
ffmpeg Michael Niedermayer release/4.3:98a7a5000f9e: avformat/kvag: Check sample_rate
<cone-535>
ffmpeg Michael Niedermayer release/4.3:2a49bf571379: avformat/mxfdec: Check body_offset
<cone-535>
ffmpeg Michael Niedermayer release/4.3:11410efb23dd: swscale/output: Fix integer overflow in yuv2rgba64_1_c_template
<cone-535>
ffmpeg Michael Niedermayer release/4.3:27326d0e09f6: swscale/output: Fix integer overflow in yuv2rgba64_full_1_c_template()
<cone-535>
ffmpeg Michael Niedermayer release/4.3:816ac9de6797: avcodec/mpegvideo_enc: Fix 1 line and one column images
<cone-535>
ffmpeg Michael Niedermayer release/4.3:62ffa50af2d7: doc/examples/demux_decode: Simplify loop
<cone-535>
ffmpeg Michael Niedermayer release/4.3:682b002d1538: avcodec/cbs_av1: Avoid shift overflow
<cone-535>
ffmpeg Michael Niedermayer release/4.3:ef80220f3278: libavutil/base64: Try not to write over the array end
<cone-535>
ffmpeg Michael Niedermayer release/4.3:a0af91c34850: avutil/tests/base64: Check with too short output array
<cone-535>
ffmpeg Michael Niedermayer release/4.3:4f5ae895ce91: avcodec/lpc: copy levenson coeffs only when they have been computed
<cone-535>
ffmpeg Michael Niedermayer release/4.3:a06d3b96fd1d: avcodec/h264_slice: Remove dead sps check
<cone-535>
ffmpeg Michael Niedermayer release/4.3:749a3cdae8fc: avcodec/fmvc: remove dead assignment
<cone-535>
ffmpeg Michael Niedermayer release/4.3:52f7861420ec: avcodec/qsvdec: Check av_image_get_buffer_size() for failure
<cone-535>
ffmpeg Michael Niedermayer release/4.3:b7deab84be3b: avcodec/jpeg2000dec: remove ST=3 case
<cone-535>
ffmpeg Michael Niedermayer release/4.3:419c77c81a22: avcodec/vp3: Replace check by assert
<cone-535>
ffmpeg Michael Niedermayer release/4.3:590a0b5f3334: avcodec/vble: Check av_image_get_buffer_size() for failure
<cone-535>
ffmpeg Michael Niedermayer release/4.3:b84b27b3606e: avcodec/mpeg4videodec: assert impossible wrap points
<cone-535>
ffmpeg Michael Niedermayer release/4.3:25826459d98d: avcodec/mpegvideo_enc: Fix potential overflow in RD
<cone-535>
ffmpeg Michael Niedermayer release/4.3:5cca477c5125: avcodec/mscc & mwsc: Check loop counts before use
<cone-535>
ffmpeg Michael Niedermayer release/4.3:6b0671a6442d: avdevice/dshow: fix badly indented line
<cone-535>
ffmpeg Michael Niedermayer release/4.3:8d969d277888: avutil/tests/dict: Check av_dict_set() before get for failure
<cone-535>
ffmpeg Michael Niedermayer release/4.3:d180fa89cb8e: avutil/tests/opt: Check av_set_options_string() for failure
<cone-535>
ffmpeg Michael Niedermayer release/4.3:b9924b244eda: swscale/yuv2rgb: Use 64bit for brightness computation
<cone-535>
ffmpeg Michael Niedermayer release/4.3:9e5d40ff9f83: avcodec/tiff: Assert init_get_bits8() success in horizontal_fill()
<cone-535>
ffmpeg Michael Niedermayer release/4.3:9c605c88a9a5: avcodec/tiff: Assert init_get_bits8() success in unpack_gray()
<cone-535>
ffmpeg Michael Niedermayer release/4.3:2b79276114c0: avformat/ape: Use 64bit for final frame size
<cone-535>
ffmpeg Michael Niedermayer release/4.3:2cca78ee6886: avformat/asfdec_f: Use 64bit for preroll computation
<cone-535>
ffmpeg Michael Niedermayer release/4.3:4546c600c639: avformat/fwse: Remove always false expression
<cone-535>
ffmpeg Michael Niedermayer release/4.3:36c061fe22fa: avcodec/tests/jpeg2000dwt: Use 64bit in err2 computation
<cone-535>
ffmpeg Michael Niedermayer release/4.3:a6502919bd7a: avcodec/tests/jpeg2000dwt: Use 64bit in comparission
<cone-535>
ffmpeg Michael Niedermayer release/4.3:99c8b0ac22d5: avcodec/wavpackenc: Use unsigned for potential 31bit shift
<cone-535>
ffmpeg Michael Niedermayer release/4.3:86a2543744fa: avcodec/scpr3: Check add_dec() for failure
<cone-535>
ffmpeg Michael Niedermayer release/4.3:9bd68c46a6c0: avcodec/tests/dct: Use 64bit in intermediate for error computation
<cone-535>
ffmpeg Michael Niedermayer release/4.3:32fc533b2c12: avcodec/notchlc: Check init_get_bits8() for failure
<cone-535>
ffmpeg Michael Niedermayer release/4.3:ea1e12e029ed: avcodec/vp8: Check mutex init
<cone-535>
ffmpeg Michael Niedermayer release/4.3:02b37af1776a: avcodec/vp8: Check cond init
<cone-535>
ffmpeg Michael Niedermayer release/4.3:b06c0e8b8384: avcodec/ilbcdec: Remove dead code
<cone-535>
ffmpeg Michael Niedermayer release/4.3:5b392f49834c: avformat/sdp: Check before appending ","
<cone-535>
ffmpeg Michael Niedermayer release/4.3:4a91134a18c6: fftools/cmdutils: Add protective () to FLAGS
<cone-535>
ffmpeg Michael Niedermayer release/4.3:413b07402e09: avdevice/xcbgrab: Check sscanf() return
<cone-535>
ffmpeg Michael Niedermayer release/4.3:c81fd6fd6db3: avformat/img2dec: Move DQT after unrelated if()
<cone-535>
ffmpeg Michael Niedermayer release/4.3:bf5ba9a1706a: avformat/libzmq: Check av_strstart()
<cone-535>
ffmpeg Michael Niedermayer release/4.3:316bd5787b18: avformat/matroskadec: Assert that num_levels is non negative
<cone-535>
ffmpeg Michael Niedermayer release/4.3:ef18ef408513: avfilter/vf_rotate: Check ff_draw_init2() return value
<cone-535>
ffmpeg Michael Niedermayer release/4.3:5dcac90c77ef: Update for 4.3.7
Krowl has joined #ffmpeg-devel
Krowl has quit [Read error: Connection reset by peer]
<ramiro>
I'm having a hard time adding unscaled yuv2rgb to checkasm. the c functions use LUTs, which give different results from simd, so memcmp() doesn't work.
<ramiro>
I am considering adding c functions that don't use LUTs and select them under SWS_BITEXACT, but I'm not exactly sure what bitexact means in the context of libswscale.
<Lynne>
bitexact flags shouldn't exist IMO
<Lynne>
definitely not in DSP code
<ramiro>
Lynne: do you think all dsp code should always give the same result? or is it some other reasoning?
<Lynne>
yes, all DSP code should always give the same results
<ramiro>
sometimes we can gain a lot of performance by cutting some corners and the result doesn't look too bad.
<nevcairiel>
sws is definitely not accurate/identical in all pathes
<wbs>
ramiro: check some of the existing swscale checkasm tests, we do have some tests that do off-by-few checks for cases when SWS_BITEXACT isn't set
MikhailAMD has joined #ffmpeg-devel
System_Error has quit [Remote host closed the connection]
<ramiro>
wbs: I thought about that, but it doesn't seem right to compare the output from functions that are essentially doing different things
System_Error has joined #ffmpeg-devel
<wbs>
sure, but still, if they're meant to output roughly the same thing, even if using a different algorithm, an off-by-few check should be good enough I think?
<Lynne>
ramiro: there's no such thing as "doesn't look too bad", there's a correct implementation and an incorrect implementation