2024-05-26 01:34
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
00:01
<
cone-435 >
ffmpeg Haihao Xiang master:a4630d479a54: lavu/hwcontext_vulkan: Support write on drm frame
00:04
System_Error has joined #ffmpeg-devel
00:12
derpydoo has joined #ffmpeg-devel
00:22
lexano has quit [Ping timeout: 252 seconds]
00:32
thilo_ has quit [Ping timeout: 256 seconds]
00:33
thilo_ has joined #ffmpeg-devel
01:02
iive has quit [Quit: They came for me...]
01:16
arch1t3cht7 has joined #ffmpeg-devel
01:18
arch1t3cht has quit [Ping timeout: 252 seconds]
01:18
arch1t3cht7 is now known as arch1t3cht
01:34
<
cone-435 >
ffmpeg Lynne master:11a8e0a4e585: aacdec_usac: dequantize scalefactors after noise synthesis
01:34
<
cone-435 >
ffmpeg Lynne master:d79fbad36689: aacdec_usac: rename spectrum decode function and remove unused arg
01:34
<
cone-435 >
ffmpeg Lynne master:bdd3c6ca5015: aacdec_usac: always zero out alpha_q values for stereo streams
01:38
ubitux has quit [Ping timeout: 268 seconds]
01:39
ubitux has joined #ffmpeg-devel
01:49
Marth64[m] has joined #ffmpeg-devel
01:52
Marth64 has quit [Ping timeout: 264 seconds]
02:02
lemourin has joined #ffmpeg-devel
03:19
Martchus has joined #ffmpeg-devel
03:20
jamrial has quit []
03:21
Martchus_ has quit [Ping timeout: 252 seconds]
03:30
mkver has joined #ffmpeg-devel
03:46
kekePower has quit [Ping timeout: 264 seconds]
03:58
Marth64 has joined #ffmpeg-devel
03:58
AbleBacon has quit [Quit: I am like MacArthur; I shall return.]
04:01
Marth64[m] has quit [Ping timeout: 264 seconds]
04:06
tufei has quit [Quit: Leaving]
04:08
darkapex has quit [Ping timeout: 255 seconds]
04:15
kekePower has joined #ffmpeg-devel
04:20
tufei has joined #ffmpeg-devel
04:34
cone-435 has quit [Quit: transmission timeout]
06:12
kekePower has quit [Quit: Ping timeout (120 seconds)]
06:13
kekePower has joined #ffmpeg-devel
06:15
rvalue has quit [Read error: Connection reset by peer]
06:16
rvalue has joined #ffmpeg-devel
07:30
derpydoo has quit [Ping timeout: 252 seconds]
07:46
sadome has joined #ffmpeg-devel
07:46
sadome has joined #ffmpeg-devel
07:46
sadome has quit [Excess Flood]
08:11
kekePower has quit [Read error: Connection reset by peer]
08:11
kekePower7 has joined #ffmpeg-devel
08:15
<
elenril >
llyyr: thanks, patch sent
08:48
sfan5_ has joined #ffmpeg-devel
08:48
sfan5 has quit [Ping timeout: 256 seconds]
08:56
Livio has joined #ffmpeg-devel
09:04
ccawley2011 has joined #ffmpeg-devel
09:12
<
elenril >
sigh, why is ffprobe printing foo=0 for every disposition
09:13
<
JEEB >
some sort of flags logic?
09:19
Marth64 has quit [Remote host closed the connection]
09:19
Marth64 has joined #ffmpeg-devel
09:42
<
Daemon404 >
elenril, always has afaik
09:42
<
Daemon404 >
why, i cannot say
09:45
<
elenril >
it means adding a new flag breaks a bunch of completely unrelated tests
09:45
<
elenril >
great design, that
09:47
<
Daemon404 >
hmmm i bet you also broke the xsd
09:47
<
Daemon404 >
i het i did too with my spatial set
09:47
<
Daemon404 >
(xsd is evil)
09:48
kekePower7 has quit [Ping timeout: 255 seconds]
09:49
<
Daemon404 >
hmm mine didnt
10:03
System_Error has quit [Ping timeout: 260 seconds]
10:06
AndrewSayers has quit [Quit: Leaving.]
10:11
j45_ has joined #ffmpeg-devel
10:11
j45 has quit [Ping timeout: 240 seconds]
10:11
j45_ is now known as j45
10:11
j45 has quit [Changing host]
10:11
j45 has joined #ffmpeg-devel
10:12
System_Error has joined #ffmpeg-devel
10:18
kekePower has joined #ffmpeg-devel
10:40
Livio has quit [Ping timeout: 240 seconds]
11:02
AndrewSayers has joined #ffmpeg-devel
11:04
av500 has quit [Read error: Connection reset by peer]
11:04
av500 has joined #ffmpeg-devel
11:08
j45_ has joined #ffmpeg-devel
11:08
j45 has quit [Ping timeout: 268 seconds]
11:08
j45_ is now known as j45
11:08
j45 has quit [Changing host]
11:08
j45 has joined #ffmpeg-devel
11:10
jamrial has joined #ffmpeg-devel
11:12
j45_ has joined #ffmpeg-devel
11:13
j45 has quit [Ping timeout: 268 seconds]
11:13
j45_ is now known as j45
11:13
j45 has quit [Changing host]
11:13
j45 has joined #ffmpeg-devel
11:15
lexano has joined #ffmpeg-devel
11:17
cone-605 has joined #ffmpeg-devel
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:0c9e04b0e22c: avcodec/ratecontrol: Pass RCContext directly in ff_rate_control_uninit()
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:685cda26aa03: avcodec/mpegvideo, ratecontrol: Remove write-only skip_count
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:68a06640b1db: avcodec/ratecontrol: Don't call ff_rate_control_uninit() ourselves
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:ba28222a14ab: avcodec/ratecontrol: Fix double free on error
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:2786d3471249: avcodec/ratecontrol: Avoid padding in RateControlEntry
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:0aaabe1fd72f: avcodec/get_buffer: Remove redundant check
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:d4b63885a601: avcodec/mpegpicture: Store linesize in ScratchpadContext
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:a30c95ed8a51: avcodec/mpegvideo_dec: Sync linesize and uvlinesize between threads
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:99543c93ec09: avcodec/mpegvideo_dec: Factor allocating dummy frames out
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:73676dd47ac3: avcodec/mpegpicture: Mark dummy frames as such
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:c85508f5a91c: avcodec/mpeg12dec: Allocate dummy frames for non-I fields
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:f44d212e0e4e: avcodec/mpegvideo_motion: Remove dead checks for existence of reference
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:5f505995db8f: avcodec/mpegvideo_motion: Optimize check away
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:55e81306bfde: avcodec/mpegvideo_motion: Optimize check away
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:60d4c8a13740: avcodec/mpegvideo_motion: Avoid constant function argument
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:1a5c21daee4b: avcodec/msmpeg4enc: Only calculate coded_cbp when used
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:ba033acb56f2: avcodec/mpegvideo: Only allocate coded_block when needed
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:582d828f754f: avcodec/mpegvideo: Don't reset coded_block unnecessarily
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:caac9740f846: avcodec/mpegvideo: Only allocate cbp_table, pred_dir_table when needed
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:71ff9217f769: avcodec/mpegpicture: Always reset motion val buffer
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:7ad13e173bf2: avcodec/mpegpicture: Always reset mbskip_table
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:a95591dbfde3: avcodec/mpegvideo: Redo aligning mb_height for VC-1
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:788892d64763: avcodec/mpegvideo, mpegpicture: Add buffer pool
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:6450cfcd100a: avcodec/mpegpicture: Reindent after the previous commit
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:bed17eba4733: avcodec/mpegpicture: Use RefStruct-pool API
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:2cbca73975c9: avcodec/h263: Move encoder-only part out of ff_h263_update_motion_val()
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:101ed72c2f33: avcodec/h263, mpeg(picture|video): Only allocate mbskip_table for MPEG-4
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:9645eeb48554: avcodec/mpegvideo: Reindent after the previous commit
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:47e43c19cb50: avcodec/h263: Move setting mbskip_table to decoder/encoders
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:3a4e7694a13e: avcodec/mpegvideo: Restrict resetting mbskip_table to MPEG-4 decoder
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:ec1eba792aed: avcodec/mpegvideo: Shorten variable names
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:1c40a179222f: avcodec/mpegpicture: Reduce value of MAX_PLANES define
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:7814dd77aa61: avcodec/mpegpicture: Cache AVFrame.data and linesize values
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:f1c4e8950ebe: avcodec/rv30, rv34, rv40: Avoid indirection
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:dda009b97d31: avcodec/mpegvideo: Add const where appropriate
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:8c59b5aa6b41: avcodec/vc1_pred: Remove unused function parameter
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:042117da7530: avcodec/mpegpicture: Improve error messages and code
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:89ca63cc9cc2: avcodec/mpegpicture: Split ff_alloc_picture() into check and alloc part
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:8225d2da7390: avcodec/mpegvideo_enc: Pass AVFrame*, not Picture* to alloc_picture()
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:2dfe7c1e404d: avcodec/mpegvideo_enc: Move copying properties to alloc_picture()
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:59422955cf38: avcodec/mpegpicture: Rename Picture->MPVPicture
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:dac15a5b6ede: avcodec/vc1_mc: Don't check AVFrame INTERLACE flags
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:fe6037fd04db: avcodec/mpegpicture: Split MPVPicture into WorkPicture and ordinary Pic
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:17501b22674a: avcodec/error_resilience: Deduplicate cleanup code
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:17b5fc2e51a3: avcodec/mpegvideo_enc: Factor setting length of B frame chain out
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:b64dfe2bd192: avcodec/mpegvideo_enc: Return early when getting length of B frame chain
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:dceb73a22d43: avcodec/mpegvideo_enc: Reindentation
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:183a67580b61: avcodec/mpeg12dec: Don't initialize inter tables for IPU
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:2f22fd7ec18e: avcodec/mpeg12dec: Only initialize IDCT for IPU
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:bbe10bcae84d: avcodec/mpeg12dec: Remove write-only assignment
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:5ea7c0e323b5: avcodec/mpeg12dec: Set out_format only once
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:b90b6764093a: avformat/riff: Declare VCR2 to be MPEG-2
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:99d26939af3f: avcodec/mpegvideo_dec: Add close function for mpegvideo-decoders
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:9ce56f91c0e7: avcodec/mpegpicture: Make MPVPicture refcounted
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:4ef98a43ee0e: avcodec/mpeg4videoenc: Avoid branch for writing stuffing
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:5475000942c8: avcodec/mpeg4videoenc: Simplify writing startcodes
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:45cf0541cf4b: avcodec/mpegpicture: Use ThreadProgress instead of ThreadFrame API
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:12fcbff44664: avcodec/mpegpicture: Avoid loop and branch when setting motion_val
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:d0f76e6a11a1: avcodec/mpegpicture: Use union for b_scratchpad and rd_scratchpad
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:a05eebee99f7: avcodec/mpegpicture: Avoid MotionEstContext in ff_mpeg_framesize_alloc()
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:4f2719a239e2: avcodec/mpegvideo_enc: Unify initializing PutBitContexts
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:7e61e77c4ecc: avcodec/mpeg12enc: Simplify writing startcodes
11:17
<
cone-605 >
ffmpeg Andreas Rheinhardt master:a13ca893cff5: avcodec/mpegvideo_dec, rv34: Simplify check for "does pic exist?"
11:18
<
cone-605 >
ffmpeg Andreas Rheinhardt master:c703a519a57b: avcodec/mpegvideo_dec: Don't sync encoder-only coded_picture_number
11:18
<
cone-605 >
ffmpeg Andreas Rheinhardt master:9bb35cfa4d7f: avcodec/mpeg12dec: Pass Mpeg1Context* in mpeg_field_start()
11:18
<
cone-605 >
ffmpeg Andreas Rheinhardt master:1f5c64cbdc9e: avcodec/mpeg12dec: Don't initialize inter_scantable
11:18
<
cone-605 >
ffmpeg Andreas Rheinhardt master:2f6768c761eb: avcodec/mpegvideo: Remove pblocks
11:18
<
cone-605 >
ffmpeg Andreas Rheinhardt master:b2ac79236738: avcodec/mpegvideo: Use enum for msmpeg4_version
11:18
<
cone-605 >
ffmpeg Andreas Rheinhardt master:34ed9d1954c1: avcodec/ituh263enc: Remove redundant check
11:18
<
cone-605 >
ffmpeg Andreas Rheinhardt master:bf087a7e2609: avcodec/mpegvideo_enc: Binarize reference
11:18
<
cone-605 >
ffmpeg Andreas Rheinhardt master:1fe2a97b6161: avcodec/vc1_pred: Fix indentation
11:44
Livio has joined #ffmpeg-devel
11:46
<
cone-605 >
ffmpeg Andreas Rheinhardt master:f0c021b6ffbb: avcodec/dnxhdenc: Remove always-false check
11:46
<
cone-605 >
ffmpeg Andreas Rheinhardt master:bf8208d0cd7c: avcodec/mpegvideo_enc: Make ff_dct_quantize_c() static
11:46
<
cone-605 >
ffmpeg Andreas Rheinhardt master:adb1521385e6: avcodec/mpegvideo_enc: Move initializing H263DSP to ituh263enc.c
11:46
<
cone-605 >
ffmpeg Andreas Rheinhardt master:832782820509: avcodec/mips/mpegvideo: Set denoise_dct only for encoder
11:46
<
cone-605 >
ffmpeg Andreas Rheinhardt master:e80bd068c519: avcodec/mpegvideo_enc: Always set c funcs in ff_dct_encode_init()
11:46
<
cone-605 >
ffmpeg Andreas Rheinhardt master:1dc1b6c5af62: avcodec/mpegvideo: Remove fast_dct_quantize
11:46
<
cone-605 >
ffmpeg Andreas Rheinhardt master:89742b88c609: avcodec/mpegvideo_enc: Don't pretend ff_dct_encode_init() can fail
12:31
<
llyyr >
elenril: thanks, tested and works
12:40
<
cone-605 >
ffmpeg Andreas Rheinhardt master:4c22b0078094: api/api-band-test: Remove write-only variable
12:40
<
cone-605 >
ffmpeg Andreas Rheinhardt master:b8cc5cf23322: api/api-band-test: Perform codec admissibility check earlier
12:40
<
cone-605 >
ffmpeg Andreas Rheinhardt master:edcfa846cfba: api/api-band-test: Query codec capabilities to decide admissibility
12:47
Krowl has joined #ffmpeg-devel
12:51
<
cone-605 >
ffmpeg Frank Plowman master:d72a5fe719c0: lavc/vvc: Prevent overflow in chroma QP derivation
13:02
jordimedia_ has joined #ffmpeg-devel
13:38
Krowl has quit [Read error: Connection reset by peer]
13:42
Krowl has joined #ffmpeg-devel
13:43
Krowl has quit [Read error: Connection reset by peer]
14:02
<
Marth64 >
i used ffprobe 7.0.1 last night, disposition was being reported correctly over here
14:03
<
Daemon404 >
yes, what was being noted was that every disposition is always shown (as 0)
14:03
<
Daemon404 >
rather than just the ones that are extant
14:04
<
Marth64 >
ahhh, yes, saw that indeed
14:07
Marth64 has quit [Remote host closed the connection]
14:07
Marth64 has joined #ffmpeg-devel
14:09
<
Marth64 >
i actually have an idea for a potential new disposition but haven't had a chance to craft my thoughts to the ML yet on it
14:10
<
Daemon404 >
of course we can only fit a few more dispositions
14:10
<
Daemon404 >
because in classic old API fasion, it's a bitmask in an int
14:11
Krowl has joined #ffmpeg-devel
14:11
<
Marth64 >
seen a lot of subtitles which are summarized, simplified language, or "kid-friendly". in DVD it is actually a disposition
14:20
Livio has quit [Ping timeout: 264 seconds]
14:22
Livio has joined #ffmpeg-devel
14:27
Marth64[m] has joined #ffmpeg-devel
14:29
Marth64 has quit [Ping timeout: 255 seconds]
14:41
Sean_McG has quit [Quit: leaving]
14:41
<
courmisch >
niiice, FATE went ENOSPC so it reports all tests successful now
15:09
<
jamrial >
you can't get failures if nothing runs
*taps forehead*
15:11
mkver has quit [Ping timeout: 264 seconds]
15:15
blb has quit [Quit: brb]
15:23
blb has joined #ffmpeg-devel
15:26
mkver has joined #ffmpeg-devel
15:29
Krowl has quit [Read error: Connection reset by peer]
15:38
<
cone-605 >
ffmpeg sunyuechi master:8d9fb7b5cf0a: lavc/vp8dsp: R-V V put_bilin_h v unroll
15:38
<
cone-605 >
ffmpeg sunyuechi master:a4901a56c653: lavc/vp8dsp: R-V V bilin_load to bilin_load_h
16:01
mkver has quit [Ping timeout: 268 seconds]
16:39
AbleBacon has joined #ffmpeg-devel
16:47
Livio has quit [Ping timeout: 255 seconds]
16:49
Livio has joined #ffmpeg-devel
16:50
System_Error has quit [Remote host closed the connection]
17:02
System_Error has joined #ffmpeg-devel
17:42
<
cone-605 >
ffmpeg Marton Balint master:e405afdd0a27: avformat/mov_chan: use the newly added channel ids for more exact mapping
17:42
<
cone-605 >
ffmpeg Marton Balint master:d9a0a6a2b5e3: avformat/mov_chan: make iso_channel_position table more compact
17:42
<
cone-605 >
ffmpeg Marton Balint master:9fb6f024c3db: avformat/mxfdec: use the newly added channel ids for more exact mapping
17:47
j45_ has joined #ffmpeg-devel
17:49
j45 has quit [Ping timeout: 268 seconds]
17:49
j45_ is now known as j45
17:49
j45 has quit [Changing host]
17:49
j45 has joined #ffmpeg-devel
17:51
System_Error has quit [Remote host closed the connection]
18:04
Krowl has joined #ffmpeg-devel
18:06
tmatth_ has joined #ffmpeg-devel
18:06
tmatth has quit [Ping timeout: 256 seconds]
18:06
tmatth_ is now known as tmatth
18:07
sfan5_ is now known as sfan5
18:13
System_Error has joined #ffmpeg-devel
18:56
<
courmisch >
jamrial: would it be faster to use the trick from neon/ (saving and restoring block[0] to preserve alignment)?
18:58
Krowl has quit [Read error: Connection reset by peer]
18:58
<
courmisch >
jamrial: I guess not because you'd just move the problem from head to tail?
18:59
<
courmisch >
jamrial: in any case, I'd purge the MMX while you're at it
19:00
<
courmisch >
and by MMX, I mean the DCT, no the loop filter
19:13
<
ramiro >
do I understand correctly that we're trying to get rid of mmx/mmxext code?
19:13
<
ramiro >
did we settle on what other minimum must be implemented? for example, to remove mmx/mmxext in libswscale/x86/yuv2rgb, should we implement sse2 instead?
19:13
<
courmisch >
I think most people want to get rid of MMX/MMXEXT
19:14
<
courmisch >
there seems to be some argument whether the baseline should be SSE2 or SSSE3
19:28
<
kepstin >
oh no, but ffmpeg on my k6-3 box… (I'm joking, please don't actually care about running modern ffmpeg on my k6-3 box)
19:30
<
Lynne >
we still keep hacks for athlon in our asm
19:30
<
Lynne >
see REP_RET
19:31
<
Lynne >
for some reason going from a conditional jump straight into a ret is slow, but if you add a few NOPs it becomes magically faster
19:31
<
courmisch >
we should deprecate all SIMDs and switch to WASM anyway
19:31
<
courmisch >
compile-once run-anywhere FFmpeg
19:31
<
courmisch >
(where anywhere is a browser)
19:32
<
Lynne >
remind me how did loops and conditional blocks work again in wasm, and the reason why?
19:32
<
courmisch >
I have not studied WASM byte code yet
19:34
<
Lynne >
google's v8 compiler was designed back to front, so as a compromise, they wrote the syntax back to front, so you have horrendus compiler-only write-only syntax with more blocks than elisp has brackets
19:52
Krowl has joined #ffmpeg-devel
19:54
<
courmisch >
SpacemiT specs go all the way to 4096-bit vectors. Somebody is ambitious
19:55
<
courmisch >
kernel is going to hate your hardware
19:55
<
courmisch >
16 KiB of vector to save/restore on context switch
20:01
<
courmisch >
matrix dot product just confusing wording for matrix product?
20:04
iive has joined #ffmpeg-devel
20:13
System_Error has quit [Remote host closed the connection]
20:18
System_Error has joined #ffmpeg-devel
20:19
<
Lynne >
modern GPUs have hundreds of megabytes to save/restore on context switches
20:21
<
courmisch >
it doesn't count if it's not in registers
20:21
<
Lynne >
its all registers
20:21
<
Lynne >
they actually started to do context switching these days, rather than letting a compute task freeze your desktop, but it involves moving task submission to the firmware rather than kernel
20:22
<
Lynne >
they mainly did this for virtualization reasons though, no point in wasting a whole GPU on a VM which doesn't do anything with it
20:27
wyatt8750 has joined #ffmpeg-devel
20:28
wyatt8740 has quit [Ping timeout: 255 seconds]
20:29
Krowl has quit [Read error: Connection reset by peer]
20:42
cone-605 has quit [Quit: transmission timeout]
20:59
j45 has joined #ffmpeg-devel
20:59
j45 has quit [Changing host]
20:59
j45 has joined #ffmpeg-devel
21:10
s55 has joined #ffmpeg-devel
21:38
System_Error has quit [Remote host closed the connection]
21:43
System_Error has joined #ffmpeg-devel
21:43
ccawley2011 has quit [Read error: Connection reset by peer]
21:44
System_Error has quit [Read error: Connection reset by peer]
21:47
System_Error has joined #ffmpeg-devel
21:58
Marth64[m] has quit [Remote host closed the connection]
21:58
Marth64[m] has joined #ffmpeg-devel
22:33
Livio has quit [Ping timeout: 268 seconds]
22:42
cone-409 has joined #ffmpeg-devel
22:42
<
cone-409 >
ffmpeg Michael Niedermayer master:959cb2e2e36c: doc/developer: Provide information about git send-email and gmail
22:51
witchymary has quit [Remote host closed the connection]
22:52
witchymary has joined #ffmpeg-devel
23:15
Gramner has quit [Remote host closed the connection]
23:16
markh has quit [Ping timeout: 246 seconds]
23:22
markh has joined #ffmpeg-devel
23:32
Gramner has joined #ffmpeg-devel
23:45
IndecisiveTurtle has joined #ffmpeg-devel
23:55
iive has quit [Quit: They came for me...]