Previous day | Jump to hour: 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23 | Next day

Seconds: Show Hide | Joins: Show Hide | View raw
Font: Serif Sans-Serif Monospace | Size: Small Medium Large

Click in the nick column to highlight everything a person has said.
The Logo icon identifies that the person is a core developer (has commit access).

#rockbox log for 2023-08-24

00:22:08Nyaadid test codec thing for opus on ipodvideo
00:22:15Nyaawithout dsp - 122.03% realtime
00:22:21Nyaawith dsp - 108.63% realtime
00:23:13Nyaaso, not a whole lot of overhead there
00:29:57 Join user890104 [0] (~Venci@freemyipod/user890104)
00:31:26 Join spork_ [0] (
00:35:29 Quit spork (*.net *.split)
00:35:29 Quit user890104_ (*.net *.split)
00:37:17 Quit m01 (Quit: Konversation terminated.)
00:39:59 Join m01 [0] (
01:03:36 Join advcomp2019_ [0] (~advcomp20@user/advcomp2019)
01:05:28 Quit advcomp2019 (Ping timeout: 256 seconds)
01:21:05 Nick spork_ is now known as spork (
01:57:46***Saving seen data "./dancer.seen"
01:58:58 Join advcomp2019 [0] (~advcomp20@user/advcomp2019)
02:01:56 Quit advcomp2019_ (Ping timeout: 260 seconds)
02:59:12 Quit CH23_M (Ping timeout: 252 seconds)
02:59:31 Join CH23_M [0] (~CH23@revspace/participant/ch23)
03:18:22 Quit CH23_M (Read error: Connection reset by peer)
03:18:40 Join CH23_M [0] (~CH23@revspace/participant/ch23)
03:57:50***Saving seen data "./dancer.seen"
05:57:52***No seen item changed, no save performed.
06:11:41 Quit jacobk (Quit: No Ping reply in 180 seconds.)
06:18:50 Join jacobk [0] (~quassel@
07:11:32buZzNyaa: decoding is faster without hw acceleration?
07:11:58buZzor is that percentage 'amount of time to decode' :)
07:52:52 Quit sebagala (Ping timeout: 245 seconds)
07:53:08 Join sebagala [0] (~Burak@
07:57:54***Saving seen data "./dancer.seen"
08:14:11speachybuZz: if "100%" means we can (just barely) decode it fast enough for realtime playback, lower numbers means it's going to stutter (at best) while higher numbers means there's spare CPU cycles for something other than decoding the audio. Like updating the UI, or possibly even going into lower-power mode.
08:15:21speachyfor comparison, IIRC xDuoo X3 can decode opus at nearly 6x realtime.
08:15:30buZzisnt DSP using hw decoding?
08:19:13gevaerts"dsp" here means effects like equaliser
08:19:36buZzso, software effects
08:19:48speachyequalizer, playback speed/pitch control, etc.
08:21:39gevaertsRockbox doesn't do any hardware decoding, except on the (very) old Archos hardware
08:37:12 Quit sebagala (Ping timeout: 240 seconds)
08:47:00_bilgusand we ripped it out
08:48:18gevaertsWell, yes and no. There hasn't been a new release yet since then!
08:48:37_bilgusthe DSP stuff uses quite a bit of CPU figure even on the devices that do OPUS pretty ok start adding echo and signal processing and your battery life is noticably affected
08:51:42_bilgusI remeber this convo with speachy about upstream opus pushing underpowered devices ever closer to the brink
08:56:14_bilguswe still pushed it 2 years later but I don't remember the specific reason that said Nyaa I'd try the downstream we had prior to see what kind of an increase you get on that hardware vs what we were testing [libopus Sansa Clip+ (Head) Desperation - Opus 64k 056.35 347.69 617.01% 031.11 ; libopus Sansa Clip+ (upstream) Desperation - Opus 64k 061.68 347.69 563.69% 034.06]
08:57:44_bilgusmind you upstream there ^ is current HEAD
09:01:41_bilgusgevaerts, noted.
09:02:09_bilgusback to FAT32 and long latencies wouldn't multiple partitions solve most of those issues?
09:02:52_bilguswe have support for that in the codebase already
09:04:09_bilgusI think the main issue with FAT32 is windows won't do the format because microsoft has stake in making exfat defacto
09:04:19_bilgusso meh Fuck MS lets do FAT32
09:04:46_bilgusyou already need tons of 3rd party tools to windows effectively whats one more?
09:05:58 Join sebagala [0] (~Burak@
09:11:06 Quit sebagala (Ping timeout: 256 seconds)
09:14:48 Join sebagala [0] (~Burak@
09:57:55***Saving seen data "./dancer.seen"
10:11:59kirvesAxe_bilgus_, well yeah, M$ is what it is... although I was a bit surprised when they published the specs for exFAT in 2019 (probably they finally realized its usage will die otherwise lol)
10:15:39kirvesAxeand yeah multiple partitions might solve some problems. but they would create others. It would be a horrible PITA to have to organize a large music collection and the playlists with that :)
10:40:49speachythere's nothing wrong with exFAT technically. or legally, any more.
10:42:27speachybut if rockbox gets exFAT support, most players we support will lose the ability to use the original device firmware.
10:43:12speachy(ie if we move to GPT partitioning and exFAT root)
11:15:24 Quit jacobk (Ping timeout: 248 seconds)
11:33:22 Join jacobk [0] (~quassel@
11:57:34 Join Schimon [0] (
11:57:58***Saving seen data "./dancer.seen"
12:25:17 Quit jacobk (Ping timeout: 248 seconds)
12:39:46 Join lebellium [0] (
12:40:34 Join paulk-bis [0] (
12:40:47 Quit paulk (Ping timeout: 244 seconds)
13:28:41 Quit CH23_M (Read error: Connection reset by peer)
13:28:51 Join CH23_M [0] (~CH23@revspace/participant/ch23)
13:29:00 Quit CH23_M (Read error: Connection reset by peer)
13:29:15 Join CH23_M [0] (~CH23@revspace/participant/ch23)
13:38:18 Quit CH23_M (Ping timeout: 252 seconds)
13:39:32 Join CH23_M [0] (~CH23@revspace/participant/ch23)
13:43:42 Quit CH23_M (Read error: Connection reset by peer)
13:43:56 Join CH23_M [0] (~CH23@revspace/participant/ch23)
13:58:02***Saving seen data "./dancer.seen"
13:59:09 Join jacobk [0] (~quassel@
14:25:02kirvesAxespeachy, well that might definately be a bad tradeoff in some cases :)
14:29:16Nyaai mean if you take a fuck microsoft stance just do ext2 :p
14:34:39*Nyaa wonders how difficult it would be to implement an MTP usb mode
14:34:54Nyaathen aside from OF support it wouldn't matter what filesystem is used
14:35:12Nyaa[at least for transferring stuff to/from it]
14:36:55Nyaawould also, in theory, allow for playing music at the same time you're doing usb stuff, though a lot of devices probably don't have enough memory/resources for that\
14:48:46 Join bilgus_ph [0] (~bilgus_ph@
14:51:09bilgus_phKirvesAxe i remember that but wasn't it only to OIC members though? As in still proprietary unless you join their cabal
14:53:58 Quit bilgus_ph (Quit: Connection closed)
15:01:15lebellium@Mt83 (logs) : regarding SonyNWDestTool for S615F : I used to work with paumary on that tool but that was several years ago and none of us is still really involved in the project.
15:01:15lebelliumHe was basically doing all the tech/dev thing and I was assisting him with "QA" (I have a huge DAPs collection) and maintaining the wiki page.
15:01:15lebelliumI just tried with my NWZ-S615F and I can confirm I get the same error (Cannot read node 'shp') as you do with your Japanese NW.
15:01:15DBUGEnqueued KICK lebellium
15:01:15lebelliumSearching through ICR logs, you can find another user getting the same issue back to 2021:
15:01:18lebellium (12:17): I also got an error with the S615 and pamaury replied it used a too old Linux kernel to be supported
15:01:21lebelliumMeanwhile he extracted the NVP map from each Sony player's Linux kernel (root/utils/nwztools/database/nvp), including the S615, that's why this device may seem to be supported in the tool and why the error message is a bit different ("This device doesn't have node 'shp'" vs "Cannot read node 'shp') but as despite the known NVP map we are not able to read the 'shp' node from the device, we can't do anything.
15:01:26lebelliumI'm afraid there is no solution here.
15:13:27kirvesAxe_bilgus_, actually you might be right, it might actually even need a licensing fee B$
15:16:53Nyaathe trick is to just not be in europe where software patents are invalid lol
15:17:03Nyaathe trick is to be in europe*
15:17:08Nyaaor not in the usa, more specifically
15:19:43kirvesAxeOnly sharing exFAT-compatible rockbox from europe-based servers? lol
15:20:17Nyaamost countries don't consider software patents valid lol
15:21:09 Quit jacobk (Ping timeout: 252 seconds)
15:21:24Nyaathey treat software like maths
15:22:20Nyaaa specific implementation can be copyright, but the method itself doesn't have any protections
15:24:46 Quit othello7 (Quit: othello7)
15:31:01Nyaatbf some of the codecs shipped in rockbox are also covered by software patents, no?
15:36:46buZzlikely even .wav is patented
15:36:58Nyaathe patent is probably expired by now
15:37:01buZzpatents dont really limit using tech though :)
15:37:11Nyaalike how mp3's patent expired
15:37:11buZzunless you plan on selling rockbox
15:37:35buZzi dont believe mpeg1 was patented?
15:38:16buZzmaybe some en/decoder implementation though
15:39:23NyaabuZz, mmm, looks like US courts have deemed noncommercial use also falls under the patents
15:39:42buZzwell, ~99.6% of nations globally arent the US ;)
15:39:58Nyaayes but the us is also basically the only place that has software patents
15:40:10Nyaa[japan does too but they're far more limited than the usa's]
15:40:11buZzeasiest to ignore then :D
15:41:40Nyaaiirc, japan's software patents are limited to the case where such software ties with a specific piece of hardware
15:46:15Nyaaah, the UK also seems to have software patents, but aside from that, doesn't look like a single software patent has been successfully enforced in the EU
15:46:57Nyaa[tricking the patent office into issuing a patent by phrasing your patent funny, and being able to actually enforce it, are different things]
15:47:10sporkthere are much bigger and richer targets than a nice jukebox software for abandonned hardware
15:47:22Nyaayeah probably, you never know though
15:47:40Nyaaa company could pull a nintendo and decide to sue you anyway
15:48:40sporkmake it a plugin
15:50:36speachyexFAT is still patented, but effectively (if not explicitly) license/royalty free due to MS's actions.
15:50:52speachyas for ext2, it's a LOT more complicated than FAT32.
15:51:41buZzpatenting something doesnt explicitly prevent people from using it
15:51:42speachyand there's the beginning of an MTP implementation in gerrit, would be a good starting place.
15:52:59speachypreventing others from using it (without permission, which usually means royalties) is the entire point of patents.
15:53:36speachy(well, "to promote progress in the arts" is the "point" of patents, but "exclusive use" is the mechanism)
15:54:24speachyof course, patents eventually expire.
15:54:37Nyaawonder if it would be possible to make filesystem stuff also loadable via a plugin-like system
15:54:49speachyand the patent holder doesn't _have_ to enforce it offensively.
15:55:06speachy(unlike trademarks, there's no "use it or lose it" principle)
15:55:57speachypossible, of course. worth the effort? I highly doubt it.
15:56:14speachywhat's the benefit?
15:56:47speachy(especially considering that, for example, our bootloaders need to be able to access the filesystems too)
15:57:07Nyaaright, but it could be done in a similar method to how EFI does it
15:57:19Nyaawhere you have a fat32 partition the bootloader loads the next stage of stuff out of
15:58:04***Saving seen data "./dancer.seen"
15:58:39 Join jacobk [0] (
16:02:30Nyaa[plus having a bootloader partition would open up opportunity for having a configurable bootloader menu, easily booting multiple versions of rockbox, etc]
16:03:23speachyat that point we might as well just use Linux.
16:04:20Nyaai mean, linux would probably handle threading better lol
16:04:22speachyand stop reinventing the wheel
16:04:49buZzdoesnt rockbox already target linux?
16:04:58buZzas standalone SDL application or something
16:05:27speachyour crappiest (ie arm7tdmi/arm9tdmi) targets are mostly SOL, but everything else has the capabilities to run a full OS with an MMU etc.
16:05:59Nyaai mean µClinux exists
16:06:01buZzarmv5 without mmu should be able to run uClinux just fine
16:06:14speachyyeah, we can run as a native Linux application. We also have an SDL application, but that's not used on actual player hardware normally.
16:06:28buZzthusfar ;)
16:06:40speachyuClinux is not worth the effort here, as it's IMO the worst of both worlds.
16:07:36speachythusfar because none of our "hosted" targets actually include SDL. We directly use the linux APIs instead of having to deal with adding SDL on top of that too.
16:07:47Nyaaprobably a good idea lol
16:09:00speachyuClinux gives us drivers and tht sort of thing but "userspace" is a complete cluster-f
16:10:08speachyif we wanted to scrap the linux builds on these players and roll our own, we could end upw ith a much nicer system overall. Except for teh minor detail that there's no kernel sources for any of these things and we'd therefore have to port Linux ourselves.
16:10:46speachy(at _best_ there's an ancient kernel drop −− I think Sony was generally GPL compliant −− but the rest... nope)
16:10:48Nyaaoh wait upstream linux can already be compiled without using mmu
16:11:41speachyyes, but uclinux (ie mmu-less linux) has some major restritions on what "userspace" entails. binaries have to be specially compiled for it, lots of restritions, and no executable relocation is possible.
16:12:13Nyaayeah but if you're going to only run a single application [rockbox] that probably doesn't matter
16:12:42speachywe still ahve to deal with the overhead even if it's just one application.
16:13:11speachy(overhead of maintaining our own complete "userspace", that is)
16:13:46Nyaaisn't there already a "userspace" of sorts
16:14:10speachyno such thing as generic userspace. it owuld need to be compiled for each target, based on how we've configured the kernel etc.
16:15:33Nyaahmm, just compile as a kernel module, problem solved /s
16:15:41speachy(by that I mean libc (+ anything else we need)
16:16:42Nyaadon't need to care about userspace if you just run in kernelspace
16:18:49speachyeither way the code has to be written.
16:20:13speachyand by doing that you lose the advantage of readily-available libraries. Plus you still need some minimal userspace, if only an init process.
16:21:16speachygiven the amount of effort (vs return) there are much better uses of time, but far be it for me to say what one should or shouldn't do.
16:44:01 Quit jacobk (Ping timeout: 258 seconds)
16:49:45Nyaai wonder if upstream opus would care about the performance on armv4 targets
16:52:14 Join jacobk [0] (~quassel@
16:59:38Nyaaalso don't all ARM7DMI chips that have a cache also have an MMU?
17:05:23CtcpIgnored 1 channel CTCP requests in 0 seconds at the last flood
17:05:23*Nyaa wonders
17:06:13Nyaaon the ipod targets, are codecs running out of SDRAM or IRAM?
17:26:37 Join othello7 [0] (
17:46:14Nyaai changed one option in the makefile and ran test again and got 120% realtime with dsp
17:46:18*Nyaa unsets it and tests again
17:51:27 Quit lebellium (Quit: Leaving)
17:58:06***Saving seen data "./dancer.seen"
18:21:00 Quit jacobk (Ping timeout: 252 seconds)
18:21:36 Join jacobk [0] (
19:33:36 Quit jacobk (Ping timeout: 252 seconds)
19:56:17_bilgusbe mindful of running stuff in the BGD i usually boot a clean install no db only the settings being tested and just the test files
19:57:01Nyaayeah i can't seem to get numbers that bad now lol
19:57:38Nyaathough it would make more sense to do tests with background stuff given that's more representative of real world usage
19:58:07***Saving seen data "./dancer.seen"
19:58:10_bilgusonly if you can replicate them
19:59:42Nyaathat's the thing about real world usage, you'll never get the same results twice :p
19:59:54Nyaabut you can repeat the test multiple times and do a statistical analysis
20:02:01_bilgusor just do it clean and assume you need 150 to be safe
20:02:48Nyaatime to mod the ipod to use the case for a heatsink and just overclock it past 80mhz lol
20:03:28_bilgusworked for my ti-86
20:03:38Nyaai have one of those lol
20:03:43Nyaaalso a ti-85 and a ti-81
20:03:50_bilgusbest eng calc ever
20:04:00Nyaathe 85 is dead though [was when i got it lol]
20:04:04Nyaabeen meaning to fix it
20:04:31Nyaamild water damage and probably a dead fuse or zener diode somewhere
20:04:57_bilguslikely bad ribbon cable water will corrode them off the board
20:05:20Nyaadon't remember seeing any damage near ribbon cables
20:05:24Nyaaonly near battery compartment
21:37:49_bilgusNyaa, id try adding back some of the ICONST tags and such and see if it fits and if it makes it faster
21:47:46Nyaaputting decoding look in IRAM would definitely help on pp5002 given hardware bug but idk if it would help on pp5020/pp5022
21:48:51Nyaawith pp5002 anything running from sdram is effectively running at half clock rate due to said bug
21:58:10***No seen item changed, no save performed.
22:01:15_bilgusprobably explains the usual ifdefs around those pp50xx
22:20:20_bilgusfinal one
22:35:08Nyaait's the SILK code mostly
22:35:20Nyaaencoded celt-only results in ~160% realtime with dsp
22:36:05Nyaawell SILK/Hybrid are the slower ones
22:38:34NyaaSILK gets ~110% realtime with dsp, Hybrid gets ~120% realtime with dsp, and CELT gets ~160% realtime with dsp
22:53:21 Quit LjL (Quit: Scappò via con la paura di arrugginire. Il giornale di ieri lo dà morto arrugginito. I becchini ne raccolgono spesso fra la gente che si lascia piovere addosso.)
22:54:45 Join LjL [0] (~ljl@user/ljl)
23:34:38 Quit buZz (Ping timeout: 246 seconds)
23:34:46 Join buZz [0] (
23:58:13***Saving seen data "./dancer.seen"

Previous day | Next day