| 00:00:47 | aaabbb | oh i see. it looks like that is mostly about the ogg demuxer |
| 00:01:13 | aaabbb | i notice a few other optimizations that were lost but not put back in, like https://github.com/Rockbox/rockbox/commit/c8e7dae362517489dc803a2ac3279aacefe0c3bc |
| 00:07:24 | aaabbb | i wonder if pgo would be useful for decoders |
| 00:20:13 | _bilgus_ | devils in the details I find it difficult pulling stuff far enough out without rewriting half the interface in the process to be able to feed into tools but the stuff that does the whole thing works nicer, asan, ubsan, coverity, valgrind |
| 00:21:15 | _bilgus_ | nice thing about the plugins is they are containerized and easy to run standalone |
| 00:21:24 | _bilgus_ | or easier.. |
| 00:29:08 | *** | Saving seen data "./dancer.seen" |
| 02:00 |
| 02:29:12 | *** | No seen item changed, no save performed. |
| 03:00 |
| 03:33:27 | | Quit buZz (Ping timeout: 244 seconds) |
| 03:35:11 | | Join buZz [0] (~buzz@23.163.0.47) |
| 03:35:35 | | Nick buZz is now known as Guest6010 (~buzz@23.163.0.47) |
| 03:45:49 | | Join OlsroFR [0] (~OlsroFR@user/OlsroFR) |
| 03:48:22 | OlsroFR | chris_s https://gerrit.rockbox.org/r/c/rockbox/+/6031 https://gerrit.rockbox.org/r/c/rockbox/+/6032 I just pushed updates to do the renames "albums" -> "album" in lang files |
| 03:48:44 | | Quit OlsroFR (Client Quit) |
| 03:50:41 | | Join OlsroFR [0] (~OlsroFR@user/OlsroFR) |
| 03:52:17 | OlsroFR | All my other pending patches: https://gerrit.rockbox.org/r/c/rockbox/+/6384 https://gerrit.rockbox.org/r/c/rockbox/+/6375 https://gerrit.rockbox.org/r/c/rockbox/+/6368 |
| 03:52:36 | | Quit OlsroFR (Client Quit) |
| 03:57:22 | | Quit winsim-buildbot (Ping timeout: 272 seconds) |
| 04:00 |
| 04:04:18 | | Quit jacobk (Ping timeout: 248 seconds) |
| 04:06:04 | | Join jacobk [0] (~quassel@47-186-65-73.dlls.tx.frontiernet.net) |
| 04:29:16 | *** | Saving seen data "./dancer.seen" |
| 04:34:54 | | Quit othello7 (Ping timeout: 252 seconds) |
| 05:00 |
| 05:27:52 | | Quit PheralSparky (Read error: Connection reset by peer) |
| 06:00 |
| 06:22:09 | | Quit user890104_ (Ping timeout: 260 seconds) |
| 06:29:20 | *** | Saving seen data "./dancer.seen" |
| 08:00 |
| 08:29:22 | *** | No seen item changed, no save performed. |
| 09:00 |
| 09:26:41 | speachy | aaabbb: that "optimization" you identitified was actually dropped over a decade ago. it may or may not have been intentional, but upstream did change that code around a lot (adding arch-specific optimised versions). |
| 09:27:39 | aaabbb | speachy: i've been looking through other optimizations and found a few that were dropped (some of which did get shuffled around like b6bcb1338e06c3cfcc7e844dd248c332743de5c6 which moved X and freq in celt.c into iram), but others like f636aa07dfb541b5828d44e8168e7b36e5ad8898 came from code that was more or less not touched by upstream |
| 09:27:40 | speachy | my recent efforts were purely versus the pre-3.15 state of affairs, and all of the stuff accidently dropped in late 2019. |
| 09:28:26 | speachy | anything further at this point needs to be backed up by benchmarks. |
| 09:29:32 | speachy | keep in mind that IRAM is a very limited resource. |
| 09:41:19 | | Quit speachy (Quit: WeeChat 4.5.1) |
| 09:43:12 | | Join speachy [0] (~speachy@hurricane.shaftnet.org) |
| 09:43:12 | | Quit speachy (Changing host) |
| 09:43:12 | | Join speachy [0] (~speachy@rockbox/developer/speachy) |
| 09:43:12 | Mode | "#rockbox +v speachy" by ChanServ (ChanServ@services.libera.chat) |
| 09:44:17 | speachy | The opus codec is currently using about 39K. |
| 09:51:38 | speachy | aac is ~76K, mp3 is ~56K, flac is ~81K. On a mini2g. |
| 09:51:43 | speachy | (of IRAM) |
| 09:52:12 | aaabbb | the older commits did include benchmarks from what i could see in the commit message |
| 09:56:02 | speachy | benchmarks versus the then-state of the codebase. |
| 09:56:24 | speachy | since that commit, for example, there have been _three_ rebases against upstream |
| 09:58:45 | speachy | including one performed by the author of most of those optimizations (who proceeded to add more afterwards) |
| 10:00 |
| 10:05:35 | | Join user890104 [0] (~Venci@freemyipod/user890104) |
| 10:29:23 | *** | Saving seen data "./dancer.seen" |
| 10:37:09 | _bilgus_ | Figure I did similar comparisons when I moved to upstream I'm pretty sure we even discussed how we were about 10% slower but the issue was that it wouldn't fit as is and without a device to test be just a well to leave it alone.. Which is what we did :P |
| 10:39:44 | _bilgus_ | my big concern at the time were these devices that could barely achieve realtime but that is apparently very few |
| 10:43:06 | _bilgus_ | Anyone have Thoughts on writing a comment when the device is started to the playback log? |
| 10:45:43 | _bilgus_ | I've already implemented file splitting so you don't need to delete or export the file occasionally |
| 10:46:57 | speachy | it's good enough to achieve realtime on high bitrate files on the m68k targets, adn those are by far the most pathological. PP-based ipods are a little laggy in the UI but otherwise work ok. |
| 10:46:57 | _bilgus_ | I figure the user who was asking about keeping track of your daily plays could use it, although (in the RTC devices) you could just parse the track timestamps |
| 10:47:50 | speachy | _bilgus_: I think that's a good idea, fwiw. |
| 10:48:24 | _bilgus_ | I actually forgot it was running and I REALLY like it |
| 10:48:45 | _bilgus_ | doubt I'll start scrobbling anytime soon but I like the log |
| 10:50:10 | _bilgus_ | it'd be pretty neat to have a plugin parse them and tell you ststs over the last n months |
| 10:50:17 | _bilgus_ | status* |
| 10:51:02 | _bilgus_ | wonder if any of the scrobble clients have that option |
| 10:55:05 | speachy | I'd be surprised if not. |
| 10:56:03 | speachy | _bilgus_: wrt opus, most of that performance hit turned out to be the inadvertant stripping of the IRAM annotations. though on higher performance targets it made a <1% difference IIRC. |
| 10:59:29 | _bilgus_ | just having a device to test should help |
| 11:00 |
| 11:00:18 | _bilgus_ | that patch that claimed 1.1MHZ decrease would be my first stop if possible |
| 11:00:36 | _bilgus_ | it doesn't even sound like a lot but yeah THAT a lot |
| 11:00:50 | speachy | I have an ihp3xx but it's d-e-d. |
| 11:01:20 | speachy | the fact it took five years for someone to complain about the perf hit though. :) |
| 11:01:25 | _bilgus_ | . |
| 11:02:42 | _bilgus_ | split up between devices and codecs though its a small pool for overlap |
| 11:03:27 | _bilgus_ | WAG 1% for the device and 33% for the codec? |
| 11:04:31 | speachy | more likely that most folks with these ancient devices don't care about newfangled codecs. |
| 11:06:01 | _bilgus_ | one thing is though and it might not be the upstream code as much as the lower stack we have not had anymore opus crashes |
| 11:07:28 | _bilgus_ | That I'm aware of* :P |
| 11:09:58 | speachy | kepping that stuff off the stack is a good thing but it may be worth putting that evicted stuff into IRAM. |
| 11:10:46 | speachy | I do need to take apart this thing and see if I can recussitate it. it accepts a charge, warms up, but won't power on. |
| 11:37:44 | | Quit jacobk (Ping timeout: 260 seconds) |
| 11:39:05 | | Join jacobk [0] (~quassel@47-186-65-73.dlls.tx.frontiernet.net) |
| 11:48:16 | | Quit rogeliodh (Quit: The Lounge - https://thelounge.chat) |
| 11:48:38 | | Join rogeliodh [0] (~rogeliodh@rogeliodh.dev) |
| 12:00 |
| 12:08:21 | | Quit jacobk (Ping timeout: 276 seconds) |
| 12:09:28 | | Join jacobk [0] (~quassel@47-186-65-73.dlls.tx.frontiernet.net) |
| 12:29:27 | *** | Saving seen data "./dancer.seen" |
| 13:00 |
| 13:06:46 | | Nick Guest6010 is now known as buZz (~buzz@23.163.0.47) |
| 13:18:09 | | Join othello7 [0] (~Thunderbi@pool-100-36-176-164.washdc.fios.verizon.net) |
| 14:00 |
| 14:29:30 | *** | No seen item changed, no save performed. |
| 16:00 |
| 16:22:35 | | Quit baltazar (Quit: leaving) |
| 16:29:31 | *** | Saving seen data "./dancer.seen" |
| 16:57:32 | | Join baltazar [0] (~baltazar@user/baltazar) |
| 17:00 |
| 17:10:24 | | Join Moriar [0] (~moriar@107-200-193-159.lightspeed.stlsmo.sbcglobal.net) |
| 17:18:40 | | Quit cstine (Quit: The Lounge - https://thelounge.chat) |
| 17:19:08 | | Join cstine [0] (~cstine@150.136.136.191) |
| 18:00 |
| 18:29:33 | *** | Saving seen data "./dancer.seen" |
| 19:00 |
| 19:09:21 | | Join massive_H [0] (~massiveH@2600:4040:a982:5400:6445:6ff7:48df:9be0) |
| 19:25:48 | | Quit massive_H (Quit: Leaving) |
| 20:00 |
| 20:29:37 | *** | Saving seen data "./dancer.seen" |
| 21:00 |
| 21:18:44 | aaabbb | speachy: are you sure it's accepting a charge then? the battery could be basically dead, "accepting" a charge, might be the linear regulator for the battery charger warming up from the voltage drop |
| 21:19:45 | aaabbb | from an internal short somewhere. maybe try with a different battery or fake battery |
| 21:24:51 | | Join PheralSparky [0] (~S|h|a|w|n@user/shawn/x-4432647) |
| 21:47:24 | | Join chris_s [0] (~chris_s@2a09:bac2:2d18:246e::3a1:64) |
| 21:53:10 | chris_s | _bilgus_: https://git.rockbox.org/cgit/rockbox.git/diff/apps/tree.c?id=f3a858e16c has introduced a regression where you are returned to the wrong folder after leaving and returning to the file browser if a folder (instead of a file) was last selected, due to the change to set_current_file_ex |
| 22:00 |
| 22:00:13 | | Quit chris_s (Quit: Client closed) |
| 22:29:38 | *** | Saving seen data "./dancer.seen" |
| 23:00 |
| 23:45:46 | _bilgus_ | chris_s are you sure I just reverted it and still have the issue |
icon identifies that the person is a core developer (has commit access).