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).

Notice: Only Gecko based browsers prior to FF4 support the multipart/mixed "server push" method used by this log reader to auto-update. Since you do not appear to use such a browser, this page will simply show the current log, and not automatically update.

#rockbox log for 2015-02-14

00:01:13 Quit AlexP (Remote host closed the connection)
00:04:33 Join saratoga [0] (123e11e0@gateway/web/freenode/ip.
00:05:06saratogawhats the easiest way to remove a file from a gerrit commit?
00:06:36 Quit ZincAlloy (Quit: Leaving.)
00:06:42marazmake another commit that deletes it.
00:07:23marazI suppose that wasn't the answer you were looking for...
00:07:26saratogawon't that delete the file?
00:08:11marazI was supposing you added it by mistake.
00:08:22marazlike, a new file.
00:08:40foolshsaratoga: maybe you're looking for "git rm −−cached ./path/to/file" and then "git commit −−amend" perhaps?
00:09:13saratogai don't think that works either
00:09:26marazdid you already push the commit to a remote?
00:09:35saratogai want to remove the file from the gerrit task, not from version control
00:09:51gevaertsgit reset to clear the entire commit, then redo the commit and copy the gerrit id?
00:12:06saratogadoes reset undo the changes to the files locally?
00:12:36gevaertsIt shouldn't, IIRC
00:12:59marazit will
00:13:11*gevaerts tries
00:13:45maraza hard reset of a commit will, reset of staged changes won't
00:13:56gevaerts"git reset HEAD^1" just removes the actual commit but leaves the actual files
00:14:20marazit won't remove the commit either
00:14:25marazit is still in refog
00:14:48gevaertsWell yes, but that's fairly irrelevant here
00:16:21*maraz fails to see what this has to do with gerrit
00:17:10gevaertsthe gerrit bit means you want to keep it in a single commit and you want to keep the change id
00:20:37m4tpamaury: thanks ^
00:20:57saratogashould the gerrit task i'm trying to fix show up in git log?
00:21:29gevaertsIf you haven't done the reset yet, yes
00:21:32m4ti get a background static on my zen. i don't recall having that with stock :/
00:21:58m4ti wonder if i can go back to stock to test?
00:22:01gevaertsIf it's not there yet, you probably have to cherry-pick it first
00:22:02saratogai see the source code changes but not the commit?
00:22:41gevaertsWithout doing git reset first?
00:22:50saratogayeah i haven't done anything yet
00:24:18gevaertsDoes "git status" show the expected things about the branch you're on and things like that?>
00:24:19saratogaoh maybe i lost them somehow
00:24:33saratogamy tree looks like a mix of different things
00:25:33saratogai did do one thing, "git branch testing" then "git checkout testing"
00:25:47saratogamy theory was that this would take what i had and give me a way to recover it if i screwed this up
00:25:54gevaertsI usually only gets stuck with git if I do things like forgetting I was halfway a rebase or things like that
00:25:56saratogabut it looks like it did something else
00:26:35gevaertsThat should have behaved as you expected...
00:27:49saratogai'll just reset and try again
00:28:26 Join man_alive [0] (
00:29:09 Quit Ivoah (Quit: Leaving...)
00:31:07man_aliveI'm using RB on a Sansa e200r and like (love) it but would like to use a more realistic battery algorithm
00:31:35man_aliveI don't think I want to compile myself and I wonder if a member of the community could help me out
00:31:40saratogacurrent algorithm just references a measured discharge curve for your player
00:32:20man_aliveYes - I've done a couple of discharges and the current algorithm is off by 20% at times
00:32:53man_aliveI've done a new algorithm but installing to compile will takes days when I only have a simple patch
00:33:24man_aliveI'm wondering if some kind soul who is already set to compile could do a patch compile for me
00:35:00man_alivecurrent default: 3300, 3680, 3740, 3760, 3780, 3810, 3870, 3930, 3970, 4070, 4160
00:35:00man_alivemy revised: 3300, 3709, 3776, 3815, 3825, 3844, 3876, 3946, 4000, 4056, 4160
00:36:30 Quit bertrik (Quit: Lost terminal)
00:36:53 Quit pamaury (Ping timeout: 252 seconds)
00:40:03saratogagevaerts: do you know much about the AS3514 or as3543?
00:40:25 Join Strife89 [0] (
00:42:29*gevaerts doesn't know much about SoC-level stuff
00:44:04saratogabertrik: would you look at ?
00:44:28saratogathis fixes a race condition on the ascodec write, but i'm curious if it should also be extended to the read?
00:54:45saratogai think its worth protecting them as well
00:58:35saratogai don't fully understand whats happening there, but my guess is that the backlight or another thread manages to corrupt the write by talking to the chip concurrently
00:58:58 Quit dfkt (Quit: Sic gorgiamus allos subjectatos nunc.)
01:09:28 Join AlexP [0] (~alex@rockbox/staff/AlexP)
01:14:31 Quit lebellium (Quit: ChatZilla [Firefox 36.0/20150209164123])
01:30:37***Saving seen data "./dancer.seen"
01:43:24 Quit AlexP (Remote host closed the connection)
02:07:59 Quit RiD (Quit: A good plan today is better than a perfect plan tomorrow.)
02:19:46 Quit mpmc (Remote host closed the connection)
03:00:55 Quit krabador (Remote host closed the connection)
03:01:25 Join krabador [0] (~krabador@unaffiliated/krabador)
03:05:27 Quit krabador (Client Quit)
03:08:58 Join krabador [0] (~krabador@unaffiliated/krabador)
03:16:51 Quit [Franklin] (Ping timeout: 244 seconds)
03:30:38***Saving seen data "./dancer.seen"
03:39:41 Join goom [0] (
04:00:33 Quit pixelma (Disconnected by services)
04:00:34 Join pixelma_ [0] (pixelma@rockbox/staff/pixelma)
04:00:36 Nick pixelma_ is now known as pixelma (pixelma@rockbox/staff/pixelma)
04:00:39 Join amiconn_ [0] (amiconn@rockbox/developer/amiconn)
04:00:39 Nick amiconn is now known as Guest91620 (amiconn@rockbox/developer/amiconn)
04:00:39 Nick amiconn_ is now known as amiconn (amiconn@rockbox/developer/amiconn)
04:03:12 Quit Guest91620 (Ping timeout: 245 seconds)
04:55:43 Quit krabador (Quit: Sto andando via)
05:01:22 Quit the-kyle (Ping timeout: 240 seconds)
05:26:22 Quit Provel (Quit: Provel)
05:30:42***Saving seen data "./dancer.seen"
05:43:56 Quit TheSeven (Ping timeout: 265 seconds)
05:45:23 Join TheSeven [0] (~quassel@rockbox/developer/TheSeven)
06:31:04 Quit goom (Quit: Leaving)
07:05:17 Join jhMikeS [0] (
07:05:18 Quit jhMikeS (Changing host)
07:05:18 Join jhMikeS [0] (~jethead71@rockbox/developer/jhMikeS)
07:05:29jhMikeSsaratoga: ping
07:30:44***Saving seen data "./dancer.seen"
07:47:36 Join saratoga_ [0] (123e1c40@gateway/web/freenode/ip.
07:47:57saratoga_jhMikeS: pong
07:48:57saratoga_jhMikeS:,48549.msg233945.html#msg233945 (several posts leading to that patch)
07:48:58jhMikeSwhat about that race condition?
07:51:39 Quit pixelma (Disconnected by services)
07:51:39 Join amiconn_ [0] (amiconn@rockbox/developer/amiconn)
07:51:40 Join pixelma_ [0] (pixelma@rockbox/staff/pixelma)
07:51:40 Nick amiconn is now known as Guest12646 (amiconn@rockbox/developer/amiconn)
07:51:40 Quit Guest12646 (Killed ( (Nickname regained by services)))
07:51:40 Nick amiconn_ is now known as amiconn (amiconn@rockbox/developer/amiconn)
07:51:42 Nick pixelma_ is now known as pixelma (pixelma@rockbox/staff/pixelma)
07:53:39jhMikeSsaratoga: it is protected from interrupts, any transfer by interrupt during the wait will be queued after the two consecutive PMU transfers
07:56:17saratoga_since that patch apparently makes a difference, there must be some kind of conflict if an interrupt happens before the write completes though?
07:56:17jhMikeSI wonder if it's done clocking in/out the last byte though by the time the next request arrives
08:01:54jhMikeSMight have to do with stacked interrupts
08:02:04 Quit Strife89 (Ping timeout: 244 seconds)
08:05:07 Quit saratoga (Ping timeout: 246 seconds)
08:06:12 Quit greatwolf (Ping timeout: 245 seconds)
08:07:02jhMikeSsaratoga: I think I see the actual race where a submission can happen while still in INT_I2C_AUDIO (never audited this part)
08:08:58 Quit saratoga_ (Ping timeout: 246 seconds)
08:17:02 Join Slasheri_ [0] (
08:20:36 Join saratoga [0] (123e1c40@gateway/web/freenode/ip.
08:22:45 Join greatwolf [0] (greatwolf@gateway/shell/panicbnc/x-saalcqnjvhtdjkjd)
08:24:43 Quit saratoga (Ping timeout: 246 seconds)
08:28:06 Quit Slasheri (*.net *.split)
08:28:06 Quit derf (*.net *.split)
08:28:07 Quit kugel (*.net *.split)
08:35:57 Join derf [0] (
08:35:57 Join kugel [0] (~kugel@rockbox/developer/kugel)
08:40:26 Join saratoga [0] (123e1c40@gateway/web/freenode/ip.
08:41:50saratogasorry, lost connectivity
09:05:26 Join petur [0] (~petur@rockbox/developer/petur)
09:18:13 Join Ivoah [0] (
09:20:22 Join wubking [0] (
09:21:21 Quit wubking (Remote host closed the connection)
09:25:04 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
09:30:48***Saving seen data "./dancer.seen"
09:34:04saratogajhMikeS: what else would be using the ascodec anyway? I see backlight using it, but that doesn't happen frequently enough
09:57:15 Quit petur (Remote host closed the connection)
09:57:54jhMikeSsaratoga: the interrupts are read regularly for various things
09:58:19saratogaso not just accesses to the pmu, but interrupts in general?
09:59:05jhMikeSyeah, charging, USB status, RTC and ADC
10:00:11saratogastrange that its not a problem on the clip zip then
10:01:04jhMikeSisn't that a different SoC?
10:01:32saratogano, same SOC
10:02:02saratogahardware is almost identical, although the fuze v2 does have a wheel which involves some extra logic
10:04:57jhMikeSIt seems it should wait for TX to be empty after sending the last byte. Right now it gives a "finished" right after writing the TX register but the byte won't have been clocked out before it tries the next one if the queue was emptied.
10:05:59 Quit pamaury (Ping timeout: 246 seconds)
10:12:19jhMikeSSome careful rearranging of ascodec_continue_req() is in order to do one more TXEMPTY after the last byte
10:33:29 Quit saratoga (Quit: Page closed)
10:38:42 Quit jhMikeS (Ping timeout: 245 seconds)
10:54:28 Join bertrik [0] (~quassel@rockbox/developer/bertrik)
10:58:33 Join ender` [0] (
11:30:50***Saving seen data "./dancer.seen"
11:33:37 Join lebellium [0] (
12:11:48 Join Provel [0] (
12:28:43 Join AlexP [0] (~alex@rockbox/staff/AlexP)
12:49:29 Quit fyre^OS (Ping timeout: 264 seconds)
12:49:39 Join fyre^OS [0] (
13:04:44 Quit AlexP (Remote host closed the connection)
13:07:24gevaertsIs that *really* a setting users should have?
13:07:46*gevaerts won't mention being unhappy about the USB mode settings that got smuggled in as well for ibasso
13:08:57 Join petur [0] (~petur@rockbox/developer/petur)
13:11:18 Join AlexP [0] (~alex@rockbox/staff/AlexP)
13:12:19 Join RiD [0] (
13:16:33 Join TheLemonMan [0] (~lemonboy@unaffiliated/thelemonman)
13:21:32 Quit kvieta (Ping timeout: 245 seconds)
13:30:53***Saving seen data "./dancer.seen"
13:40:10 Join rela [0] (~x@pdpc/supporter/active/rela)
13:53:27lebelliumcan someone tell me when and why did the setting "load to ram" for the database actually disappear?
13:55:29 Quit lbrown (Ping timeout: 265 seconds)
13:56:06gevaertslebellium: IIRC it vanished in jhMikeS's filesystem overhaul
13:57:13gevaertsI'm more interested in knowing why the german translation for the clip+ is gone on the builds on the server!
13:58:55lebelliumbecause Germans can speak english, they don't need a german translation
13:59:17lebelliumit's more useful to us French or Italian or Spanish people \o/
13:59:46gevaertsYes, but why the clip+ specifically?
14:02:01 Join lbrown [0] (
14:17:50 Quit AlexP (Remote host closed the connection)
14:22:33 Join AlexP [0] (~alex@rockbox/staff/AlexP)
14:48:13 Quit rela (Quit: Leaving)
14:52:49 Join kvieta [0] (~kvieta@
14:52:49 Quit kvieta (Excess Flood)
14:53:12 Join kvieta [0] (~kvieta@
15:04:47 Nick Guest75406 is now known as olspookishmagus (
15:10:24 Join y4n [0] (~y4n@unaffiliated/y4ndexx)
15:30:57***Saving seen data "./dancer.seen"
15:36:00 Quit bluebrother (Disconnected by services)
15:36:06 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
15:39:53 Quit fs-bluebot (Ping timeout: 265 seconds)
15:40:15 Join fs-bluebot [0] (
15:58:51 Join krabador [0] (~krabador@unaffiliated/krabador)
16:15:43 Quit Ivoah (Quit: Leaving...)
16:18:03 Join Strife89 [0] (
16:42:12 Quit krabador (Quit: Sto andando via)
16:42:38 Join nap [0] (
16:43:02 Nick nap is now known as Guest8772 (
16:43:15 Nick Guest8772 is now known as Napych (
16:44:47 Quit AlexP (Read error: Connection reset by peer)
16:45:02 Join AlexP [0] (~alex@rockbox/staff/AlexP)
16:45:23 Quit RiD (Quit: A good plan today is better than a perfect plan tomorrow.)
16:49:27 Quit krnlyng (Read error: Connection reset by peer)
16:50:13 Join krnlyng [0] (~liar@
16:53:11 Join RiD [0] (
17:00:16 Join [Franklin] [0] (~franklin@unaffiliated/franklin)
17:05:17 Quit TheLemonMan (Remote host closed the connection)
17:06:24 Quit RiD (Quit: A good plan today is better than a perfect plan tomorrow.)
17:30:59***Saving seen data "./dancer.seen"
17:42:19 Join jhMikeS [0] (
17:42:19 Quit jhMikeS (Changing host)
17:42:19 Join jhMikeS [0] (~jethead71@rockbox/developer/jhMikeS)
17:43:21 Quit Provel (Read error: Connection reset by peer)
17:43:32 Quit lebellium (Read error: Connection reset by peer)
17:43:46 Join Provel [0] (~Provel@
17:44:07 Join lebellium [0] (
17:50:37 Join Jinx [0] (Dojo@unaffiliated/jinx)
18:36:56 Join rela [0] (~x@pdpc/supporter/active/rela)
18:52:10 Join thomasjfox [0] (~thomasjfo@rockbox/developer/thomasjfox)
18:57:51 Join cmhobbs [0] (~cmhobbs@fsf/member/cmhobbs)
18:59:32 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
19:05:15 Quit rela (Ping timeout: 250 seconds)
19:07:57 Quit krnlyng (Remote host closed the connection)
19:11:08 Join krnlyng [0] (~liar@
19:16:59 Join rela [0] (~x@pdpc/supporter/active/rela)
19:21:44 Quit rela (Ping timeout: 265 seconds)
19:25:21 Quit petur (Quit: Leaving)
19:31:02***Saving seen data "./dancer.seen"
19:33:21 Join rela [0] (
19:33:29 Quit rela (Changing host)
19:33:29 Join rela [0] (~x@pdpc/supporter/active/rela)
19:37:02 Quit Strife89 (Ping timeout: 250 seconds)
19:37:45 Quit rela (Ping timeout: 250 seconds)
19:41:31 Quit kvieta (Ping timeout: 245 seconds)
19:47:09 Join kvieta [0] (~kvieta@
19:49:29 Join rela [0] (~x@pdpc/supporter/active/rela)
19:54:02 Join saratoga [0] (123e11e0@gateway/web/freenode/ip.
19:54:13 Quit rela (Ping timeout: 250 seconds)
19:54:45saratogajhMikeS: is there some way to measure the CPU time spent setting up DMA transfers without hacking up the code?
19:59:16saratogaanyone have an AMSv1 device handy?
20:05:49 Join rela [0] (~x@pdpc/supporter/active/rela)
20:06:19 Join dfkt [0] (~dfkt@unaffiliated/dfkt)
20:07:50 Quit thomasjfox (Quit: Konversation terminated!)
20:08:52 Join krabador [0] (~krabador@unaffiliated/krabador)
20:09:06jhMikeSsaratoga: in what exactly?
20:10:15 Quit rela (Ping timeout: 250 seconds)
20:21:55 Join rela [0] (~x@pdpc/supporter/active/rela)
20:26:59 Quit rela (Ping timeout: 265 seconds)
20:31:31saratogajhMikeS: AMSv2, i'm testing using larger DMA transfers for audio playback
20:33:15jhMikeSHow much larger? Too large and the mixer will get clunky.
20:34:26jhMikeSI don't think it's gonna spend much time setting anything up. It just set a few registers.
20:37:11saratogaDMA_S1 -> DMA_S16
20:37:13jhMikeSIf you mean the actual buffer, then it'll spend time proportional to the buffer size.
20:37:26saratogai'm assuming that means one word, but i'm still looking at the datasheet
20:37:35jhMikeSoh, that sort of thing
20:38:07 Join rela [0] (~x@pdpc/supporter/active/rela)
20:40:00saratogathe source/destination burst size
20:42:56 Quit rela (Ping timeout: 265 seconds)
20:44:02saratogadatasheet just says that the larger the burst size, the less overhead for the transfer
20:44:47saratogai guess you'd have to record how quickly the transfer occurs
20:45:23saratogaplayback, fm, recording all work fine with a large burst, and apparently battery life is improved a little, so i may just commit after i test a few more devices's a 32-bit output register
20:54:20 Join rela [0] (~x@pdpc/supporter/active/rela)
20:55:12jhMikeSI guess the overhead would be on the DMA controller and peripheral
20:56:20saratogaalso strange that apparently 16 is the largest working burst size, datasheet says it has a 128 entry fifo
20:56:27saratoga128 word
20:56:38 Quit kvieta (Quit: greedo shot first)
20:59:00saratoganeed to test on amsv1 as well
20:59:22 Quit rela (Ping timeout: 265 seconds)
21:02:34 Quit Napych (Quit: Napych)
21:02:37jhMikeSmaybe it's sending a new DMA request when it has 16 words free
21:03:49 Quit Jinx (Read error: Connection reset by peer)
21:08:18 Join Jinx [0] (Dojo@unaffiliated/jinx)
21:16:46jhMikeSsaratoga: I have a clipv1 btw
21:21:38jhMikeSsaratoga: does that mean the transfer has to be a multiple of 16 words because that won't always be true
21:26:16 Join rela [0] (~x@pdpc/supporter/active/rela)
21:27:30fs-bluebotBuild Server message: 3New build round started. Revision 8a078d7, 255 builds, 24 clients.
21:30:51 Quit rela (Ping timeout: 250 seconds)
21:31:03***Saving seen data "./dancer.seen"
21:34:29fs-bluebotBuild Server message: 3Build round completed after 421 seconds.
21:34:32fs-bluebotBuild Server message: 3Revision 8a078d7 result: All green
21:34:43jhMikeSsaratoga: it's 16 because the memory boundary is 16, duh :)
21:42:38 Join rela [0] (~x@pdpc/supporter/active/rela)
21:47:19 Quit rela (Ping timeout: 250 seconds)
21:50:27 Join Mihail [0] (4d43b278@gateway/web/freenode/ip.
21:54:21MihailjhMikeS: any way to set bigger dma burst size?
21:58:24jhMikeSChange the DMA driver to allow setting it differently for source and destination
21:58:53 Join rela [0] (~x@pdpc/supporter/active/rela)
21:59:51MihailI try this with no success
22:01:41jhMikeSit is possible to unabandon a change in gerrit?
22:02:46jhMikeSshit, I guess not
22:03:38 Quit rela (Ping timeout: 265 seconds)
22:05:55fs-bluebot3Gerrit review #1145 at : 3Fix a race condition in as3525 I2C driver caused by stacked ISRs. by Michael Sevakis
22:15:11 Join rela [0] (~x@pdpc/supporter/active/rela)
22:19:49 Quit rela (Ping timeout: 250 seconds)
22:34:54 Join [Saint] [0] (77e0243f@rockbox/staff/saint)
22:42:32gevaerts[Saint]: I'd say it's better to delete bot posts right away. That way they go back to 0 posts, which means they need approval again
22:43:31[Saint]gevaerts: Oh, yeah, sure. I just felt uneasy about a: deleting saratoga's post and/or b: leaving it there after I pruned the spam.
22:43:44[Saint]had no one replied to it it'd be gone.
22:44:19[Saint]In my mind it was better to let saratoga clean up than wonder where his post went.
22:44:40gevaertsYou can always tell him somewhere else, such as here
22:45:44gevaertsNot too important though
22:58:05Mihailsaratoga: Have a time to review g1146 and g1147?
22:58:13fs-bluebot3Gerrit review #1146 at : 3AS3543: Fix recording volume setting by Mihail Zenkov
22:58:13fs-bluebot3Gerrit review #1147 at : 3AMS: Fix pop-clicks noise on play/stop. by Mihail Zenkov
22:59:31saratogaMihail: volume setting looks fine (assuming you tested it and it works)
22:59:41saratoganeed to look at the pop-click fix a little more
22:59:49saratogathats the one caused by disabling the audio mixer right?
23:01:53saratogajhMikeS: which memory boundary?
23:02:16Mihailsaratoga: we discuss it early,50609.0.html
23:02:47jhMikeSsaratoga: the natural 16-byte alignment (cache line)
23:03:39saratogaMihail: so basically we trade some power for some tiny increase in SNR?
23:03:46saratoga(because the clock is on when paused)
23:04:32saratogajhMikeS: isn't that 16 words (so 64 bytes)?
23:06:38jhMikeSI'm honestly not confident because of the was it's described
23:07:12jhMikeS> You must set this value to the burst size of the destination peripheral, or if the
23:07:12jhMikeSdestination is memory, to the memory boundary size.
23:07:57jhMikeSBut before that it says "number of transfers"...soo....
23:08:24Mihailsaratoga: Yes, we have bigger power consumption on pause and no SNR difference. Just fix annoying clicks.
23:08:48 Quit saratoga (Ping timeout: 246 seconds)
23:08:50*jhMikeS has no clicks on his Fuzev2
23:08:53 Join saratoga_ [0] (123e11e0@gateway/web/freenode/ip.
23:09:42saratoga_Mihail: we could just revert the mixer change, in which case we'd save some power on pause but slightly increase distortion
23:09:45saratoga_which do you prefer?
23:10:06saratoga_IMO the power consumption and quality differences are both very small
23:11:52MihailWithout mixer changes we still have clicks but not so sharp. I prefer fix it and get best sound quality :)
23:13:44jhMikeSthe clicks happen about 3 seconds after pausing?
23:14:06jhMikeSah, ok
23:14:32jhMikeSIs the mixer change pushed because I don't hear anything.
23:15:00jhMikeSderp, I guess if it can be reverted
23:16:37MihailI describe easy way to reproduce it on forum.
23:16:38saratoga_sounds like it will use smaller bursts if not a multiple
23:17:11saratoga_jhMikeS: yes the mixer change is pushed, but you need sensitive headphones to hear it (i didn't notice it on mine)
23:17:50saratoga_i don't really care if we just revert the mixer change (very small effect on distortion) or keep the clock enabled (very small increase in power consumption when paused)
23:18:13saratoga_both effects are probably too small to notice without measurement equipment
23:18:59MihailIt have big difference in quality with external amp.
23:19:26*jhMikeS can hear it if he turns up the volume
23:21:10[Saint]I have some pretty bloody sensitive IEMs and a nice external amp, and, its definitely noticeable - but not something I'd write home about.
23:21:19 Quit krabador (Quit: Sto andando via)
23:21:37[Saint]And, honestly, if I'm in the top percentile of equipment users and its only barely noticeable, Joe User won't care at all.
23:22:16jhMikeSit seems the mixer change makes PCM not work if the radio is playing, which makes voice not work too
23:22:57[Saint]well, I guess that settles it, lol
23:23:02saratoga_which change?
23:23:03jhMikeSso there's that side effect
23:23:45jhMikeSsaratoga_: the mixer being off. it needs to mix analog and DAC while using the radio but it doesn't seem to work now.
23:23:50saratoga_oh i suppose because the mixer is need to combine the ADC input (radio) with the DAC output (voice)
23:24:07jhMikeSthat makes voice unusable with the radio
23:24:12saratoga_could probably make the driver force the mixer on if the radio is playing though
23:24:51jhMikeScould do that too; I just thought I'd point it out :)
23:25:28Mihailyes, we can fix this behavior
23:27:28jhMikeSif any input input is analog, mixer must be on
23:28:15Mihailyes and set it to 'sum' mode
23:28:50jhMikeSbasically disable only for DAC-only monitoring
23:30:54jhMikeSmonitoring and input source are supposed to remain independent
23:31:06***Saving seen data "./dancer.seen"
23:31:27saratoga_jhMikeS: when would you have a non-mulitple of 16 word DMA transfer over I2S?
23:32:20jhMikeSsaratoga_: because PCM isn't restricted to any block size. everthing is able to play any length of clip.
23:32:48saratoga_do they actually happen regularly? i'd like to test if possible
23:33:20jhMikeSplugins would be more likely to do it
23:33:55saratoga_Mihail: do you want to fix the mixer problem and the recording volume problem together and i'll do one push?
23:34:23jhMikeSthe only universal requirement is a 32-bit word-aligned buffer
23:35:33jhMikeSThe PCM mixer I think could just be set to an arbitrary buffer size to test it out in a controlled way
23:37:28jhMikeS /firmware/export/pcm_mixer.h, set MIX_FRAME_SAMPLES to whatever; I think that should go
23:37:32MihailWe have 8188 for DMA transfer size (MAX_TRANSFER) so non-mulitple transfer should occur regulary.
23:40:32jhMikeSnot for playback, that only does 1K at a time
23:41:29saratoga_i have to run, but it sounds like this is easy to test
23:41:29MihailWhy so small?
23:43:30jhMikeSmikmod looks like it does huge buffers and it uses PCM directly
23:43:44jhMikeSMihail: want low latency
23:44:47jhMikeSthe longer the latency, the longer it takes any new channel to start up
23:45:47Mihail23ms for 44100, right?
23:45:48jhMikeSalso, you spend more time in the ISR at once copying/mixing
23:47:27Mihailah, stereo 16bit
23:47:29jhMikeSthat interrupt load isn't so much, ~173 fps
23:50:17jhMikeShmm...mikmod doesn't align its char audio buffer
23:53:06jhMikeSPP5002 gets hit though because it can only do very short blocks because it's not using hardware DMA

Previous day | Next day