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 2020-11-01

00:15:22 Quit [7] (Disconnected by services)
00:15:28 Join TheSeven [0] (~quassel@rockbox/developer/TheSeven)
00:41:39***Saving seen data "./dancer.seen"
01:08:44 Quit mendel_munkis_ (Remote host closed the connection)
01:10:07 Join mendelmunkis [0] (
01:12:21 Quit massiveH (Quit: Leaving)
01:41:42***Saving seen data "./dancer.seen"
03:03:21 Join _bilgus_ [0] (~bilgus@
03:04:25 Quit _bilgus__ (Ping timeout: 264 seconds)
03:41:46***Saving seen data "./dancer.seen"
03:47:01 Join lebellium [0] (
04:17:34 Join johnb3 [0] (
04:39:38 Quit prof_wolfff (Ping timeout: 264 seconds)
05:36:08 Join pamaury [0] (~pamaury@rockbox/developer/pamaury)
05:41:50***Saving seen data "./dancer.seen"
05:56:57 Join MrZeus [0] (~MrZeus@
07:24:26fs-bluebot_Build Server message: New build round started. Revision f9ba96c, 293 builds, 8 clients.
07:39:00fs-bluebot_Build Server message: Build round completed after 874 seconds.
07:39:04fs-bluebot_Build Server message: Revision f9ba96c result: All green
07:41:53***No seen item changed, no save performed.
07:56:02 Join prof_wolfff [0] (
08:15:34 Quit pamaury (Ping timeout: 246 seconds)
08:17:38 Join livvy [0] (~livvy@gateway/tor-sasl/livvy)
09:28:44 Quit prof_wolfff (Ping timeout: 258 seconds)
09:35:03 Part edhelas
09:35:44 Join edhelas [0] (9d94237298@2a01:7c8:aab8:6b9:5054:ff:fec9:fd84)
09:41:55***Saving seen data "./dancer.seen"
10:20:18 Join pamaury [0] (~pamaury@rockbox/developer/pamaury)
10:57:12 Quit johnb3 (Ping timeout: 272 seconds)
11:19:38_bilgus_lebellium, do you have a physical OndaVx?
11:21:08_bilgus_well here is the probalem even at head I can not reproduce the issues I see in the Onda sim on the other physical player I have
11:21:20_bilgus_maybe I should try another sim
11:21:43_bilgus_this is with the failsafe theme so I doubt its a theme thing
11:24:40_bilgus_either that or maybe touch screen targets in general but the issue I see is that in failsafe the text is drawn and then cleared and never refreshed again (even with the dirty vp stuff disabled)
11:26:01_bilgus_I try to force redraw anywhere in the stack above it still doesnt redraw but if I go to the place where text is drawn and set it update=true it works fine but redraws too much then
11:26:35_bilgus_someone is removing my redraw but I can't seem to find where
11:27:18lebelliumwe don't have many touch screen sims
11:27:35lebelliumIf we had a YP-R1 sim, I wouldn't use the Onda VX
11:27:48lebelliumI'm wondering is anyone is still using this device
11:28:36_bilgus_I think I might just hack it so the Onda forces redraw
11:29:04_bilgus_the issue is reduced battery life but at least it redraws properly
11:41:26 Join prof_wolfff [0] (
11:41:57***Saving seen data "./dancer.seen"
11:43:40_bilgus_lebellium, do you know how to check out a branch prior to a commit its just checkout Commit-ID right?
11:46:05_bilgus_anyway I assume thats right anyway the issue was existing (at least in the sim)
11:46:39_bilgus_coding around it and I'll revisit if we get something concrete
11:46:40speachylebellium: if you could take hi-res photo of the YP-R1 I can hack a sim target into being around it.
11:48:08lebellium_bilgus_: I don't know much about git. I always need the wiki page
11:48:31_bilgus_I checked the source files I did it right
11:49:47_bilgus_its an existing issue
11:50:56lebelliumspeachy: thanks. I'll try with daylight
11:51:24 Quit Acou_Bass (Ping timeout: 240 seconds)
12:02:19 Join Acou_Bass [0] (
12:37:57 Join johnb3 [0] (
12:40:26 Quit Acou_Bass (Ping timeout: 272 seconds)
12:43:45fs-bluebot_Build Server message: New build round started. Revision 0c99a3f, 293 builds, 8 clients.
12:44:02 Join Acou_Bass [0] (
12:44:23_bilgus_lebellium, The other issues with the onda sim are fixed I in this patch
12:45:05_bilgus_I had the FS issue fixed sometime in the process of trying to fix this bug but I'll have to revisit that later
12:48:37 Nick mendelmunkis is now known as mendel_munkis (
12:50:55mendel_munkis_bilgus_: how is a function call faster than an if?
12:57:19fs-bluebot_Build Server message: Build round completed after 813 seconds.
12:57:23fs-bluebot_Build Server message: Revision 0c99a3f result: All green
13:07:46 Quit johnb3 (Quit: Nettalk6 -
13:08:01_bilgus_I don't know that I could say it is without benchmarks and a caveat of depends on arch and compiler but it gets rid of a branch
13:08:39_bilgus_the code no longer has to go two ways to exit it just calls the function and returns
13:10:11_bilgus_now if we start talking about multiple ifs then the cost is amortized and its just better and better we can do processing outside the hot path
13:12:43_bilgus_did you find the fn ptr code to be much slower than the original?
13:19:31mendel_munkisI haven't actually benchmarked it I was just thinking about the fact that a function call requires stack overhead.
13:22:53_bilgus_tbh I'd expect the compiler to inline the thing in the standard case
13:23:58mendel_munkisin which case it becomes an if again just with an added function call. (if I understand inlining correctly)
13:24:02_bilgus_once you add other conditionals though all bets are off
13:24:18_bilgus_no function call if inlined
13:24:46mendel_munkisin the non standard case. unless you're say it would inline both?
13:25:39_bilgus_don't know tbh depends on the compiler
13:26:07_bilgus_it might inline both but that depends on so many vars
13:26:37speachywe're using -Os (except the codecs) so the compiler will generally prefer strategies that produces the smallest code.
13:27:26speachyI think GCC describes it as "enable everything from -O2 except where it results in a larger binary size"
13:28:07_bilgus_we can worry about this but really its already at 75 fps pretty good
13:31:22_bilgus_now the 30% slow drawing routines OTOH
13:31:37 Quit emacsomancer (Read error: Connection reset by peer)
13:32:07_bilgus_I'm gonna try for 10% hopefully the end up 15% slower
13:33:55mendel_munkisI got identical speeds with both builds. (according to test_fps) but I prefer patchset 2 if there is no optimization difference
13:34:38mendel_munkisalso what kind of glitches were you seeing and where they only with screen_flip enabled?
13:36:38_bilgus_yes only when flipped
13:37:09_bilgus_it was like tearing with a white flash
13:37:49_bilgus_its your patch I prefer 3 but tbh IDC enough to argue with you or i'd do it myself :P
13:38:29_bilgus_but i'll be back later I have to go put a starter on my car I'm tired of driving the backup it kills me in gas
13:38:44speachy_bilgus_: good luck.
13:38:53*speachy still reeks of gasoline
13:39:22_bilgus_that was last week when I fixed the backup because I anticipated this starter issue
13:39:42_bilgus_it needed a new filler neck
13:39:43 Quit livvy (Ping timeout: 240 seconds)
13:41:58***Saving seen data "./dancer.seen"
13:43:01 Join emacsomancer [0] (
13:50:57mendel_munkisnothing. if you're using cabbie I can't think of anything else that would cause it.
14:36:20 Quit pamaury (Ping timeout: 272 seconds)
15:18:03 Join pamaury [0] (~pamaury@rockbox/developer/pamaury)
15:42:01***Saving seen data "./dancer.seen"
16:08:25 Quit Acou_Bass (Ping timeout: 240 seconds)
16:10:35 Join Acou_Bass [0] (
16:15:25 Quit Acou_Bass (Ping timeout: 240 seconds)
16:39:35 Join Acou_Bass [0] (
17:12:22 Quit lebellium (Quit: Leaving)
17:42:03***Saving seen data "./dancer.seen"
18:00:04 Quit mendel_munkis (Remote host closed the connection)
18:00:18 Join mendel_munkis [0] (
18:18:15 Quit Acou_Bass (Ping timeout: 260 seconds)
18:21:42 Join Acou_Bass [0] (
18:44:05 Quit Acou_Bass (Ping timeout: 240 seconds)
18:47:53 Join Acou_Bass [0] (
19:09:12 Quit pamaury (Quit: this->disconnect())
19:36:31 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
19:37:50 Join fs-bluebot [0] (
19:39:45 Quit fs-bluebot_ (Ping timeout: 240 seconds)
19:39:49 Quit bluebrother^ (Ping timeout: 264 seconds)
19:42:05***Saving seen data "./dancer.seen"
20:06:18 Join mendel_munkis_ [0] (
20:08:45 Quit mendel_munkis (Ping timeout: 240 seconds)
20:38:49 Quit MrZeus (Ping timeout: 256 seconds)
20:55:31 Quit speachy (Ping timeout: 260 seconds)
21:03:09_bilgus_mendel testing I grabbed the patchset from gerrit and it is bouncing around when flipped just scrolled the menu to the top and it bounces I'll try again after rebase
21:14:41_bilgus_mendel_munkis_, OK same deal at head after flipping, in main menu hold the up key and watch after it hits the stop it keeps bouncing around
21:15:45_bilgus_test_viewports plugin the screen tears in lines 1 - 11 pixel wide horizontal bands that glitch across the screen with equal spacing
21:15:57_bilgus_the scrolling text doesn't work either
21:17:11_bilgus_right side up main menu is stable test viewports works fine and scrolling text is working
21:17:23 Nick mendel_munkis_ is now known as mendel_munkis (
21:17:31mendel_munkisgive me a second to rebase on head.
21:17:39 Join speachy [0] (~speachy@
21:17:54_bilgus_its the same there just test_viewports has a stack ovfl :)
21:19:25braewoods_bilgus_: i think i'm going to rewrite iriver_flash and submit it when I'm done fully testing it... i'll need to retain compatibility with some old stuff but beyond that...
21:20:01_bilgus_braewoods, patches welcome :)
21:20:20braewoodsi feel like i keep patching old assumptions and it feels like i need to start from scratch and just use the old one as reference
21:20:33braewoodsjust to wrap my head around the whole logic
21:23:29mendel_munkisA. I can't reproduce on head B. that sounds identical to patchless behavior.
21:27:29_bilgus_ok lets rebase the commit on gerrit ands I'll try pulling it again
21:28:14_bilgus_well actually its not too deep I'll just verify its there
21:29:25_bilgus_yes 3 is there I'll try copy pasta 2
21:31:52mendel_munkisdoes if else order matter for optimization?
21:32:36_bilgus_sometimes but I use LIKELY or UNLIKELY to the same effect
21:33:14_bilgus_just don't use it wrong
21:34:28_bilgus_gonna be a minute I nuked the build dir to be sure
21:36:52_bilgus_mendel as a general rule of thumb unless you know its a bottleneck no real point in bothering
21:37:36mendel_munkisWell you were concerned about the inefficiency of using an if.
21:38:40_bilgus_and .5 fps is not much
21:40:15_bilgus_also I'm not sure what the most likely path is amongst fuze+ users
21:40:34 Quit Acou_Bass (Ping timeout: 256 seconds)
21:40:41_bilgus_the player is begging to be upside down its even weighted towards that end
21:42:07***Saving seen data "./dancer.seen"
21:42:12mendel_munkisI find that using it upside down tends to result in my thumb covering half the screen.
21:46:18_bilgus_same deal with patchset 2
21:46:27 Join Acou_Bass [0] (
21:47:38mendel_munkiswhats the version number?
21:48:42_bilgus_send me your version
21:48:51mendel_munkisthe zip?
21:49:18_bilgus_lets remove this first
21:49:35_bilgus_then we can start questioning device differences
21:50:51mendel_munkiscan you send me your cfg while we are at it?
21:51:08mendel_munkisit's uploading now
21:51:18_bilgus_in debug we have hw info>target it says lcd kind ili9325
21:51:43_bilgus_my config for rb? its a clean install too
21:52:00_bilgus_I presume it might have the lcd flip maybe?
21:52:43mendel_munkisok it's not a config difference.
21:53:04mendel_munkisbut seeing as I have a st7781 that's probably the cause
21:53:12_bilgus_I wanted to remove all vars I nuked everything
21:53:52mendel_munkisso I need to update to check lcd type.
21:53:54_bilgus_ah there we go ok so mine either needs special handling or is a bad implementation
21:54:10mendel_munkisnope mine is the bad implementation that needs special handling.
21:54:26mendel_munkisthe patch was to introduce the special handling.
21:54:28_bilgus_pretty sure mine glitched at head too
21:54:44_bilgus_let me double check that
21:55:22mendel_munkisdon't tell me they have different bad handling
21:56:37_bilgus_ah also it would not be a compile time constant either :(
21:58:17 Join Stanley00 [0] (~stanley00@unaffiliated/stanley00)
21:59:24mendel_munkisThe data sheet has the same quote about locations except it uses GRAM not DRAM
21:59:38_bilgus_ok so your patch improved the part in the main menu BUT
22:00:10_bilgus_the scrolling lines work in test viewport no glitches they are just misplaced
22:00:33_bilgus_so you know which part did work
22:00:44_bilgus_and the register part did not
22:01:02mendel_munkisI have three Fuze+s and they are all st7783 so I have no way to test.
22:01:55mendel_munkisalso I don't see why moving a chunk of video memory should cause scrolling glitches but I will see if I can figure it out.
22:02:09_bilgus_ok so the viewport part leqave it for everyone and the register changes -ifdef- nope not ifdef because its not a compile time constant
22:02:30mendel_munkisthe registers are the same according to the datasheets.
22:02:47_bilgus_hopefully you can do that stuff in the menu and not in the draw routine
22:03:16_bilgus_I'm saying my device works better without the register changes you made
22:03:49mendel_munkisso the datasheet is wrong.
22:04:53mendel_munkisalso what viewport part?
22:05:07_bilgus_OH I thought you also sent another set of commands to the lcd for flip but it looks like you just added height-y
22:05:14mendel_munkisI didn't touch the viewport
22:05:22_bilgus_that is the vp
22:05:32_bilgus_its just the default one
22:05:50mendel_munkisah. i was thinking of it as the registers containing the rectangle to be updated.
22:06:29mendel_munkisbecause technically it doesn't have to be a viewport.
22:06:48_bilgus_no but it helps to think of it when it comes to transforms
22:07:05_bilgus_or windows if you prefer
22:07:22mendel_munkisfair enough. I just thought that the rockbox viewport os something else and didn't want to confuse them.
22:07:48_bilgus_its all bitmaps
22:09:06mendel_munkisso register 21 is what was causing problems?
22:17:32_bilgus_? no idea
22:18:36_bilgus_changing it back to y but leaving the other two makes the scrolling lines show right but its still glitchy
22:19:02_bilgus_and it is off like 6 lines everywhere else
22:19:19_bilgus_its odd and makes me want to do it in software
22:19:27mendel_munkisthat makes sense. It's why I set it in the first place
22:20:15_bilgus_it just seems like too much effort for one target when they can just all get it
22:20:37mendel_munkisbutdoing it in software is so much slower
22:21:20_bilgus_I doubt reverse memset is much slower than forward memset
22:21:42braewoods_bilgus_: meaning where they start from is different?
22:23:40_bilgus_yes one goes 0- 1228000 and the other goes 1228000 - 0
22:25:17mendel_munkis_bilgus_ what happens if you add 128 to reg_3_val?
22:25:25mendel_munkisthat's ili specific
22:26:42_bilgus_where should I add it? its checking &0x20 do I need to add it there too?
22:28:06mendel_munkisno in lcd_set_flip instead of 0x1000 use 0x1080
22:29:33mendel_munkisalthough I need to think for a few minutes if that would be with the old or new value of register 21.
22:30:35_bilgus_its almost perfect and I put it back before I changed it
22:30:42 Quit Acou_Bass (Ping timeout: 272 seconds)
22:30:53_bilgus_so h-y for reg 21
22:31:30mendel_munkistry it with just y.
22:31:33_bilgus_it has just the slightest flutter up and down in the menus and no glitch in the plugin
22:32:29 Quit Stanley00 (Read error: Connection reset by peer)
22:32:38 Join Acou_Bass [0] (
22:33:13 Join Stanley00 [0] (~stanley00@unaffiliated/stanley00)
22:34:37_bilgus_menu worse
22:34:46_bilgus_vp_test no change
22:35:13_bilgus_I'd call it good enough to say you've got it
22:36:04mendel_munkisok now I have to make sure the that bit flag won't blow up the st7783
22:36:32_bilgus_oh itd be nice if they go together I figured you'd have to split on them
22:37:07mendel_munkisthe data sheet just has it zerod out. either it needs to be zero or it doesn't have much effect.
22:38:11_bilgus_so in your math LCD_HEIGHT-(y+h-1));
22:38:33_bilgus_that paren is superfluous
22:39:46mendel_munkisok. I never remember the order of operations so I usually am very liberal with parenthesis on the basis that it cant hurt.
22:39:57_bilgus_instead do int y_f = height - y and use it in place so the compiler can optimize
22:40:32_bilgus_right now unless its smart those are at best 2 and at worst 3 calcs
22:40:38mendel_munkisy_f for y floor?
22:40:50_bilgus_y_f as in flipped lol
22:41:57_bilgus_or yr for y reflected?
22:57:13mendel_munkisI made a few minor changes to make it cleaner and fix a small artifact. (and managed to screw up committing them) can you test patchset 9?
22:59:40_bilgus_is it the latest?
23:03:11_bilgus_run oscilloscope in flipped mode
23:03:48_bilgus_get a wall of red on first run through?
23:05:42_bilgus_so off by 1 maybe?
23:06:13mendel_munkishow would an off by one cause that?
23:06:33_bilgus_grabbing the wrong line?
23:07:27mendel_munkisso I fixedd one off by one and caused another.
23:07:45_bilgus_off by 1/2?
23:08:39_bilgus_just thinking of how oscope works that line is stationary its the copying code that moves it
23:09:04_bilgus_but it could be something weird about the lcd too
23:09:37_bilgus_hell if I know you are the one who figured it out :)
23:14:13 Quit TheSeven (Disconnected by services)
23:14:23 Join [7] [0] (~quassel@rockbox/developer/TheSeven)
23:15:33mendel_munkisI can;t figure out where it gets any red in the first place.
23:17:01mendel_munkisalso the rectangles it's updating seem weird.
23:18:32 Quit _bilgus_ (Remote host closed the connection)
23:22:14 Join _bilgus [0] (~bilgus@
23:23:08_bilgusmendel rest assured it is weird
23:23:38_bilgusbut it follows the implementation to the T it is a very good test
23:24:00_bilgusany little thing off breaks it in horrible ways
23:24:13mendel_munkisok. oscilloscope was always weird to me. when I first realized you can set it wrap it changed everything.
23:24:39_bilgusI was neck deep in there for a week
23:42:09***Saving seen data "./dancer.seen"
23:43:37 Quit Stanley00 (Read error: Connection reset by peer)
23:44:06 Join Stanley00 [0] (~stanley00@unaffiliated/stanley00)
23:46:22mendel_munkisthe red bar is also at head
23:48:18_bilgushmm then carry on
23:48:56_bilgusnot making it worse at least
23:49:35_bilgusthere will be more bugs appearing with the framebuffer stuff too
23:50:52mendel_munkisI haven't noticed anything blowing up because of the register, and I cant think offhand of how these registers cause it (although I did noarrow it down to line 980 not saving the old state)
23:51:54mendel_munkisI think if the only issue you noticed was the oscilloscope it's committable.
23:52:08mendel_munkisdi the last patchset help ith the menu flutter?
23:55:56_bilgusidk maybe the magnitude is less
23:58:30mendel_munkisI think it's probably an off by one. but that's one I cant test.

Previous day | Next day