| - wagga (QUIT: Quit: Client closed) (~wagga@node-1w7jra22ildhvg0963fkd9csj.ipv6.telus.net) | 01:31 | |
| + jcs (~jcs@jcs.org) | 01:59 | |
| chartreuse | Noticed a neat thing with the clear trackball buttons, they flouresce slightly under sunlight, gives them a slight purplish hue | 02:23 |
|---|---|---|
| mntmn | yeah they respond to UV more than the ones we printed before | 02:42 |
| doppler | huh, neat | 03:14 |
| + reform10226 (~jamin@2603-8000-b400-1b02-06f0-21ff-fe91-0625.res6.spectrum.com) | 03:42 | |
| * reform10226 -> acdimalev | 03:42 | |
| - acdimalev (QUIT: Client Quit) (~jamin@2603-8000-b400-1b02-06f0-21ff-fe91-0625.res6.spectrum.com) | 03:43 | |
| + reform26164 (~jamin@2603-8000-b400-1b02-06f0-21ff-fe91-0625.res6.spectrum.com) | 03:45 | |
| * reform26164 -> acdimalev | 03:46 | |
| chartreuse | Think I'm going to do my first circuit board level mod. Replacing C127 and C147 47uF ceramics with 100-330uF audio electrolytics (gotta check what I have) to improve the headphone output bass response greatly | 06:14 |
| chartreuse | With my 24 ohm over-ears 47uF has a cutoff of 150Hz, and with my 16ohm ear buds it's worse at 220Hz | 06:16 |
| + chartreu1e (~chartreus@S0106f0f249dfd9c3.cg.shawcable.net) | 07:45 | |
| - chartreuse (QUIT: Ping timeout: 272 seconds) (~chartreus@S0106f0f249dfd9c3.cg.shawcable.net) | 07:49 | |
| - chartreu1e (QUIT: Quit: leaving) (~chartreus@S0106f0f249dfd9c3.cg.shawcable.net) | 07:54 | |
| + chartreuse (~chartreus@S0106f0f249dfd9c3.cg.shawcable.net) | 08:35 | |
| chartreuse | Okay well the soldering was a bit sketchy trying to solder some rather large 100uF 16v audio caps I had onto 0603 pads, but it's a massive improvement in sound and bass | 08:36 |
| chartreuse | I've got some better/smaller 220uF 6.3v ones on the way on digikey, didn't think the 0603 pads were big enough for a surface mount electrolytic so got through hole ones again | 08:37 |
| chartreuse | Just going to mount them in much the same way though they should be slightly smaller as well so might fit neater | 08:37 |
| chartreuse | Not like the motherboard can be changed now, but replacing those ceramics with holes for small diameter through hole caps would likely fit in the area. Or possibly some surface mount electrolytics would fit with a little shuffling | 09:08 |
| ex-parrot | Nice | 09:16 |
| ex-parrot | I should do a similar mod | 09:17 |
| chartreuse | Makes a nice improvement even just going up to 100uF. Do be careful if you do do it, it's a small area and the pads will be relatively fragile | 09:28 |
| chartreuse | I don't like MLCC for audio, but another idea could be to stack another 47uF in parallel with the originals | 09:29 |
| chartreuse | Posted a picture of the jank on the community forum | 09:30 |
| - acdimalev (QUIT: Quit: .) (~jamin@2603-8000-b400-1b02-06f0-21ff-fe91-0625.res6.spectrum.com) | 10:43 | |
| chartreuse | Also forcing Firefox to use native wayland makes a large difference if you have a video playing, even when the video is off screen. I tried it a few days ago and I thought I had issues with firefox starting, but trying it again it works just fine. | 10:46 |
| chartreuse | Does give different graphical glitches at the moment, sometimes the window flickers transparant when loading a new page | 10:59 |
| - wiedi (QUIT: Read error: Connection reset by peer) (~wiedi@37.228.191.159) | 11:51 | |
| + wagga (~wagga@node-1w7jra22ildhybaszosec4fjh.ipv6.telus.net) | 12:39 | |
| - erlehmann (QUIT: Read error: Connection reset by peer) (~erle@dynamic-046-114-036-024.46.114.pool.telefonica.de) | 13:24 | |
| + erlehmann (~erle@dynamic-046-114-032-213.46.114.pool.telefonica.de) | 13:37 | |
| mntmn | chartreuse: i saw your mod, cool cool | 13:41 |
| mntmn | chartreuse: and yes, i got the same flickering @ firefox with forced layer acceleration | 13:41 |
| - wagga (QUIT: Quit: Client closed) (~wagga@node-1w7jra22ildhybaszosec4fjh.ipv6.telus.net) | 13:42 | |
| mntmn | chartreuse: in the forum i wrote about this once > My unconfirmed intuition is that the glitches are some CPU/GPU buffer sharing problem. I discovered that setting dom.ipc.processCount to 1 instead of the default 8 greatly reduces the problem | 13:42 |
| mntmn | chartreuse: (if you have layers.acceleration.force-enabled = true) | 13:42 |
| chartreuse | I've already got the ipc count down to two. I had a slightly different glitchiness (rarely) before forcing wayland, though after this happens almost every time I change pages | 13:50 |
| chartreuse | I do have the layers acceleration force enabled at the moment. | 13:51 |
| chartreuse | The main difference was forcing wayland with MOZ_ENABLE_WAYLAND=1, which made certain things way faster but added quite a bit of flickering in places | 13:51 |
| chartreuse | Going to try looking into the errors given on startup: Crash Annotation GraphicsCriticalError: |[0][GFX1-]: glxtest: DRM device has no render node (t=0.891441) [GFX1-]: glxtest: DRM device has no render node | 13:53 |
| chartreuse | And a couple others | 13:53 |
| mntmn | chartreuse: i think those errors are irrelevant | 13:55 |
| mntmn | chartreuse: the bugs you are seeing are probably mesa/etnaviv bugs related to buffer sharing | 13:55 |
| chartreuse | Okay turning off the layers force acceleration got rid of the flickering, and is still running faster than when I had it on but on xwayland (the default) | 13:56 |
| chartreuse | Alright | 13:56 |
| mntmn | chartreuse: i recorded the session with apitrace (incl glitches), but when playing it back, the glitches are gone | 13:56 |
| chartreuse | Hmm alright | 13:56 |
| mntmn | so it's complicated | 13:56 |
| chartreuse | I've seen reports of old sway issues that sound similar, though they've long been fixed | 13:57 |
| mntmn | there is an interesting mesa PR that we should test soon https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9780 | 13:57 |
| mntmn | chartreuse: yeah that's true, these kind of issues can also happen on other GPUs | 13:58 |
| mntmn | and it's often different stuff/causes that look similar | 13:58 |
| mntmn | wlroots/sway have also made bigger changes in the meantime, but we're on a bit older/more stable version. updating always needs good testing | 13:58 |
| chartreuse | I haven't looked much into it besides trying to enable it, but Firefox also has the WebRender as well, rather than trying to make the OpenGL renderer (layers.accelleration) work | 14:01 |
| chartreuse | Force enabling webrenderer results in dozens of similar errors to the drm device ones as well as failing to create GL contexts | 14:04 |
| mntmn | yeah, unfortunately webrender needs some futuristic OpenGL version that we don't have ;) | 14:05 |
| mntmn | 3.2 or something | 14:05 |
| mntmn | you can try setting MESA_GL_VERSION_OVERRIDE and MESA_GLSL_VERSION_OVERRIDE but idk which features are missing that firefox needs | 14:06 |
| mntmn | for example MESA_GL_VERSION_OVERRIDE=3.3 and MESA_GLSL_VERSION_OVERRIDE=330 | 14:06 |
| chartreuse | Ah of course that makes sense | 14:07 |
| mntmn | we have the same problem with blender 2.80+ | 14:07 |
| mntmn | that's why i ship a special build of 2.79b | 14:07 |
| chartreuse | It's wanting GLSL 1.5 so failing on that error: GLSL 1.50 is not supported. Supported versions are: 1.10, 1.20, and 1.00 ES | 14:07 |
| mntmn | you can try MESA_GLSL_VERSION_OVERRIDE=150 | 14:08 |
| - erlehmann (QUIT: Quit: Just say no, then the virus can not enter your body without your consent.) (~erle@dynamic-046-114-032-213.46.114.pool.telefonica.de) | 14:08 | |
| + erlehmann (~erle@dynamic-046-114-032-213.46.114.pool.telefonica.de) | 14:08 | |
| mntmn | the compiler frontend can handle any version but the question is if all required ops are supported by etnaviv | 14:08 |
| chartreuse | And now it just gives up: | 14:08 |
| chartreuse | Unhandled NIR tex src type: 3 | 14:08 |
| chartreuse | Unhandled NIR tex type: 4 | 14:08 |
| chartreuse | and segfaults | 14:08 |
| mntmn | ah very interesting though | 14:09 |
| mntmn | so this tells us that they use unsupported texture types | 14:09 |
| chartreuse | It repeats that 11 times and that's the only error (besides the same 3 DRM ones it always has) | 14:09 |
| mntmn | the error comes directly from etnaviv: | 14:10 |
| mntmn | > gallium/drivers/etnaviv/etnaviv_compiler_nir.c: compile_error(c, "Unhandled NIR tex src type: %d\n", | 14:10 |
| mntmn | the unsupported tex src type is "nir_tex_src_offset" | 14:12 |
| mntmn | > An integer value that is added to the texel address before sampling | 14:12 |
| chartreuse | I would have thought mesa would be what's handling this rather than the kernel gpu driver. Or is that path in mesa? | 14:15 |
| mntmn | that is a path in mesa | 14:16 |
| mntmn | the "gallium" drivers handle the command buffer construction in mesa | 14:16 |
| mntmn | so that is the userspace part of the driver | 14:16 |
| mntmn | that gets sent to the kernel, the kernel driver does more low level plumbing and buffer/memory handling of the gpu | 14:16 |
| chartreuse | Mesa 21.0.0's patch notes make mention of nir_tex_src_offset "zink: use ConstOffset for nir_tex_src_offset" | 14:17 |
| mntmn | zink is an opengl implementation on top of vulkan | 14:18 |
| mntmn | so if you have a working vulkan driver you can "emulate" opengl on it with zink | 14:18 |
| mntmn | we do not have vulkan though | 14:18 |
| chartreuse | Debian's mesa is 20.3.5 | 14:18 |
| chartreuse | Ah okay | 14:18 |
| mntmn | the "Unhandled NIR tex type: 4" is nir_texop_txf, /**< Texel fetch with explicit LOD */ | 14:18 |
| mntmn | so etnaviv doesn't know what the TXF instruction is | 14:19 |
| chartreuse | iMX 8M fact-sheet lists vulcan support for the gpu, is it a matter of there not being an open or linux driver for it? | 14:20 |
| mntmn | correct | 14:20 |
| mntmn | the etnaviv driver is written by reverse engineering the vivante blob driver | 14:21 |
| mntmn | chartreuse: i wrote about this once https://www.crowdsupply.com/mnt/reform/updates/graphics-improvements-4k-hdmi-and-proper-kicad-support#2-early-z-bug | 14:21 |
| mntmn | like, how reversing a single feature could be done | 14:22 |
| mntmn | (this earlyz and alpha test stuff is all fixed in the meantime) | 14:22 |
| mntmn | so, one writes a GL test case that is supposed to trigger the feature you want to see, and run that on the vivante blob driver, and then dump the command buffers and examine them if you find anything different | 14:23 |
| mntmn | so for example to support the TXF instruction (if it exists on the vivante gpu) one would make a minimal openGL program containing this instruction and look at what the blob does | 14:24 |
| + erle (~erle@dynamic-046-114-038-151.46.114.pool.telefonica.de) | 14:25 | |
| - erlehmann (QUIT: Killed (NickServ (GHOST command used by erle!~erle@dynamic-046-114-038-151.46.114.pool.telefonica.de))) (~erle@dynamic-046-114-032-213.46.114.pool.telefonica.de) | 14:25 | |
| mntmn | chartreuse: these two errors are related, first it sets up the coordinate it wants to read from the texture and then it wants to read a pixel/texel from the texture. | 14:26 |
| * erle -> erlehmann | 14:26 | |
| chartreuse | Alright I'll look into that. Going to take some getting up to speed on my part on GL programming and driver development | 14:26 |
| chartreuse | Hopefully not going to end up fixing that and it just then presents some unfixable error after | 14:27 |
| mntmn | related https://github.com/servo/webrender/blob/5aa779686d3b56c1b74ecb5df171a73c235c691f/webrender/res/base.glsl#L18 | 14:28 |
| mntmn | well adding new ops etc to etnaviv is always extremely desirable, because support for new GL versions is mostly a collection of supporting new ops... | 14:28 |
| mntmn | see also https://mesamatrix.net/ | 14:29 |
| chartreuse | Okay so a lot missing compared with other drivers mainstreamed | 14:34 |
| mntmn | yeah but this doesn't show OpenGL 2.1 or GLES 2 which are fully supported | 14:36 |
| - chartreuse (QUIT: Quit: leaving) (~chartreus@S0106f0f249dfd9c3.cg.shawcable.net) | 15:15 | |
| + wiedi (~wiedi@37.228.191.159) | 18:05 | |
| - S0rin (QUIT: Ping timeout: 268 seconds) (~S0rin@user/s0rin) | 22:10 | |
| + S0rin (~S0rin@user/s0rin) | 22:13 | |
| + chartreuse (~chartreus@S0106f0f249dfd9c3.cg.shawcable.net) | 23:24 | |
| + ephase (~ephase@2a01:e0a:168:1211::885) | 23:29 | |
| ephase | Hi | 23:36 |
| chartreuse | Hello | 23:40 |
| ephase | how are you chartreuse? | 23:44 |
| chartreuse | Doing alright still slowly getting used to the reform. You? | 23:47 |
| ephase | The same, I'm finishing wrinting an article about it | 23:59 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!