| T42 | <FakeShell> i was working on some vulkan fixes on hybris and got some PRs up | 02:02 |
|---|---|---|
| T42 | <FakeShell> https://github.com/libhybris/libhybris/pull/604 | 02:02 |
| T42 | <FakeShell> https://github.com/libhybris/libhybris/pull/610 | 02:02 |
| T42 | <FakeShell> https://github.com/libhybris/libhybris/pull/611 | 02:02 |
| T42 | <FakeShell> https://github.com/libhybris/libhybris/pull/612 | 02:02 |
| T42 | <FakeShell> https://github.com/libhybris/libhybris/pull/613 | 02:02 |
| T42 | <FakeShell> | 02:02 |
| T42 | <FakeShell> hopefully someone with enough interest can give it a review and maybe pull it in | 02:02 |
| T42 | <FakeShell> | 02:02 |
| T42 | <FakeShell> on a different note, is there any icd implementation for vulkan (even experimental)? i'm having issues getting vulkan loader to deal with function pointer output of icdGetInstanceProcAddr | 02:02 |
| T42 | <fredldotme> I started work on ICD-ifying libhybris Vulkan, but hit a roadblock wrt dispatch table buildup. The Android-side dispatch table would, during creation, call into a magic number and cause a SIGBUS. (re @FakeShell: i was working on som...) | 07:07 |
| T42 | <fredldotme> If there is still interest I can link you to my experiments. | 07:07 |
| T42 | <FakeShell> similar issue here, sigbus and I can't pinpoint where it happens (re @fredldotme: I started work on IC...) | 07:08 |
| T42 | <fredldotme> You might have some success loading the Android debug symbols into gdb and find it, I assume it's the same dispatch table buildup that fails. (re @FakeShell: similar issue here, ...) | 07:08 |
| T42 | <fredldotme> Just to be sure, would you find the address that causes the SIGBUS in vulkan-loader sources? | 07:10 |
| T42 | <FakeShell> I had better luck using libGLES_mali from vendor since that bypasses androids wrapper for vulkan but obviously that's not a solution usable for all targets | 07:11 |
| T42 | <FakeShell> on my device libGLES exposes raw syms dlopenable with hybris, much like libvulkan itself | 07:11 |
| T42 | <FakeShell> I have the call stack in gdb from experiments today so I know where it fails at least in vulkan loader (re @fredldotme: Just to be sure, wou...) | 07:13 |
| T42 | <FakeShell> getting anything past vulkan loader itself is the part I couldn't figure out, I'm not very good at using gdb with things in android world | 07:14 |
| T42 | <fredldotme> Need assistance? I at least know that the upstream libhybris symbol loader python script needs some adjustments so that it can match the library addresses. | 07:16 |
| T42 | <FakeShell> sure I wouldn't mind | 07:16 |
| T42 | <FakeShell> for now assuming your changes for vulkan will be pulled on, I also pulled it into my beanch | 07:17 |
| T42 | <FakeShell> for now assuming your changes for vulkan will be pulled on, I also pulled it into my branch (edited) | 07:17 |
| T42 | <fredldotme> Source this from gdb : https://irc.thaodan.de/.imgstore/7a015c71/load_sym_files.py | 07:17 |
| T42 | <FakeShell> for now assuming your changes for vulkan will be pulled in, I also pulled it into my branch (edited) | 07:17 |
| T42 | <fredldotme> also, make a symlink from `android` to `.` next to `system` and friends. | 07:19 |
| T42 | <fredldotme> in your debug symbols folder | 07:19 |
| T42 | <FakeShell> I can try tomorrow | 07:20 |
| T42 | <fredldotme> feel free to ping me, I just re-learned using this a few weeks back | 07:20 |
| T42 | <FakeShell> https://github.com/FuriLabs/libhybris/commits/feature/forky/vulkan | 07:21 |
| T42 | <FakeShell> in case you want to follow along my progress | 07:21 |
| T42 | <FakeShell> I was fixing up crashes and things with vulkan up until today, I have gtk apps and all vulkan benchmarks running without crashes | 07:21 |
| T42 | <FakeShell> but obviously no ICD == can't redistribute properly without diversion hacks so all that ain't of any use until I get ICD functional | 07:22 |
Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!