**** BEGIN LOGGING AT Fri Jan 01 03:00:13 2021 Jan 01 03:00:40 my next actions will involve attempting to compile and load my beaglebone's installed wifi adapter's niche driver (kernel module). On debian 10 on AM335x this takes ~25 min to compile, and on debian 9 it might be a challenge if there are complications due to the older operating system, though one always hopes not Jan 01 03:02:22 if the SGX530 access on that image seems promising but i can't connect my bb with wifi quickly, i will try to change my setup to allow ethernet access instead if i think it will let me get back into ssh and continue working faster. Jan 01 03:03:55 Okay. Good luck. I am trying to alter a .doc file into .pdf on Linux so I can view this instruction of the OMAP info. that is listed in .doc format. Jan 01 03:06:31 there's an open source software for linux capable of the task you described, the debian package is libreoffice-writer. however requires X server. i haven't learned of any command line only tool capable of that. Jan 01 03:13:56 No issue. Jan 01 03:51:03 You played me like a plaid shirt. Jan 01 03:51:21 It is okay. All is forgiven in the land of the beagle. Jan 01 03:55:33 misternumberone: I don't see any reason to use an obsolete debian release... if you want to try the 1.14 drivers it would suffice to downgrade the kernel to one that uses them Jan 01 03:56:04 like I said earlier, it's a bit annoying rcn's sgx modules packages don't list the version Jan 01 05:54:20 the image here https://debian.beagleboard.org/images/bone-debian-9.12-imgtec-armhf-2020-04-06-4gb.img.xz contains a full driver stack that is capable of accessing the SGX530 and precompiled demos that were sufficient to convince me they were executing OpenGLES 2.0 instructions on the device. Jan 01 05:56:16 the installed files were extremely similar in configuration to essentially everything I tried on Debian 10 with one major difference, Jan 01 05:58:11 which was that all of the user mode SDK files I had been attempting to use were replaced with a single package from the rcn-ee.com repository: ti-sgx-ti33x-ddk-um Jan 01 05:58:17 which i was unaware existed Jan 01 06:12:56 this package contains a /etc/init.d/ti-sgx-ti33x-ddk-um that replaces the functionality of the /etc/init.d/rc.pvr i was trying to use before Jan 01 06:14:26 it also contains libGLESv2.so.1.14.3699939, libEGL.so.1.14.3699939 and other required libraries for the graphics stack that i was attempting to use before Jan 01 06:15:39 the only problem: yes, indeed, this package is marked as only for Debian 9, NOT Debian 10 Jan 01 06:18:57 zmatt: the reason I see to use an obsolete debian release is that the official image stating it contains SGX530 support is Debian 9, the official Beaglebone repository for Debian only contains the user space portion of the driver for Debian 9, and all attempts to use any versions of the driver I've possibly been able to find on Debian 10 have failed. Is there any obvious, functional way I could actually use this in Debian 10 Jan 01 06:18:57 instead? Jan 01 06:21:27 what do you mean by "only contains the user space portion of the driver for Debian 9" ? it doesn't seem very likely that it depends on a specific distro release, in fact the userspace libs don't target debian at all, but they don't really have any dependency on a specific linux distro let alone a specific release thereof Jan 01 06:22:46 isn't the reason I couldn't view this package in Debian 10 because its filename contains "stretch"? Jan 01 06:22:56 ??? Jan 01 06:22:59 what package? Jan 01 06:23:09 ti-sgx-ti33x-ddk-um Jan 01 06:23:31 rcn packaged it for stretch but not for buster? weird Jan 01 06:23:42 when I downloaded it manually through web, I see the file "ti-sgx-ti33x-ddk-um_1.14.3699939-git20171201.0-0rcnee9~stretch+20190328_armhf.deb" only Jan 01 06:24:19 it probably installs just fine, but you'll have the same problem you had with my debian package Jan 01 06:24:35 which is incompatibility with the 1.17 kernel driver Jan 01 06:25:02 (at least, I think that's what the problem was) Jan 01 06:26:24 try downgrading to the latest 4.14-ti kernel Jan 01 06:27:21 I found his build scripts and it looks like he switched to 1.17 for 4.19 while 4.14 used 1.14 Jan 01 06:27:50 when I find a bit of energy I'll try it myself Jan 01 06:28:50 if that were really all there was to it, then surely I'd simply install ti-sgx-ti335x-modules-4.14.108-ti-r131, the kernel space driver package I found actually installed on the image, and the kernel to go with that, 4.14.108-ti-r131 Jan 01 06:29:25 that sounds plausible yes Jan 01 06:29:42 in reality, i already tried something like that and it didn't work, unless this very specific kernel revision + modules is the only one that works Jan 01 06:29:44 if that's the latest 4.14-ti kernel Jan 01 06:29:51 hmm Jan 01 06:30:11 i can again of course since I've now seen it working with my own eyes Jan 01 06:30:59 ( i believe the one i actually had tried was 4.14.108-ti-r137, iirc) Jan 01 06:32:14 ah I see he switched to 1.17 for kernel 4.19 because he had no option anyway: https://github.com/rcn-ee/ti-omap5-sgx-ddk-linux/branches/all Jan 01 06:33:01 also, he hasn't updated that mirror in quite a while Jan 01 06:33:18 oh, never mind, upstream also hasn't Jan 01 06:33:37 ( https://git.ti.com/cgit/graphics/omap5-sgx-ddk-linux/refs/heads ) Jan 01 06:40:02 well, since i tried using the 1.17 version without success, and the kernel that comes loaded in the latest image is only compatible with version 1.17, and the repository package is only compatible with version 1.14, does it make more sense to reflash with Debian 10 and then intentionally downgrade the kernel, or use dist-upgrade on the Debian 9 image that already has everything installed, without changing the kernel? I don't Jan 01 06:40:02 see what the problem would be if everything is still working after dist-upgrade Jan 01 06:40:39 downgrading the kernel makes more sense Jan 01 06:40:56 since 4.14 is still in use on buster too (the TIDL images for bbai use a 4.14-ti kernel) Jan 01 06:42:41 like, there's not really such a thing as a dependency on a distro version... it can have a dependency on a version of the kernel / driver version, and it may have dependencies on certain shared libraries (though its shared library dependencies appear fairly mundane), but a distro release is just a collection of packages in the end, most of which have nothing to do with this Jan 01 06:48:44 ok i am loading the debian 10 image to sd card and flashing, then i will need to recompile kernel space portion of my wifi software again, I did save a module but only for 4.19 kernel so i don't think that can speed it up Jan 01 06:49:07 why not use ethernet for now and deal with wifi later once sgx works? Jan 01 06:49:29 especially since last time I saw, your wifi driver was spitting out tracebacks in the kernel :P Jan 01 06:58:52 i started writing up a pretty long explanation of why, but does it make enough sense just to say my current hardware sitution makes it inconvenient to have network access on both the beagleboard and the device i prefer using irc on at the same time, without using wifi on the beagleboard? Jan 01 06:59:31 i understand the beaglebone head features intended to make tcp/ip over usb easy, but that is also inconvenient atm Jan 01 06:59:35 has* Jan 01 07:00:05 beaglebone has no specific features to that, that's all just linux Jan 01 07:00:31 well, not every computer i use linux on has a usb-b port, i guess that's all i mean Jan 01 07:00:48 (by saying the bbb makes that easier) Jan 01 07:08:12 if i keep having to change kernels for some reason, i will probably spend some time to try to change the hardware setup to cut out wifi. the way things stand, though, I've only seen the SGX530 functional on that one kernel, and the ~25 min compile time is shorter. Jan 01 09:49:21 hi Jan 01 17:53:20 Hi I wanted to ask about the Beaglebone AI board's supported layers. I see that deconvolution is supported but when I tried using STM's board, tensorflow's deconv layer threw errors because it involved a shape layer, which wasn't supported. Shape doesn't appear in the Beaglebone supported layer list either Jan 01 17:53:37 So does it support tensorflow's deconv layer? **** BEGIN LOGGING AT Fri Jan 01 17:57:33 2021 **** ENDING LOGGING AT Sat Jan 02 02:59:56 2021