**** BEGIN LOGGING AT Sun Nov 04 02:59:59 2018
Nov 04 04:32:53 <NishanthMenon> zmatt, poked internally
Nov 04 05:01:59 <zmatt> NishanthMenon: so, the exact issue is (afaict): the cortex-a15 uses WRAP mode (linefill-like) bursts for reads to memory that has been mapped using the "normal, non-cacheable" memory type.  TILER however does not support WRAP mode bursts (to the 8/16/32-bpp views) and throws a bus error
Nov 04 05:03:58 <zmatt> so it's basically an integration issue... a slightly odd design decision of the cortex-a15 in combination with a limitation of TILER that was never a huge problem
Nov 04 05:07:28 <NishanthMenon> zmatt, let me see what the guys come back with. i have been a bit out of touch recently
Nov 04 05:09:15 <zmatt> the weird thing is, the idea behind the cortex-a15 doing this (afaict) is that it could combine reads to normal non-cacheable memory that are performed closely together... but this clearly doesn't happen: performing back-to-back sequential single-word reads from DDR3 memory mapped as normal non-cacheable is in fact pig slow, much slower even than when mapping is as device or strongly-ordered (by a ...
Nov 04 05:09:21 <zmatt> ...factor of 5)
Nov 04 05:10:37 <zmatt> NishanthMenon: yeah, no rush... allocating an erratum number for it is already 3.5 years late anyway ;)
Nov 04 05:11:28 <zmatt> I was just giving my findings, since the commit message in linux is very vague about the problem: https://patchwork.kernel.org/patch/6087391/
Nov 04 05:19:30 <zmatt> (I'm currently working on patches to improve TILER's performance in linux on omap5/dra7 again to the point of being very much usable again)
Nov 04 05:28:17 <NishanthMenon> zmatt, i have poked tomba on it. best might be to drop a note on linux-omap ML.. will let tomi directly respond on list
Nov 04 05:28:44 <zmatt> fair point
Nov 04 06:08:20 <ayjay_t> its been just under two weeks since i started with golang and i'm already using their unsafe types
Nov 04 06:08:48 <ayjay_t> trying to steal the 16 MSB of a 64-but UNIX timestamp to store user data
Nov 04 20:08:52 <\dev\cache> Hi guys
Nov 04 20:08:57 <\dev\cache> can anyone recommend a case for an X15?
Nov 04 21:54:15 <florian> \dev\cache: too big for regular 3d printed cases... https://en.wikipedia.org/wiki/North_American_X-15
Nov 04 21:54:24 <florian> scnr ;)
Nov 04 22:25:59 <ayjay_t> hehe
Nov 05 02:41:11 <set_> Charge your batteries!
Nov 05 02:41:26 <set_> Robot Day = Monday
Nov 05 02:42:51 <set_> Has anyone ordered the RelayCape or LoadCape yet?
Nov 05 02:43:02 <set_> Oh!
Nov 05 02:43:25 <set_> I went to mouser.com to get a new BBBW. They are backlogged for umpteen months. Boo!
Nov 05 02:43:47 <set_> People love WiFi.
Nov 05 02:48:53 <ds2> depends on where you live
Nov 05 02:49:06 <ds2> WiFi is really blowing around here...the 2.4GHz space is so saturated :(
Nov 05 02:51:32 <set_> Oh? That is not all that good for people.
Nov 05 02:51:50 <set_> ds2: I am in the states right now.
Nov 05 02:52:14 <set_> Things are working for me, sort of.
Nov 05 02:56:05 <set_> Are you in a large city right now?
Nov 05 02:56:15 <set_> Is that why things are so "saturated?"
Nov 05 02:56:48 <set_> High rises and Skyscrapers may take away from 2.4GHz b/c of the pop.
Nov 05 02:57:22 <set_> Like I say, "Too many gadgets, so little time."
**** ENDING LOGGING AT Mon Nov 05 03:00:00 2018