| 00:39.31 | *** join/#brlcad kintel (~kintel@unaffiliated/kintel) |
| 01:37.42 | *** join/#brlcad LordOfBikes (~armin@dslb-088-065-180-115.088.065.pools.vodafone-ip.de) |
| 03:16.36 | *** join/#brlcad kintel (~kintel@unaffiliated/kintel) |
| 03:44.11 | *** join/#brlcad brlcad (~sean@104.225.5.10) |
| 04:57.40 | *** join/#brlcad kintel (~kintel@unaffiliated/kintel) |
| 09:04.18 | *** join/#brlcad Caterpillar (~caterpill@unaffiliated/caterpillar) |
| 10:15.08 | *** join/#brlcad davezd (~davezd@71-83-188-23.dhcp.lnbh.ca.charter.com) |
| 10:59.08 | *** join/#brlcad teepee (~teepee@unaffiliated/teepee) |
| 13:55.53 | *** join/#brlcad kintel (~kintel@unaffiliated/kintel) |
| 14:24.24 | *** join/#brlcad teepee (~teepee@unaffiliated/teepee) |
| 14:28.50 | *** join/#brlcad kintel (~kintel@unaffiliated/kintel) |
| 15:50.31 | *** join/#brlcad ARChak (~virususer@117.214.48.190) |
| 16:05.53 | *** join/#brlcad tandoorichick (~rakshika@60.243.170.241) |
| 16:27.21 | *** join/#brlcad ``Erik (~erik@pool-100-16-14-17.bltmmd.fios.verizon.net) |
| 16:31.44 | Notify | 03BRL-CAD:starseeker * 69397 (brlcad/trunk/NEWS brlcad/trunk/src/rt/viewweight.c): Have rtweight's loading of .density file use BU_DIR_SEPARATOR define. |
| 17:11.08 | *** join/#brlcad Caterpillar2 (~caterpill@unaffiliated/caterpillar) |
| 17:56.09 | *** join/#brlcad merzo (~merzo@91.151.136.88) |
| 17:56.40 | *** join/#brlcad teepee (~teepee@unaffiliated/teepee) |
| 18:33.18 | *** join/#brlcad kintel (~kintel@unaffiliated/kintel) |
| 19:17.59 | *** join/#brlcad yorik (~yorik@2804:431:f721:6f6b:290:f5ff:fedc:3bb2) |
| 19:35.49 | Apra | brlcad: Hey the channel seems to be quite empty :) |
| 19:42.57 | Apra | anyways what are your plans about this years gsoc? |
| 20:11.57 | *** join/#brlcad merzo (~merzo@91.151.138.242) |
| 20:18.48 | *** join/#brlcad Stragus_ (~alexis@modemcable090.29-19-135.mc.videotron.ca) |
| 20:23.32 | *** join/#brlcad kintel (~kintel@unaffiliated/kintel) |
| 20:37.14 | ``Erik | hw.byteorder: 1234 |
| 20:37.20 | ``Erik | the world seems right again O.o |
| 20:38.07 | ``Erik | hrm, scratch that, the x86 says the same thing, hopefully I didn't accidently put this thing in lil' endian mode |
| 20:50.31 | Stragus | Playing on ARM? |
| 20:56.07 | ``Erik | yeah, it's my new 'main' server, a rpi2 replaced the old thunderbird |
| 20:56.11 | ``Erik | FreeBSD fenris 12.0-CURRENT FreeBSD 12.0-CURRENT #0 r313175: Mon Feb 6 08:06:12 EST 2017 erik@rpi2:/mnt/da0s1/usr/obj/mnt/da0s1/usr/src/sys/RPI2 arm |
| 20:56.41 | Stragus | I absolutely love ARM. Neon is wonderful |
| 20:57.04 | ``Erik | O.o just a few years ago, you said arm was crap and it was all about amd64 :D |
| 20:57.15 | Stragus | I said that? Darn :( |
| 20:57.46 | Stragus | That was before I discovered Neon surely, but that still surprises me |
| 20:58.19 | Stragus | NEON is everything MMX/3dnow!/SSE/SSE2/SSE3/SSSE3/SSE4.1/SSE4.2/AVX/AVX2 should have been... implemented properly right away, on the first shot |
| 20:58.31 | ``Erik | iirc, we were chatting about server farms and your focus at the time was sse heavy C |
| 20:58.44 | ``Erik | erm, and tons of x86 asm *shudder* :D |
| 20:58.59 | Stragus | Eheh, so definitely before I discovered Neon |
| 21:00.01 | ``Erik | I wonder how neon compares to, say, altivec |
| 21:00.15 | ``Erik | (cuz altivec was everything that mmx/3dnow/sse*/etc was supposed to be) |
| 21:00.54 | Stragus | Right... I read the Altivec docs but never used it. But Neon has some amazing stuff you won't find in Altivec either, as far as I know |
| 21:01.02 | ``Erik | Stragus: one of the projects I'm sorta attached to has an rpi3 driving a tx1 in a mobile package O.o pheer |
| 21:02.08 | Stragus | Neon can read a bunch of 16 floats (xyzw,xyzw,xyzw,xyzw), de-interleave and store in 4 separate registers (xxxx,yyyy,zzzz,wwww). And that's a... single cheap instruction |
| 21:02.39 | Stragus | It's just fantastic, everything people need to do in real life |
| 21:04.03 | ``Erik | interesting, but why not ship that to the gpu? |
| 21:05.27 | Stragus | Sure, CUDA is also fabulous... 32-wide SIMD with fast scattered loads/stores, per-lane predicates and branching, on-chip shared memory, hardware switching to other threads whenever waiting for *anything* |
| 21:05.30 | Stragus | CPUs suck :p |
| 21:06.06 | Stragus | Still, in the realm of CPUs, ARM NEON does a pretty good job |
| 21:08.04 | ``Erik | (I'm still saddened that the rpi is setting the arm to LE, and doubt the port is good enough to do BE) |
| 21:08.53 | ``Erik | but it replaces a noisy powerhungry slow pos and seems to be doing a solid job so far, so *shrug* :) |
| 21:08.59 | Stragus | I think the world is setlling on little endian, just like it settled on bytes being 8 bits |
| 21:09.17 | Stragus | settling* |
| 21:09.40 | ``Erik | bleh, VHS instead of beta, bluray instead of hddvd or whatever, ... |
| 21:10.20 | *** join/#brlcad kintel (~kintel@unaffiliated/kintel) |
| 21:11.02 | ``Erik | I wonder if BRL-CAD will run on it... I bet it'll perform like complete crap in the benchmark :D |
| 21:12.17 | Stragus | Eheh. I'm not aware of optimization advantages to big endian, perhaps I'm missing something... |
| 21:13.07 | ``Erik | the only advantage of big endian is that I'm old and back in the day, the cool machines all used big endian, so now I'm sitting here shaking my cane and remembering the good old days |
| 21:13.17 | ``Erik | y'know, make 'murika grrt agn |
| 21:13.34 | Stragus | Oh, right, of course! :) |
| 21:14.49 | ``Erik | the 'server room' is now very quiet, however, with the multiple fans and old 20g 5400 shut down, the only moving part is the pidrive |
| 21:15.17 | ``Erik | which is basically just a laptop drive with some modified plugs and less case |
| 21:15.18 | Stragus | is trying to think of any optimization only possible on big endian... |
| 21:17.20 | ``Erik | the only real advantage is converting to/from network order (noop) |
| 21:18.34 | ``Erik | iirc, even simple things like an adder uses a few less transistors in big endian, so there's a cost in silicon real estate... though the arm (and several other more modern chips) have switchable endian |
| 21:18.41 | Stragus | I was trying to think of rather inherent advantages |
| 21:19.49 | Stragus | Ahah, Google found something! memcmp() on big-endian is more efficient (assuming aligned pointers, etc.) |
| 21:23.06 | ``Erik | when you do neon stuff, is that in asm, or is there an intrinsics package or api or something? |
| 21:23.38 | Stragus | They have intrinsics. The compilers don't do a good job though, so people often rely on inline assembly |
| 21:24.01 | Stragus | SSE/AVX has received more loving care in GCC than Neon |
| 21:24.48 | Stragus | Have a look at arm_neon.h |
| 21:26.36 | ``Erik | hm, not part of my clang distro :) |
| 21:28.28 | ``Erik | not in the raspbian image, either... ah well, if I get to needing it, I'll find it |
| 21:32.11 | Stragus | Feel the call of GCC, it's claiming for you |
| 21:38.39 | ``Erik | swift! sbcl! |
| 21:47.16 | *** join/#brlcad kintel (~kintel@unaffiliated/kintel) |
| 21:52.27 | Notify | 03BRL-CAD:starseeker * 69398 brlcad/trunk/src/external/CREO/creo-brl.cpp: Next problem is in assemblies - got a crasher in here somewhere. |
| 22:16.00 | *** join/#brlcad teepee (~teepee@unaffiliated/teepee) |
| 22:17.16 | *** join/#brlcad Caterpillar2 (~caterpill@unaffiliated/caterpillar) |
| 23:53.50 | *** join/#brlcad kintel (~kintel@unaffiliated/kintel) |