00:00:00 --- log: started forth/15.05.17 00:22:23 --- quit: cataska (Quit: Connection closed for inactivity) 01:18:14 --- join: true-grue (~grue@95-27-138-9.broadband.corbina.ru) joined #forth 01:35:37 --- quit: scoofy (Ping timeout: 256 seconds) 02:20:19 --- join: scoofy (~scoofy@catv-89-135-64-110.catv.broadband.hu) joined #forth 02:29:42 --- quit: mnemnion (Ping timeout: 265 seconds) 02:43:19 --- join: cataska (uid12717@gateway/web/irccloud.com/x-otqirjbrtskvcras) joined #forth 02:52:00 --- join: Bahman (~Bahman@89.43.90.117) joined #forth 03:00:55 --- quit: scoofy (Ping timeout: 240 seconds) 03:02:48 --- join: scoofy (~scoofy@catv-89-135-64-110.catv.broadband.hu) joined #forth 03:16:00 --- join: GeDaMo (~GeDaMo@dyn-62-56-76-198.dslaccess.co.uk) joined #forth 03:26:08 --- quit: Bahman (Read error: Connection reset by peer) 03:54:37 --- join: proteusguy (~proteusgu@180.183.110.30) joined #forth 03:54:37 --- mode: ChanServ set +v proteusguy 04:04:31 --- join: Bahman (~Bahman@89.43.90.117) joined #forth 04:24:57 --- join: vanila (~vanila@unaffiliated/vanila) joined #forth 04:34:00 --- quit: ASau (Remote host closed the connection) 04:34:32 --- join: ASau (~user@176.0.114.49) joined #forth 05:00:43 --- join: scoofy_ (~scoofy@catv-89-135-80-2.catv.broadband.hu) joined #forth 05:01:22 --- quit: scoofy (Ping timeout: 272 seconds) 05:08:41 --- quit: Bahman (Ping timeout: 250 seconds) 05:12:38 --- join: Bahman (~Bahman@89.43.90.117) joined #forth 05:29:22 --- quit: proteusguy (Remote host closed the connection) 05:32:00 --- join: proteusguy (~proteusgu@180.183.110.30) joined #forth 05:32:00 --- mode: ChanServ set +v proteusguy 05:42:24 --- quit: cataska (Quit: Connection closed for inactivity) 05:42:41 --- quit: Bahman (Quit: Ave atque vale) 06:31:36 --- join: Bahman (~Bahman@89.43.90.117) joined #forth 06:52:20 --- quit: Bahman (Quit: Ave atque vale) 06:53:40 --- join: Bahman (~Bahman@89.43.90.117) joined #forth 07:07:30 --- quit: true-grue (Read error: Connection reset by peer) 07:25:40 --- quit: Bahman (Quit: Ave atque vale) 07:32:53 --- join: cataska (uid12717@gateway/web/irccloud.com/x-tlpddoughtqusktz) joined #forth 07:37:38 --- quit: darkf (Quit: Leaving) 08:17:59 --- join: nighty^ (~nighty@hokuriku.rural-networks.com) joined #forth 08:20:12 --- join: kumul (~mool@adsl-64-237-238-168.prtc.net) joined #forth 08:39:10 --- quit: nighty^ (Read error: Connection reset by peer) 08:39:28 --- join: nighty^ (~nighty@hokuriku.rural-networks.com) joined #forth 08:51:40 --- quit: proteusguy (Remote host closed the connection) 09:06:09 --- join: mnemnion (~mnemnion@c-24-23-168-196.hsd1.ca.comcast.net) joined #forth 09:12:35 --- join: true-grue (~grue@95-27-138-9.broadband.corbina.ru) joined #forth 09:15:47 --- join: proteusguy (~proteusgu@ppp-110-168-229-123.revip5.asianet.co.th) joined #forth 09:15:47 --- mode: ChanServ set +v proteusguy 09:44:46 --- quit: impomatic (Ping timeout: 245 seconds) 09:52:20 --- join: bedah (~bedah@dyndsl-095-033-179-093.ewe-ip-backbone.de) joined #forth 10:05:07 --- join: impomatic (~impomatic@31.185.153.201) joined #forth 10:15:59 --- join: ASau` (~user@176.0.80.53) joined #forth 10:19:27 --- quit: ASau (Ping timeout: 246 seconds) 10:21:27 --- join: vsg1990 (~vsg1990@cpe-67-241-148-119.buffalo.res.rr.com) joined #forth 10:41:36 --- quit: kumul (Ping timeout: 272 seconds) 10:47:52 --- quit: impomatic (Ping timeout: 252 seconds) 10:52:24 --- quit: cataska (Quit: Connection closed for inactivity) 10:53:43 --- join: impomatic (~impomatic@31.185.153.201) joined #forth 11:14:29 --- join: fantazo (~fantazo@089144224125.atnat0033.highway.a1.net) joined #forth 11:34:17 --- join: Zarutian (~Adium@46.22.110.168) joined #forth 11:43:43 --- join: Mat4 (~quassel@ip5b407cc8.dynamic.kabel-deutschland.de) joined #forth 11:45:37 hello 11:58:35 h'lo Mat4 12:01:07 been thinking a bit about micro architecture of CPUs 12:02:11 hi Zarutian 12:02:47 and what are your conclusions so far ? 12:02:47 one of them choices in that space is pipelining which is a result of, I think, of too long decoding and performace stages. 12:03:55 my preliminary tentative conclusions so far is that there is all this crap heaped on because there werent much thought on what the problem is they were trying to solve 12:06:14 the problem in the coarse can be expressed by the refrain "this program is so damn fucking slow!" 12:06:34 (or just "it's so fucking slow!") 12:08:53 "The Machine puts data first. The concept is to build a system around nonvolatile memory with processors sprinkled liberally throughout the memory. When you want to run a program you send the program to a processor near the memory" http://highscalability.com/blog/2014/12/15/the-machine-hps-new-memristor-based-datacenter-scale-compute.html 12:09:13 Zarutian: yes and no, pipelining ease also design for very short clock ranges in GHz range. I would say for superscalar out-of-order archictectures is it a condition for reach higher clock rates 12:10:04 ^to reach higher clock ranges 12:10:25 Mat4: lets break that down a bit. Superscalar means SIMD in most cases? (like bliting of yore being a very crude example) 12:11:35 Superscalar means multible pipelines for different units (a simplification) 12:12:28 indeed, but that results in complex circuitry, yes? 12:13:02 an out-of-order architecture encode operation codes to some kind of instruction pool from which these pipelines are filled *out of order* 12:13:36 most of these pipelines, at any given time, seem to be stalled 12:14:12 resulting in most of the circuitry being idle 12:14:21 mostly not, there are stalled only in case of mispredicted branches 12:15:09 and because the circuit complexity is high and pipelines tend to be long this limit performance significant 12:15:19 which is fine and dandy if you only have nearly straight line code but which incure instruction cache penalty 12:15:29 exactly 12:15:56 that's the main limitation of such architectures 12:16:55 all of these cost also energy 12:17:00 of course... 12:18:53 I think that's a dead end 12:21:45 but for sure some kind of pipelining is needed for processing high clock rates because otherwise logic paths get simply to wide. You can see also a connection between increasing of pipeline stages and clock rate 12:52:24 --- quit: TodPunk (Ping timeout: 246 seconds) 12:56:00 --- join: kumul (~mool@adsl-64-237-236-232.prtc.net) joined #forth 13:03:32 --- quit: tsumetai (Ping timeout: 252 seconds) 13:15:25 --- quit: kumul (Ping timeout: 276 seconds) 13:19:02 --- quit: fantazo (Quit: Verlassend) 13:21:03 --- join: TodPunk (Tod@50-198-177-186-static.hfc.comcastbusiness.net) joined #forth 13:25:27 --- join: pointfree (~pointfree@c-73-142-162-42.hsd1.ma.comcast.net) joined #forth 13:27:40 Mat4: what do you mean by 'logic paths get simply too wide"? 13:29:34 besides I think Von Neuman arcitechture is okay in the small and as units but not at higher design levels. 13:31:15 these CPU's have a large set of units, some even duplicated and really complex. In addition there exist dependencies of instructions and intermediate results to be handled. All this add to the time instructions need for execution. 13:32:09 indeed, which mean that those instructions arent atomic or simple enough 13:32:15 If the clock rate is high the time needed for total instruction execution get higher than the time follwing instructions are issued 13:32:31 ^following 13:32:54 frankly I would rather have something like transputer with extremely simple cores (each with local memory) 13:34:30 Or like the GA144? :P 13:35:05 --- join: Bahman (~Bahman@89.43.90.117) joined #forth 13:35:34 note sure about the GA144. Something more like J1 cores I could live with. 13:35:38 not* 13:42:02 probably the Propeller MCU is something for you 13:43:04 I am bit familiar with it. 13:45:31 it only has what? eight cores? I am talking about at least 16x16 grid of cores+local_mem 13:47:48 you need a bus and some external sram. Communication can be done like the Transputer 13:48:14 the hive project realize something similar with 3 propellers 13:49:12 (the main difference is that these MCU's are speciallized) 13:49:51 one for memory managment, one for video generation and one for program execution 13:51:34 well, what I have in mind is something like J1 or RTX2010 cores where half of or all of each local memory is FRAM based. 13:51:51 and minimum that 16x16 grid 13:52:28 FRAM? 13:52:28 what you described, Mat4, sounds basically like what most PCs are today. 13:53:20 GeDaMo: Ferretic RAM iirc. Texas Instruments has a bit about that technology. Basically it is as fast as SRAM but more durable than Flash. 13:53:35 Yeah, I'm just looking 13:54:01 Did you see my link earlier on HP's Machine? 13:54:27 so what I described is a bit in the direction of FPGAs and CPLDs. 13:54:34 GeDaMo: yes 13:55:16 I like memristors. But what FRAM has over memristors now is that FRAM parts are aviable today. 13:56:06 there are even pin, voltage and protocol comapatible chips meant to replace Flash ones. 13:56:52 GeDoMa: beats me though how the hell you program those HP machines. 13:57:20 there exist variants of TI's MSP series with integrated FRAM memory 13:57:28 even very cheap 13:57:30 Something like map reduce I assume 13:57:49 GeDoMa: at least what I described can be programmed by the Flow Based Programming paradigm (oh, and nice FBP enviroment I have been recommended is Pure Data) 13:58:04 Or like a distributed database query 13:58:14 Zarutian: http://hive-project.de/ 13:58:17 functional or declarative programming (thing prolog) 13:58:26 think* 13:58:41 Or array programming 13:59:01 GeDaMo: array programming? like APL? 13:59:10 Yeah 13:59:28 Mat4: Oh, I see 14:04:29 GeDaMo: hmm I think array programming is basically what modern day GPUs are programmed 14:04:52 is it? hrm. 14:05:09 Similar idea, yes 14:05:12 yes i suppose so. 14:05:13 SIMD 14:05:29 * tangentstorm uses J, another language by the guy that created APL. 14:05:37 tangentstorm: think of the shaders working with textures and vertices as working with arrays of certain kinds. 14:05:52 I like the concepts in array languages, not keen on the syntax :P 14:06:22 concepts such as? 14:06:26 hi tangentstorm 14:06:35 heya Mat4 14:07:40 btw, is mosis a nice company for IC designer hobbyists like me? 14:14:29 --- nick: ASau` -> ASau 14:32:22 --- quit: bedah (Quit: quickquit) 14:35:22 ciao 14:35:31 --- quit: Mat4 (Quit: http://quassel-irc.org - Chat comfortably. Anywhere.) 14:50:56 --- quit: GeDaMo (Remote host closed the connection) 15:38:38 --- quit: true-grue (Read error: Connection reset by peer) 15:42:25 --- quit: Bahman (Quit: Ave atque vale) 16:30:43 --- quit: Zarutian (Quit: Leaving.) 16:46:12 --- quit: nighty^ (Quit: Disappears in a puff of smoke) 17:34:41 --- quit: vanila (Quit: Leaving) 17:36:44 --- join: johnmark_ (~johnmark@24.1.223.32) joined #forth 17:37:29 --- nick: johnmark_ -> JohnMarkM 18:00:18 --- quit: vsg1990 (Quit: Leaving) 18:37:43 --- join: MrMobius (~Joey@h220.165.28.71.dynamic.ip.windstream.net) joined #forth 18:46:07 --- join: darkf (~darkf___@unaffiliated/darkf) joined #forth 18:55:40 --- join: atommann (~atommann@58.251.2.94) joined #forth 18:56:13 --- quit: MrMobius (Read error: Connection timed out) 19:39:17 --- join: kumul (~mool@adsl-64-237-237-40.prtc.net) joined #forth 20:06:46 --- join: cataska (uid12717@gateway/web/irccloud.com/x-neqrnhwfzjbgsghu) joined #forth 20:29:25 --- quit: JohnMarkM (Quit: Leaving) 20:52:38 --- quit: dys (Ping timeout: 272 seconds) 21:25:37 --- quit: proteusguy (Ping timeout: 256 seconds) 21:37:39 --- join: proteusguy (~proteusgu@ppp-110-168-229-136.revip5.asianet.co.th) joined #forth 21:37:39 --- mode: ChanServ set +v proteusguy 22:15:48 --- quit: atommann (Ping timeout: 264 seconds) 22:18:30 --- quit: ASau (Remote host closed the connection) 22:20:05 --- join: ASau (~user@176.0.80.53) joined #forth 22:26:01 --- quit: ASau (Ping timeout: 245 seconds) 22:50:35 --- join: dys (~andreas@2a01:1e8:e100:8296:f66d:4ff:fec1:4421) joined #forth 23:01:17 --- quit: proteusguy (Remote host closed the connection) 23:20:37 --- join: atommann (~atommann@58.251.2.94) joined #forth 23:59:59 --- log: ended forth/15.05.17