IRC logs for #openrisc Monday, 2016-09-19

--- Log opened Mon Sep 19 00:00:18 2016
wallentoolofk, I see01:49
wallentobut I am still a bit concerned about the filesets in this context. They were kind of transparent to he backend so far, right?01:50
olofkNot sure what you mean02:22
wallentoSo far all sources, includes etc. were flattened02:32
wallentoso there is no simple way to distinct two dpi libraries via filesets, right?02:35
olofkNo, exactly. That needs to be solved if we want multiple dpi sections02:35
wallentoeven, one library may be split over multiple filesets02:35
wallentoI am not sure if we want them, but lets assume so02:35
olofkI think we could assume that. One reason could be that we want to load different files depending on the simulator02:36
wallentois there any restriction or plan to restrict filesets?02:36
olofkrestrict how?02:36
wallentoso far its only a structural element on the input side02:36
wallentoexactly, via the usage02:36
wallentobut then it will not appear in one simulator, right?02:36
wallentoso its virtually only one02:36
olofkCan you elaborate?02:37
wallentotwo DPI: the usage will filter out which files to use, right?02:38
wallentoregarding restriction: Like saying all DPI files must be in one fileset if they form a library02:38
wallentobut actually it may not be necessary02:38
wallentoone last thing: how to distinct DPI and VPI then?02:38
wallentocSource can be both, right?02:39
wallentoactually it can also be none of both, but a file to be compiled with verilator02:39
wallentoI will do the fileset stuff, but it would be great if you could send me an example fileset02:39
wallentohow DPI vs. verilator should look like02:39
wallentoafk for some hours02:40
olofkYes. I'm starting to think that maybe we should have both02:44
olofkPut the files in filesets to get the automatic export, usage, scope etc. And then mark them as source files in the respective vpi/dpi sections02:44
olofkThe drawback will of course be that they need to be specified twice, so I'm not sure it's the best solution02:45
olofkIf anyone happens to be in Stockholm 13 October, I'll be doing a talk on FOSSi stuff
olofkwallento: Looking through the IP-XACT standard to see if there is anything there that can be useful for the DPI stuff06:06
olofkOnce again I'm reminded of what a complete disaster of a standard that is06:06
olofkI'm amazed how they specify some things so hard, while other are completely left open06:06
olofkSo far FuseSoC has only used the sane parts of the standard, and it will remain like that06:07
olofkDesign by comittee taken to its extreme06:08
olofkFor example, they have decided to enumerate the valid return types of a function. The list contains of int and void06:09
stekernolofk: is that breakfeast seminar open for everybody? (i.e. do you mind if I hint about it to some friends in sthlm?)06:41
olofkstekern: Please spread the word. It's open to everyone and I hope to see some new faces06:44
stekerncool, will do06:44
stekernI'll be on the other side of the pond at that date, otherwise I might have tried to attend06:47
olofkAh ok, but when I said it's open for everybody, I didn't mean it's open for you06:48
olofkWe do have some standards06:48
stekernoh, I should have figured :(06:48
stekernI would of course been coming disguised06:49
stekern...and you know, asking obnoxious questions, just like your friends did during your thesis presentation.07:27
stekern(or mine at least)07:28
olofkMy thesis presentation was so bad that no one knew where to begin asking questions :)07:38
SMDwrkolofk: what was the topic?07:39
olofkSMDwrk: Turning simulink models into HDL code and execute them in LabVIEW's FPGA environment07:48
olofkSpend six months drawing blocks and wires and struggling with really crappy and expensive tools07:49
shornestekern: some interesting responses came in from Jonas on the kernel patches I sent10:38
-!- Mentally_Ill is now known as hammond10:45
shornewallento: thanks for you replies on the mailing list, maybe jeremybennett can help us get those last assignments and we can get our toolchains upstreamed12:11
jeremybennettshorne: I've pinged Damjan on LinkedIn12:43
jeremybennettHowever I'm not sure how effective that will be. He is shown as CEO of Kulfun Games, but that appears to be defunct - looks like the games were acquired by another company. He may be running a company called Kulone in Belgrade now, but I couldn't easily track it down.12:44
stekernshorne: yes, I saw12:48
jeremybennettstekern: shorne: Had a reply straight back from Damjan. He did a FSF copyright assignment years ago, which the FSF should have on file.13:34
stekernjeremybennett: yes, it wasn't he that was missing13:37
jeremybennettstekern: Have you checked if Matjaz has an assignment on file already. He might do if the Flextronics guys submitted en masse.13:50
jeremybennettThere might even be a generic Flextronics assignment.13:50
stekernI haven't checked anything ;)14:00
stekernbut I know that much that he didn't do that14:00
stekernjust noticed that github reviews have got some nice features14:00
stekernnice *new* features14:00
jeremybennettstekern: shorne: Just emailed you a little more info that may help.14:04
-!- hammond is now known as hammond-ey16:30
shornejeremybennett: Damjan replied and said he alrady assigned hit copyrights to fsf.  Thank you very much17:06
shornejeremybennett: I see, you also had got a reply17:09
olofkHmm.... looks like the embedded wb_sdram_ctrl in wb_altera_ddr_wrapper is killing much of my memory bandwidth17:11
olofkI think I have some improvements I made two years ago somewhere17:12
olofkBut I don't know if that's enough17:12
olofkMight have an idea now17:17
olofkIf I expose the burst length from mor1kx, I could just use a regular arbiter with an avalon interface on the downstream side17:17
olofkstekern: burst lengths are static or at least easily predictable in mor1kx, right?17:18
olofkWishbone really needs a burst length field. It would make interfacing with other protocols so much easier17:19
-!- Netsplit *.net <-> *.split quits: wallento, andrzejr17:51
-!- Netsplit over, joins: wallento17:51
-!- Netsplit over, joins: andrzejr17:51
stekernolofk: burst lengths depend on cache line width22:37
stekernso, yes they are static (for a given setup)22:38
--- Log closed Tue Sep 20 00:00:20 2016

Generated by 2.15.2 by Marius Gedminas - find it at!