#milkymist IRC log for Friday, 2012-02-03

GitHub129[scripts] xiangfu pushed 1 new commit to master: http://git.io/BRvp0g01:53
GitHub129[scripts/master] m32 toolchain build script: fix parallel build - David Kuehling01:53
cladamwawpwrak, do you think we need completely remove C238?02:04
GitHub34[scripts] xiangfu pushed 1 new commit to master: http://git.io/9_VGRg02:09
GitHub34[scripts/master] compile-flickernoise: removes unnccessary subshell parens - Xiangfu Liu02:09
wpwrakcladamwa: yes, C238 must die :)02:11
cladamwawpwrak, since to duplicate PTC current protection, I hope I can find smd type of Zener to replace through hole ones. not sure if I can find. I'll try. since current through hole will annoy me. ;-)03:05
cladamwawpwrak, since I don't want to bend pins of Zener before sending them to factory. see works i did: http://en.qi-hardware.com/wiki/File:Bending_zerner_diode.JPG03:08
wolfspraulcladamwa: please search on digikey (or octopart) first03:08
cladamwawpwrak, too much work though.03:09
wolfspraulpreferably digikey03:09
cladamwawolfspraul, sure, will search digikey first.03:09
wolfspraulyes, definitely. good! :-)03:09
wolfspraul(no bending)03:09
wpwrakthe bending looks lovely ;-)03:19
cladamwa(newest) http://downloads.qi-hardware.com/hardware/milkymist_one/sch/tmp/MILKYMISTONE.pdf03:25
cladamwawpwrak, I added second 2A protection for usb in Power.SchDoc, few details need to chat with you. ;-)03:26
cladamwa1. since I expected extra-ext. usb and int. usb will occupied present area a lot, so I'd like to keep second circuit of F3/D28/D29 etc still close to J11 DC jack area, how so you think?03:29
cladamwa2. if like this, then the net of "USB5V' will go long far then get reach usb c ~ f circuit routing area. if this being established, it would be still added a C286/220uF to reduce noise while long routings. The C286 will be placed close to usb c~f area, how do you think?03:32
cladamwa3. if against item 1&2, we may have another alternative, thus is placing second F3/etc. to the bottom side of pcb of planed usb c~f area. then C286/220uF could be reduced as possible.03:36
cladamwathe 3) is probably not work since also arising routing complexity in those area, so item 1&2 would seemly be possible solution. or do you have any idea?03:39
qi-botThe Firmware build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/milkymist-firmware-20120203-0300/03:47
wpwrak(1) keeping it close to DC jack sounds good04:00
wpwrak(2) hmm, c-f are internal and external ? may be easier to have all the external usb on one protection circuit, and everything that's internal (usb and the rest of the system) on the other04:02
cladamwaa~c now are external, e/f are internal.04:27
cladamwawpwrak, your (2) idea is much better. tks. will change it.04:29
qi-botThe Firmware build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/milkymist-firmware-20120203-0458/05:45
cladamw(Zener smd 5.6V 5W) http://search.digikey.com/us/en/products/SMBJ5339B-TP/SMBJ5339B-TPMSCT-ND/163620507:26
cladamwin R3, we used a 0.57 usd/ea: http://search.digikey.com/us/en/products/1N5339BG/1N5339BGOS-ND/147407907:27
cladamwI'd prefer to go for this smd type even it's @0.9 usd/ea, that could save productive time a lot of course. The problem is available sotck of DO-2144A, SMB is only 105pcs. A though hole type is 6,496 pcs available.07:32
cladamw(SMBJ5339B) good news on Mouser that is 4,447 pcs available now, don't need to worry. so last time R3 production, didn't see Mouser has this part came into their warehouse.07:37
GitHub76[llvm-lm32] jpbonn pushed 150 new commits to master: http://git.io/GtoUnA07:54
GitHub76[llvm-lm32/master] jalr should use t9 ($25) for indirect calls regardless of the relocation model - Akira Hatanaka07:54
GitHub76[llvm-lm32/master] Rename WrapperPIC. It is now used for both pic and static. - Akira Hatanaka07:54
GitHub76[llvm-lm32/master] Typo. - Chad Rosier07:54
GitHub177[clang-lm32] jpbonn pushed 92 new commits to master: http://git.io/vlajrQ07:54
GitHub177[clang-lm32/master] This patch extends thread safety analysis with support for the scoped_lockable attribute. - DeLesley Hutchins07:54
GitHub177[clang-lm32/master] add tree test for suse on ppc64 (r146142) - Hal Finkel07:54
GitHub177[clang-lm32/master] Fix an edge case in IRGen for conditionals.  PR11509. - Eli Friedman07:54
Fallenoularsc lekernel : is it normal that at some point in lm32 simulation mem[0] becomes 0x00000000 , then mem[1] as well etc gradually all instruction SRAM clears itself09:06
Fallenouuploading bios.elf to show you the code I am running which is pretty straight forward09:07
Fallenouand does not do any memory load/store09:07
lekernelno, it is not09:12
Fallenousource code is here : http://pastebin.com/5tmrBnYM09:12
Fallenouand EBA s initialized to 0 in lm32_include.v09:12
Fallenouso next I will try to understand why sram is zeroing itself up :/09:12
lekernelby the way, why bother with the full BIOS if you're only testing a few MMU instructions?09:12
FallenouI removed almost everything09:13
lekernelbtw if you commented out bi      main it will continue into save_all etc.09:13
FallenouI just zero r0, set EBA and jump to _ctr009:13
Fallenouand then do a while(1) nop09:13
lekerneltry simply _reset_handler: bi _reset_handler:09:14
lekernelmaybe with xor     r0, r0, r009:14
lekernelbut no moe09:14
Fallenouhum ok09:15
larsclekernel: i send you a migen pull request on github09:21
lekernelah you already did... for some reason it didn't show up in my notifications09:24
larscuh, yes, sorry, s/send/sent/09:24
larscmy brain is still frozen ;) -15C outside09:25
GitHub111[migen] sbourdeauducq pushed 4 new commits to master: https://github.com/milkymist/migen/compare/3143608...504a16909:25
GitHub111[migen/master] Lower required python version to 3.1 - Lars-Peter Clausen09:25
GitHub111[migen/master] fhdl/namer: Add support for STORE_DEREF opcode - Lars-Peter Clausen09:25
GitHub111[migen/master] Use enumerate(x) instead of zip(range(x), x) - Lars-Peter Clausen09:25
Fallenoulekernel: at startup the icache flushes and invalidates itself before anything happen, ok normal, then cpu I_ADR_O starts incrementing (0, 4, 8 etc) and at this exact moment sram starts to zero itself09:30
Fallenouand so the way_data of the icache starts to refill itself, storing the 0 of the sram09:31
Fallenouinstead of storing what used to be the real instructions09:31
Fallenoumakes me crazy09:31
FallenouI did as you said xor r0,r0,r0 then bi _reset_handler09:31
Fallenouand this is a non modified lm32 cpu, the lm32_dcache is the original one, not the mmu-modified-one09:32
Fallenouvery weird09:33
lekernelwhat's the SRAM WE signal?09:35
lekernelnote that I did not test this stuff, so there might be some bugs09:35
Fallenouit seems there is no WE signal09:36
FallenouI cannot see one in soc.V09:36
Fallenouoh sorry yes there is09:36
Fallenouok let's add it to the wave window09:37
Fallenoulekernel: yep frag_sram0[3:0] keeps toogling between 4'd0 and 4'b111109:40
Fallenougrmbl why09:41
lekernelshould be better now *g*09:43
GitHub113[milkymist-ng] sbourdeauducq pushed 1 new commit to master: http://git.io/6iDnLA09:43
GitHub113[milkymist-ng/master] sram: fix WE signal - Sebastien Bourdeauducq09:43
Fallenoufrag_sram0[k] <= ((sram0_wishbone_cyc_i & sram0_wishbone_stb_i) & sram0_wishbone_sel_i[k]);09:43
Fallenouah yep09:44
Fallenouthanks :)09:44
Fallenouwhy 3-i ? wishbone is little endian ?09:46
lekernelno, it's big endian...09:46
lekernelbut migen has reversed signal indexes (compared to verilog) to be consistent with python slices09:48
Fallenouoh yep I just noticed09:48
Fallenoufrag_sram0[0] means mem[n][7:0] and not [31:24]09:49
Fallenoulet's try again09:49
Fallenoulekernel: ok awesome !10:03
Fallenouworks way better now10:03
Fallenoufetches 0, 4, 8, c then 0 and stays at 0 for a reason I don't know yet10:03
Fallenouand way_data starts to fill up with real instructions :)10:04
FallenouI guess bi 0 is not super nice to lm3210:04
lekernelno, it should work10:11
lekernelcladamw: seems you inadvertently flipped U6 in your latest schematics10:30
lekernelbe careful ...10:30
cladamwlekernel, ;-) no, i tried intendly to modify its frame but didn't in success, wanted to set its apparent feature, then later just change its frame color. ;-) yuh, I'm very carefully on U6. ;-)10:33
lekernelbut http://downloads.qi-hardware.com/hardware/milkymist_one/sch/tmp/MILKYMISTONE.pdf has messed up connections10:34
cladamwi didn't listed up all details on each page. but here i recorded them firstly: http://en.qi-hardware.com/wiki/Milkymist_One_Schematic_Change_History#R4_.28not_finished_yet.2910:36
lekernelhttp://milkymist.org/midiflip.png <= this is wrong, period10:39
cladamwlekernel, oah...yes, you got it. I need to mirror it again. tks.10:41
cladamwlekernel, updated.10:46
cladamw(crowbar) any further idea? Do we feel strongly enough about it to implement and test that crowbar?10:50
lekernelthe thyristor thing? it's not on this schematics, right?10:55
lekerneland no, I think we should not implement it10:55
cladamwyes, not added them now, at least needs to discuss and even if we really need it also need to test before add.10:57
lekernelwpwrak: can you post again that autodependency makefile you pointed to? I can't find it again in the irc log11:10
lekernelah, http://projects.qi-hardware.com/index.php/p/fped/source/tree/master/Makefile11:11
GitHub109[milkymist-ng] sbourdeauducq pushed 1 new commit to master: http://git.io/myJj2w11:15
GitHub109[milkymist-ng/master] Copy some software code from the original Milkymist SoC. - Sebastien Bourdeauducq11:15
Fallenoutime for lunch !11:21
GitHub126[milkymist-ng] sbourdeauducq pushed 1 new commit to master: http://git.io/p6CkRg11:31
GitHub126[milkymist-ng/master] software: dependencies the Werner way - Sebastien Bourdeauducq11:31
lekernelwolfspraul: http://commons.wikimedia.org/wiki/User:Evan-Amos/Everyday12:23
wpwraknice picture collection13:13
xiangfuwpwrak, Hi.13:24
xiangfuyou may saw my message. I got the HID_REPORT_DESCRIPTOR correct now. :) my plan is parse the REPORT_ID first. since it's very simple.13:24
xiangfuthen it can maybe most 2 interfaces Keyboard+Mouse working. :)13:25
dvdkwolfspraul: about the LM32 license.  Just looking at it.  Unfortunately I can't agree that it is 100% bsd-style open.  Maybe "95%" :).  Also it has a few points that make it incompatible with GPL:14:32
dvdk6. export control according to US law14:33
dvdkand 1.: forbids to charge recipients for distribution14:33
dvdkjust my $(1/50)14:34
whitequarkhah! 3..+inf days and I'll be a happy M1 owner15:10
whitequark(+inf) you never know for sure with russian post.15:11
qi-botThe Firmware build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/milkymist-firmware-20120203-2021/21:08
--- Sat Feb 4 201200:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!