Tuesday, 2019-06-04

*** tpb has joined #symbiflow00:00
*** futarisIRCcloud has quit IRC02:01
*** citypw has joined #symbiflow02:31
*** citypw has quit IRC02:50
*** Bertl is now known as Bertl_zZ03:05
*** futarisIRCcloud has joined #symbiflow03:56
*** proteusguy has quit IRC04:29
*** proteusguy has joined #symbiflow05:47
*** futarisIRCcloud has quit IRC07:05
*** OmniMancer has joined #symbiflow07:07
*** Xark has quit IRC08:32
*** Xark has joined #symbiflow08:35
*** kraiskil has joined #symbiflow08:43
*** citypw has joined #symbiflow08:50
*** proteusguy has quit IRC08:58
*** kraiskil has quit IRC08:59
*** proteusguy has joined #symbiflow08:59
*** citypw has quit IRC09:28
*** Bertl_zZ is now known as Bertl09:30
*** Maya-sama has joined #symbiflow11:25
*** Miyu has quit IRC11:28
*** kraiskil has joined #symbiflow11:30
*** hackkitten has joined #symbiflow11:30
*** Maya-sama has quit IRC11:34
*** kraiskil has quit IRC11:52
*** kraiskil has joined #symbiflow12:06
*** proteusguy has quit IRC12:38
*** bjorkintosh has joined #symbiflow12:39
*** kraiskil has quit IRC13:53
*** proteusguy has joined #symbiflow14:00
*** kraiskil has joined #symbiflow14:07
*** lopsided98 has quit IRC14:32
*** lopsided98 has joined #symbiflow14:34
*** Bertl is now known as Bertl_oO15:06
mithroacomodi: Any chance you have looked at https://github.com/SymbiFlow/symbiflow-arch-defs/pull/779 ?16:00
tpbTitle: xc7: Use the common_slice definition when LUTRAM used as a LUT. by mithro · Pull Request #779 · SymbiFlow/symbiflow-arch-defs · GitHub (at github.com)16:00
*** kraiskil has quit IRC16:01
sf-slack2<mkurc> Morning !16:03
mithroIs anyone awaiting reviews from me?16:06
*** kraiskil has joined #symbiflow16:20
*** OmniMancer has quit IRC16:37
*** Bertl_oO is now known as Bertl16:50
sf-slack2<acomodi> mithro: not yet in detail, was traveling today. I will get to it first thing tomorrow so that the PR can be merged ASAP16:58
mithrotmichalak: I opened https://github.com/SymbiFlow/prjxray/issues/867 to guide us towards figuring out what needs to be done in prjxray17:00
tpbTitle: Create a test design using LiteX, LiteEth and LiteDRAM and figure out what bits are still needed · Issue #867 · SymbiFlow/prjxray · GitHub (at github.com)17:00
sf-slack2<mkurc> @mithro @litghost. Some time ago I did looking what bits we are missing in the 7-series. These are tiles that have pips and we do not have segbits for them: HCLK_IOI3, CMT_TOP_R_LOWER_B, CMT_TOP_R_LOWER_T, CMT_TOP_R_UPPER_B, CMT_TOP_R_UPPER_T, CMT_FIFO_R, CMT_TOP_L_LOWER_B, CMT_TOP_L_LOWER_T, CMT_TOP_L_UPPER_B, CMT_TOP_L_UPPER_T, CMT_FIFO_L17:05
litghostmkurc: Maybe file an issue with details?17:07
mithromkurc: how did you do that?17:08
sf-slack2<mkurc> I manually looked up `ppips_*.db` files to see which tiles that I am not familiar with have non-pseudo pips. Then I looked into corresponding `segbits_*.db`files and checked whether there are bits for those pips.17:45
sf-slack2<mkurc> I focused mainly on the clock tree.17:46
*** kraiskil has quit IRC18:10
sf-slack2<mkurc> Now when I think of that, I can write a script which does the same to double check my findings.18:11
mithromkurc: But how did you figure out which bits are *needed* ?18:13
sf-slack2<mkurc> @mithro Well, each non-pseudo pip should have at least some (I guess)18:14
sf-slack2<mkurc> I identified which pips do not have bits18:15
mithromkurc: But if a design never uses these pips, then we don't care all that much?18:15
sf-slack2<mkurc> Oh, I didn't check which bits would be required for a particular design18:16
hackerfoolitghost: I sketched an idea about compressing and sampling the cost matrix for lookahead:18:16
* hackerfoo posted a file: Notes122.pdf (109KB) < http://sandbox.hackerfoo.com:8008/_matrix/media/v1/download/sandbox.hackerfoo.com/NLIgEXUhdecwhprjwwukdsVA >18:17
hackerfooLet me know if that makes sense, or you'd like me to write up something more detailed.18:18
hackerfooFrom my work on simulated annealing, most connections will be near the diagonal, but I didn't use a space filling curve, because I didn't care about compression.18:20
hackerfooSo this should focus the diagonal better.18:21
litghostThat link does not appear to work?18:27
hackerfoolitghost: Drive link: https://drive.google.com/file/d/1DqPwElYrXBJZn5D8f2IErOFD7xU8m0BK/view?usp=sharing18:28
tpbTitle: Notes122.pdf - Google Drive (at drive.google.com)18:28
litghostYa, 1-10 GB for 1 million nodes is still a loser18:29
litghost1 million nodes is a small FPGA18:29
hackerfooReally? Why?18:29
litghostBecause we need to scale to 10 - 100 million nodes18:29
litghostThe current a7 graph is 1/5 of a 50k part18:29
litghosteven in artix 7-series, there is s 300k part18:29
hackerfooOh. The width of the diagonal is likely much less than sqrt(N), but I don't know.18:30
litghostonce you get to uS+, you have 1 M parts18:30
litghostSo we are still working on toy sized graphs18:30
litghostWe have +1-2 order of magnitude to go to reach the big parts18:30
hackerfooSo this is the number of pins? Can pins be grouped together?18:31
litghostNodes, which can be either pin or interconnect18:31
hackerfooI would guess the lower bound to be O(N log^2 N), which could scale to 100 million. But that's probably too optimistic.18:34
hackerfooMaybe this coupled with some symmetry tricks could do it. At least it's a start.18:35
litghostTo be clear, N in this case is number of nodes?18:35
hackerfooWe could keep improving the compression without affecting the results, since its just a way to store the true data, although sampling all of that data would take forever, hence random sampling.18:38
hackerfooThere's probably good ways to accelerate that, though, since you shouldn't have to route every src -> dst individually.18:40
*** kraiskil has joined #symbiflow18:56
*** kraiskil has quit IRC20:43
*** kraiskil has joined #symbiflow20:55
*** kraiskil has quit IRC21:58
*** Bertl is now known as Bertl_zZ23:03
*** bjorkintosh has quit IRC23:20
*** futarisIRCcloud has joined #symbiflow23:45

Generated by irclog2html.py 2.13.1 by Marius Gedminas - find it at mg.pov.lt!