*** tpb has joined #yosys | 00:00 | |
*** Cerpin has quit IRC | 00:19 | |
*** bitmapper has joined #yosys | 00:30 | |
*** N2TOH_ZZz is now known as N2TOH | 00:38 | |
*** rohitksingh has quit IRC | 01:15 | |
*** tlwoerner has quit IRC | 01:25 | |
*** bitmapper has quit IRC | 02:06 | |
*** Cerpin has joined #yosys | 02:33 | |
*** m_w has quit IRC | 03:15 | |
*** rohitksingh has joined #yosys | 03:42 | |
*** citypw has joined #yosys | 03:58 | |
*** rohitksingh has quit IRC | 04:03 | |
*** m_w has joined #yosys | 04:56 | |
*** Jybz has joined #yosys | 04:57 | |
*** unkraut has quit IRC | 05:08 | |
*** tlwoerner has joined #yosys | 05:18 | |
*** s_frit has quit IRC | 06:03 | |
*** s_frit has joined #yosys | 06:04 | |
*** az0re has joined #yosys | 06:15 | |
*** Jybz has quit IRC | 06:31 | |
*** citypw has quit IRC | 06:49 | |
*** citypw has joined #yosys | 06:55 | |
*** X-Scale` has joined #yosys | 07:34 | |
*** X-Scale has quit IRC | 07:37 | |
*** X-Scale` is now known as X-Scale | 07:37 | |
*** X-Scale` has joined #yosys | 07:45 | |
*** X-Scale has quit IRC | 07:47 | |
*** X-Scale` is now known as X-Scale | 07:47 | |
*** m4ssi has joined #yosys | 07:59 | |
*** dys has quit IRC | 08:07 | |
*** dh73 has joined #yosys | 08:32 | |
*** parataxis has quit IRC | 08:46 | |
*** dys has joined #yosys | 08:47 | |
*** fsasm has joined #yosys | 09:03 | |
*** captain_morgan has quit IRC | 09:29 | |
*** m4ssi has quit IRC | 09:33 | |
*** fsasm has quit IRC | 09:54 | |
*** m_w has quit IRC | 09:56 | |
*** az0re has quit IRC | 10:46 | |
*** vidbina_ has joined #yosys | 11:00 | |
*** vidbina_ has quit IRC | 11:06 | |
*** vidbina_ has joined #yosys | 11:08 | |
*** m4ssi has joined #yosys | 11:23 | |
*** Cerpin has quit IRC | 11:27 | |
*** Cerpin has joined #yosys | 11:45 | |
*** az0re has joined #yosys | 13:04 | |
*** vidbina_ has quit IRC | 14:10 | |
*** rohitksingh has joined #yosys | 14:26 | |
*** captain_morgan has joined #yosys | 14:32 | |
*** vidbina_ has joined #yosys | 14:34 | |
*** vidbina has joined #yosys | 15:03 | |
*** vidbina_ has quit IRC | 15:04 | |
*** rohitksingh has quit IRC | 15:14 | |
*** rohitksingh has joined #yosys | 15:15 | |
*** s_frit has quit IRC | 15:32 | |
*** s_frit has joined #yosys | 15:32 | |
*** rohitksingh has quit IRC | 15:34 | |
*** rohitksingh has joined #yosys | 15:38 | |
*** s_frit has quit IRC | 15:46 | |
*** s_frit has joined #yosys | 15:46 | |
lukego | Hey whaddayareckon about this ebay deal on salvaged Kintex UltraScale+ chips for ~ 1/10th of list price? https://www.ebay.com/itm/Xilinx-Kintex-XCKU15P-Fpga-On-Board/274145389372 (Maybe some prototype network card from Silicom or something?) | 16:09 |
---|---|---|
*** emeb has joined #yosys | 16:15 | |
ZirconiumX | xcup+ would require RapidWright, so it wouldn't be a completely FOSS flow. If that doesn't bother you, then sure. | 16:16 |
ZirconiumX | Although consider that list price figures are generally majorly inflated compared to business prices | 16:17 |
daveshah | Due to cross-die placement and routing issues, anything other than Zynq UltraScale+ and perhaps the smallest Kintex if it's single die isn't supported by nextpnr anyway atm | 16:17 |
ZirconiumX | Okay, fair | 16:17 |
lukego | ZirconiumX: Sure but hackers don't have access to better than list prices generally right? everybody else does but not us... | 16:18 |
ZirconiumX | Hackers don't, primarily due to economies of scale | 16:18 |
lukego | OK. So yeah it's not a very exciting card from a toolchain perspective, but still exciting to me to see an affordable hardware for 100G ethernet dev. I'm planning to work with ECP5/10G but if I wanted to do anything commercial I'd probably need to use 100G and it's nice to see a potential path | 16:19 |
daveshah | You'd also have your work cut out doing pinout RE, etc, with a board like that | 16:20 |
lukego | Yeah. Maybe make a new board and move the chip across. | 16:20 |
lukego | Guess it's not anything I need right now but still nice to know that if I wanted to prototype a 100G NIC it's possible I could pick up the FPGA for hundreds rather than thousands of dollar.s | 16:21 |
ZirconiumX | You can also source old Stratix chips for not too much | 16:22 |
ZirconiumX | Sadly though the toolchains for them still cost a bomb | 16:24 |
daveshah | Oh yeah, that's the other problem with XCKU15P, you still need a Vivado license for official bitgen with RapidWright | 16:25 |
daveshah | although you can pay by the hour on AWS, at least in theory | 16:25 |
lukego | vivado license expensive and/or difficult? | 16:25 |
daveshah | expensive ($3000 ish iirc) | 16:26 |
lukego | ah, fuck that | 16:26 |
ZirconiumX | Quartus is about that too | 16:26 |
ZirconiumX | I feel like there's a razor and blades model going on here | 16:27 |
ZirconiumX | Though the razor is still really expensive | 16:28 |
*** citypw has quit IRC | 16:28 | |
daveshah | Definitely not, just another way of reducing the support burden from people they don't care about | 16:28 |
daveshah | Vivado license would be free for anyone important enough | 16:28 |
lukego | The model I'd like to see is "cheap hardware that poor hackers can use to build expensive stuff for rich companies to buy." Like if I could spend $350 on an UltraScale+ and use that to write some gateware that a stock exchange spends a million dollars on, etc. | 16:28 |
lukego | seems like the vendors encourage this model but do it via universities. | 16:29 |
ZirconiumX | It's a crappy situation. | 16:30 |
lukego | Could even be a positive that the production hardware/software is expensive if that increases the perceived value of products based on it. But chicken egg problem if the prototyping gear is too expensive for hackers to take a chance on. | 16:31 |
lukego | Thankfully this is a second-order problem and for now I just need to do something with ECP5 using dave &co's fantastic toolchain :) | 16:31 |
ZirconiumX | Actually, daveshah, how big does Crosslink go? | 16:32 |
lukego | I'm n00bing around a little bit. I ended up blasting my ECP5 with hot air at 500C for a couple of minutes. Curious to see if it works. | 16:32 |
daveshah | CrossLink-NX? 40k magic fairies, 32k LUTs | 16:32 |
ZirconiumX | So they're building a... Smaller chip than the ECP5? | 16:33 |
daveshah | Yes, because it's not the ECP5 successor | 16:33 |
daveshah | I understand there are ECP5 and iCE40 successors both to be released this year | 16:34 |
ZirconiumX | Ah, okay | 16:34 |
ZirconiumX | I suppose being 28nm means faster interconnect and/or lower power? | 16:37 |
ZirconiumX | I'm really not up to date on these things | 16:37 |
daveshah | Yes, a bit of both, particularly being FDSOI | 16:38 |
*** rohitksingh has quit IRC | 16:41 | |
somlo | daveshah: https://pastebin.com/q52CqYnU (for nextpnr vs. python 3.9) | 16:43 |
tpb | Title: [Diff] diff --git a/CMakeLists.txt b/CMakeLists.txt index 8131d4a..92b5d18 100644 --- - Pastebin.com (at pastebin.com) | 16:43 |
*** citypw has joined #yosys | 16:44 | |
daveshah | somlo: done | 17:13 |
daveshah | thanks | 17:13 |
*** m4ssi has quit IRC | 17:27 | |
*** fsasm has joined #yosys | 17:37 | |
*** emeb has left #yosys | 17:39 | |
*** vidbina has quit IRC | 17:42 | |
*** rohitksingh has joined #yosys | 17:49 | |
*** dys has quit IRC | 18:03 | |
*** Jybz has joined #yosys | 18:43 | |
*** rohitksingh has quit IRC | 18:47 | |
*** rohitksingh has joined #yosys | 18:55 | |
awygle | > consider that list price figures are generally majorly inflated compared to business prices | 18:57 |
awygle | people say this all the time but i have not actually figured out how to actualize it | 18:58 |
awygle | it doesn't look like the procurement people at the company i work for are doing much that i wouldn't do myself | 18:58 |
awygle | maybe we're not big enough | 18:58 |
daveshah | Amazon certainly have - unless they are making a massive loss on EC2 F1 | 18:58 |
daveshah | I've heard of another large customer (name confidential) getting about a 5x discount buying 1 million+ iCE40s for a consumer product | 18:59 |
daveshah | I think the quantities to trigger it are higher than most people expect when they hear about them | 18:59 |
awygle | must be | 18:59 |
awygle | i'm currently working on a project at an awkward scale, where i can't get enough from e.g. Digikey's stock, but also Lattice and Arrow won't take my calls :( | 19:00 |
daveshah | Avnet were happy to quote for those kind of quanities for Xilinx in the UK (pricing wasn't great though) | 19:00 |
daveshah | don't know if they do Lattice too | 19:00 |
awygle | does Avnet distri-yes that | 19:01 |
awygle | they do not | 19:01 |
daveshah | Surprised that Arrow won't, they seem about the equivalnet | 19:01 |
awygle | i am gonna try again with Arrow, maybe if i get somebody on the phone they'll understand that i have a) budget b) schedule to absorb lead time | 19:01 |
daveshah | or DigiKey/Mouser RFQ? | 19:02 |
awygle | they were just like "yeah we don't have that many, so no-quote, talk to this company we own that calls Mouser and Digikey for you" | 19:02 |
*** dys has joined #yosys | 19:10 | |
*** rohitksingh has quit IRC | 19:25 | |
*** rohitksingh has joined #yosys | 19:25 | |
*** rohitksingh has quit IRC | 19:42 | |
*** rohitksingh has joined #yosys | 19:48 | |
*** Seraxis has joined #yosys | 19:53 | |
*** rohitksingh has quit IRC | 20:06 | |
*** rohitksingh has joined #yosys | 20:07 | |
*** az0re has quit IRC | 20:09 | |
thasti | is it possible to access cell/module attributes using pyosys currently? | 20:16 |
*** rektide has quit IRC | 20:31 | |
*** rektide has joined #yosys | 20:31 | |
*** rohitksingh has quit IRC | 20:35 | |
*** bitmapper has joined #yosys | 20:55 | |
*** rohitksingh has joined #yosys | 21:04 | |
*** Seraxis has quit IRC | 21:05 | |
*** seraxis has joined #yosys | 21:06 | |
*** fsasm has quit IRC | 21:14 | |
*** seraxis has quit IRC | 21:22 | |
*** seraxis has joined #yosys | 21:22 | |
*** seraxis has quit IRC | 21:24 | |
*** seraxis has joined #yosys | 21:25 | |
*** Jybz has quit IRC | 21:27 | |
*** dh73 has quit IRC | 21:43 | |
*** rohitksingh has quit IRC | 22:15 | |
*** Forty-Bot has quit IRC | 22:24 | |
*** rohitksingh has joined #yosys | 22:25 | |
*** Forty-Bot has joined #yosys | 22:28 | |
lukego | So I'm curious about Xilinx software license costs now... if you pay $3k for vivado, will they also nickle-and-dime you on basic features like e.g. SERDES? and I guess there's a substantial risk that you need to buy it again a couple of years later if there's some killer new feature e.g. rapidwright? | 22:30 |
lukego | also, can't run rapidwright on AWS for the ultrascale chips, because the AWS toolchain is device-locked on that crazy $50k virtex that they use? | 22:31 |
*** rohitksingh has quit IRC | 22:54 | |
*** bitmapper has quit IRC | 23:31 | |
*** bitmapper has joined #yosys | 23:33 |
Generated by irclog2html.py 2.13.1 by Marius Gedminas - find it at mg.pov.lt!