XgF is now known as gay
<whitequark> furan-: it took quite a bit of work to get python to be able to keep up
<whitequark> especially with the amount of babysitting USB requires to work reasonably well
<emily> the core transfer code will be ported to rust for revE right?
<whitequark> on revE it'll just be TCP/IP on the device
<emily> er, right
<emily> well i remember you saying some rust was going to be involved at least
<whitequark> that would be for revABCD
<hl> ...is revE not using USB?
<whitequark> dual USB and Ethernet interfaces
<hl> nice
<whitequark> with USB presenting one of the two Ethernet encapsulations
<whitequark> it'll be IPv6-only of course
<hl> nice
<hl> wait are you doing IPv6 on an iCE40?
<whitequark> revE would use ECP5
<hl> aah
<hl> with a softcore?
<whitequark> no, IPv6 in gateware
<hl> huh, ok
<hl> are you going with cypress's USB3 FX stuff or sticking with USB2?
<whitequark> FX3, of course
<hl> ah
<hl> 1000BASE-T?
<whitequark> yep
<hl> ah
<whitequark> ideally with PoE or at least a PoE option
<hl> yeah, faster ethernet is a bitch for the ECP5 it seems. XGMII PHYs seem unobtanium nowadays, and XAUI eats up FPGA transceivers
<whitequark> hm, really? interesting
<whitequark> oh, wait, you mean 10GbE
<hl> I think there's ONE XGMII PHY still available that I was able to find by Microsemi (ehhh, Microsemi) but the datasheet seemed to imply it didn't also support 1000BASE-T, so *shrug*
<hl> yeah
<whitequark> it makes no sense to provide 10GBASE-T on the board
<hl> Ah, it doesn't?
<whitequark> it does make sense to provide a QSFP cage
<whitequark> 10GBASE-T sucks
<hl> oh right, you mean how 10GBASE-T sucks
<hl> yes, I know that
<whitequark> if I provide a QSFP cage I can just make an adapter to PCIe for example
<whitequark> many more uses
<hl> how would you do a QSFP cage though? Do you have enough transceivers on the ECP5?
<whitequark> there are 4 lanes
<whitequark> that can use the same clocking
<hl> hm
<sorear> pity there's no good way to actually do 5 Gbit/s with USB connection
<hl> sorear: ...srsly?
<whitequark> FX3 limits you to 3 Gbps
<emily> how does 10GBASE-T suck, ooc?
<hl> ouch
<whitequark> FT601 lets you do 5 Gbps but it would require *another* microcontroller on the board
<whitequark> and I think it's not such a significant perf hit
<whitequark> emily: mostly, it's that there are no reasons to pick 10G over twisted pair over either SFP direct attach or fiber
<hl> it has horribly high power consumption by comparison
<whitequark> yes, it
<whitequark> it's basically four 2.5 Gbps analog modems
<whitequark> or more like a 4 channel MIMO SDR would be a better way to describe it
<whitequark> in terms of power consumption at least
<hl> supposedly they made it worse too by trying to make it work on crappy cable
<whitequark> i think even 1 Gbps PHYs work a bit like SDRs on inside
<whitequark> they transmit and receive in both directions at once
<hl> i vaguely recall some proposal for a 10GBASE-TX which required cat6a or something that never went anywhere, but would have had less awful power consumption
<whitequark> so you have to do echo cancellation and stuff
<whitequark> which at those bit rates ends up very computationally intensive
<whitequark> if you look at a block diagram of a 1GBASE-T PHY you'll see a bunch of high speed ADCs
<hl> keep in mind that 10GBASE-T is so bad, you literally _cannot get_ 10GBASE-T SFP+s because the power consumption exceeds the thermal/power envelope of the SFP+ specification
<whitequark> lmao i was not aware of that
<hl> read: if you buy a switch, you literally have to choose between 10GBASE-T or SFP+ ports
<hl> ...there _are_ some 10GBASE-T SFP+s which aren't _supposed_ to exist and surely aren't officially supported by anyone, but which supposedly work by limiting the range to 30m instead of 100m, to reduce power consumption
<hl> (funfact: there are also VDSL2 SFPs which pretend to be ethernet. wtf)
<whitequark> lol
<whitequark> i mean GPON SFPs are weird, they terminate DHSP
<whitequark> *DHCP
<hl> yeah jesus christ
<hl> those things are necessarily "smart". wonder what chips are inside
<whitequark> i have a GPON SFP i won't use any time soon, i can carefully open it up
<hl> huh
<whitequark> GPON ONT SFP to be precise. the customer premises one
diverger has quit [Read error: Connection reset by peer]
diverger has joined #glasgow
kmehall_ has quit [Remote host closed the connection]
kmehall has joined #glasgow
ronyrus has joined #glasgow
<gruetzkopf> the O_L_T SFPs are quite normal
yuriks has quit [Read error: Connection reset by peer]
yuriks has joined #glasgow
Stormwind_mobile has quit [Read error: Connection reset by peer]
Stormwind_mobile has joined #glasgow
ronyrus has quit [Remote host closed the connection]
bvernoux has joined #glasgow
ronyrus has joined #glasgow
fibmod has quit [Ping timeout: 265 seconds]
fibmod has joined #glasgow
fibmod has quit [Ping timeout: 240 seconds]
Stephie has joined #glasgow
thaytan has quit [Ping timeout: 240 seconds]
thaytan has joined #glasgow
bvernoux has quit [Quit: Leaving]
Stormwind_mobile has quit [Ping timeout: 240 seconds]
Stormwind_mobile has joined #glasgow
<gruetzkopf> bad but somehow working idea of the evening: i'm stuck at my workplace due to trees on rail lines (there's really far worse places to get stuck than your own electronics lab). wanted to poke at cc2531, but the windows vm with IAR on it is running on a host at home
<gruetzkopf> SPICE usb forwarding over ssh to the rescue! works really well considereing 45ms rtt (which i could get down to 16 by not routing over vpn)
<gay> [03:43:03] <whitequark> i mean GPON SFPs are weird, they terminate DHCP <-- what the fuck
gay is now known as XgF
<jschievink> the P in DHCP stands for Plug, right?
<XgF> this makes cable modem fuckery look sensible
<whitequark> XgF: you specify the password via DHCP options
<whitequark> it's... weird
<XgF> ;_;
<XgF> pls just PPPoE or something moderately sensible
<XgF> i'm just. unreasonably upset about this. nobody expects a router in their sfp
<whitequark> lol
<whitequark> i'm not sure if it's a router exactly. i mean it might be. it definitely also terminates ethernet