01:11 <whitequark> binutils patch is in

04:07 <sb0> rjo, re. DRTIO with KC705 SERDESes. it will definitely work at ~500Mbps (using 2x the SERDES bits to implement a spartan-6-style phase detector). that may be enough, and on the satellite device, transceivers are spec'd to lock at such a low bitrate as well.

04:10 <sb0> the expensive solution is to use a "serial connectivity optimized" virtex that has lots of transceivers on the core board, then we can run each link at several Gbps

04:10 <sb0> but maybe the kintex7 can be hacked to provide phase detection at 1Gbps... I'll have to look at that

04:13 <sb0> or, without a phase detector, the link can be periodically recalibrated - but it will have downtime

04:20 <sb0> (with the synchronous serdes system, a recalibration is basically needed when the round trip time varies by more than ~0.5ns)

09:41 <GitHub186> [misoc] enjoy-digital pushed 1 new commit to master: http://git.io/v3ELf

09:41 <GitHub186> misoc/master 5a3f4e4 Florent Kermarrec: tools/flterm.py: some cleanup and fix last frame data that was not transmitted

10:34 <whitequark> hm, no, it's back out

18:25 <rjo> sb0: you are maximixing the number of p2p channels here?

18:25 <rjo> sb0: yes. that oversampling would work.

18:28 <rjo> but even 24 gt* transcievers on a kintex seems to be abundant. and would even be less code duplication between the master and the satellite.

18:31 <rjo> the serdes oversampling transcievers are a good extension. i am a bit uncertain about the low frequency corner of the electrical-to-optical things. they all seem 1.25 Gb/s with a pretty high low frequency corner (or else inter symbol interference).

18:33 <rjo> if we do the oversampling serdes on the satellite side, the problem is a very jittery time base (1ns jitter) because in the best case you can correct your local clock once you have to recalibrate/slip the phase. but that is jitter.