you can't do PCIe over USB because the former needs a hard worst-case latency guarantee and the latter explicitly avoids one
-
-
Replying to @whitequark @hikari_no_yume
Wait, PCIe has latency guarantees? Didn't
@marcan42 tunnel PCIe over some crazy-slow setup for PS4 stuff and have it work fine?2 replies 0 retweets 1 like -
I believe yes, see PCIe Base Specification 3.5.2.1 and Table 3-4
1 reply 0 retweets 0 likes -
Replying to @whitequark @11rcombs and
I'm sure you could work around that if you can control the entire system, like by gating the TLP Transmitter's clock
1 reply 0 retweets 1 like -
ahahaha
1 reply 0 retweets 0 likes -
or maaaaybe by terminating PCIe on either end and then using flow control to effectively stall the transmitter? no idea about PS4
1 reply 0 retweets 0 likes -
You do not tunnel the data link layer. Those timeouts are for TLP transmission acks, not TLP to completion (which is what we care about).
1 reply 0 retweets 2 likes -
Replying to @marcan42 @whitequark and
For comparison, USB has tight reply ACK timing requirements too (but nobody who isn't writing a USB device controller has to deal with that)
2 replies 0 retweets 2 likes -
Replying to @marcan42 @whitequark and
Really tight! Incidentally, this is part of where the cable length limits come in. You're running into light speed (in copper) delays!
1 reply 0 retweets 2 likes -
Replying to @erincandescent @oshepherd and
Yup. I once ran into this due to @xmos's crappy half-software UDC. Its API has critical undocumented timing constraints.
1 reply 0 retweets 1 like
Basically it'll stop NAKing packets if you delay between actually sending a reply and polling for the next transaction.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.