<larsc>
if you want to get drunk take a shot everytime the speaker says amazing
<mumptai>
well, maybe its soo AMAZING ..
<larsc>
the idea is good
<larsc>
but the world isn't ready yet ;)
<mumptai>
but partial reconfig, well
<mumptai>
the amount of layers that have to be juggled by the developers is a serious limit to the spread of these architectures
<larsc>
yes, the make it sound as if it is all figured out
<larsc>
but in the end acknowledge that there is nothing
<sb0>
"a very fast bus"
<sb0>
haha ;)
<mumptai>
huge fpga -> two 32bit CPUs
<sb0>
"bitcoini mining"
<mumptai>
bitcoins are such an poor example
<mumptai>
actually, what feedback do you people get with mm and related projects?
<mumptai>
and who are those people? cs&ee-students, professional? or how dedicated have the amateurs to be to handle these systems
<sb0>
professional. everyone else pretty much gives up after a half-assed attempt.
<mumptai>
i'm not surprised
<mumptai>
my old university robotics club just dropped the cortex-m+fpga architecture, as they just weren't able to handle it. and that system was already running and competition proved. too sad
<mumptai>
fpgas are so much more suited to those IO and timing-heavy task than mcus ..
<sb0>
we should get direct edif output from migen at some point
<sb0>
it's not that hard
<sb0>
the really difficult point here is figuring out the fpga routing
<sb0>
which xdl still gives you, even for 7-series, but in expanded form
<mumptai>
is there a real chance to get around the vendors? a path that will work decently?
<larsc>
make it a competetive advantage for them to open up their endcoding and mapping
<mumptai>
basically make them pay the developers supporting it
<mumptai>
well, i guess a language&toolchain with abstract system descriptions could be a part of that
<sb0>
bah, xilinx is open enough that someone who wants to tweak a bitstream a bit will
<sb0>
but sorting out all those details for writing a complete toolchain is a pita
<mumptai>
verilog & vhdl both have a kind of glass-cailing in abstractions, and this whole inference approach tends to be brittle
<sb0>
"MIT Media Lab, Lifelong Kindergarten Group"
<larsc>
your new empolyee?
<sb0>
"A 40kW argon laser-powered hot tub". now that's interesting. except that I have a better idea than warming up a hot tub
<sb0>
well, in that case it was the woman operating the laser :)
<mumptai>
okay, i take that as prove *g*
<larsc>
maybe it should be called SAF
<sb0>
+1
<mumptai>
okay, agreed
<sb0>
I wonder how contacting AMD to have the vuln fixed before constitutes "responsible" disclosure. all it does is prevent people from reprogramming the CPU they have bough.
<sb0>
*bought
<larsc>
or other people from reprogramming your cpu
<mumptai>
the fix actually introduces code running on your system, that you can
<sb0>
wouldn't they need root anyway?
<mumptai>
't inspect
<larsc>
yes, but this can be used to build a rootkit that you can't detect by ordinary means
<sb0>
who needs those. there are so much easier hacks you can use to commit crimes.
<larsc>
you want to monitor people
<sb0>
well the NSA does, and they can just serve a subpoena to AMD to get the key
<mumptai>
its just one more corner to hide stuff in, and it does have it's own private cpu
<sb0>
finally! got that goddamned gtk to work with asyncio
trem has joined #m-labs
trem is now known as Guest56132
Guest56132 is now known as brujah
brujah has left #m-labs [#m-labs]
kilae has joined #m-labs
sb0 has quit [Quit: Leaving]
Alain has quit [Quit: ChatZilla 0.9.91.1 [Firefox 34.0.5/20141126041045]]
kilae has quit [Quit: ChatZilla 0.9.91.1 [Firefox 34.0.5/20141126041045]]
mumptai_ has joined #m-labs
mumptai_ has quit [Remote host closed the connection]