<Bob_Dole>
I think there's another hat like that that uses HX parts
<sorear>
in order to verify that the arm cores booted they would have had to have access to at least one piece of hardware
<sorear>
.oO( debug console over physical covert channel )
<SolraBizna>
the UART works
<sorear>
aren't there rpi alternatives that are less cursed
<Bob_Dole>
there's some that are more cursed
<SolraBizna>
yeah, but I wasn't gifted one for free by someone who didn't know better
<SolraBizna>
*one of the non-cursed alternatives
<Bob_Dole>
is the Efika MX still a totally non-functional device in the modern era?
<Bob_Dole>
because I got one still
<sorear>
do you _want_ one? could you accept one without fucky financial effects?
<SolraBizna>
I do and also don't
<Bob_Dole>
wat
<SolraBizna>
I accepted this R.Pi because I needed something to sit on the network, manage my backup array, be stable, and not consume gobs of electricity
<SolraBizna>
not consuming gobs of electricity is the only thing it's been pretty successful at, and that's in exchange for consuming gobs of my time instead
<rvense>
i've got one of those wandboards... been using it for mail and irc and other things for years
<SolraBizna>
for things requiring lots of GPIOs and a bit of processing power (e.g. the 65test), I have an Arduino Due
<Bob_Dole>
I've got 2 dues, and I think BOTH are knock-offs based off of early revision that had too weak of a resistor somewhere. I think manually resetting it was the usual fix?
<Bob_Dole>
like, every time you wanted to use it, needs reset button pushed.
<Bob_Dole>
I like that you actually know where to find a blog post about it. (I suspect it lists what resistor to replace too...)
<SolraBizna>
the 65test is immune to that problem because it uses the reset-from-USB as part of normal functioning
rohitksingh has joined ##openfpga
<SolraBizna>
I would love to finish 65test, but that Arduino is currently acting as a really enormous, expensive UART-to-USB adapter
<Bob_Dole>
should I mail you one of my spares and the 65c02?
ayjay_t has quit [Read error: Connection reset by peer]
ayjay_t has joined ##openfpga
<SolraBizna>
65test is still all wired up, it's just that it's currently also connected via mangled fan connector to that R.Pi
<Bob_Dole>
oh I thought your parents lost it for you.
<SolraBizna>
they lost the KCCU and its Arduino
<SolraBizna>
and my soldering station and my previous laptop and my mini desktop and ...
<SolraBizna>
oh, and a whole bucket-bag of PCL
<Bob_Dole>
that's expensive stuff :/
<SolraBizna>
I managed to rescue my LC III + accessories and my G5 (though not the G5's funky power cable)
<Bob_Dole>
funky power cable?
<SolraBizna>
some of the G5s had a nonstandard connector on their power supply
<Bob_Dole>
also does your LC3 still need re-capped?
<SolraBizna>
we recapped it, it runs like a dream (as much as any LC III can)
<SolraBizna>
that recapping project is the reason I know what hot tantalum dust smells like
<Bob_Dole>
I bet you could make some classic mac expansion cards as a Product. >.>
<SolraBizna>
I already made one, sort of... a serial-based external "hard disk"
<Bob_Dole>
level shifters. fpga. SATA controller.
<Bob_Dole>
and maybe modern ethernet.
<sorear>
nubus to pcie
<Bob_Dole>
I was just talking about that the other day
<Bob_Dole>
specifically thinking of his lc3
<kc8apf>
SolraBizna: those are still standard. Its a IEC C20
<SolraBizna>
good to know
<SolraBizna>
now that I think about it, I don't remember whether my G5 was one of those
<sorear>
anyway if you're talking about things that cost a day's living expenses I could probably be convinced
<Bob_Dole>
did macs ever have ramdisk expansion cards? I know Amigas did that, just stuffing a bunch of ram on a zorro card, and the OS would use it as slow-ram
* SolraBizna
adds "butter up sorear" to to-do list
<SolraBizna>
>_>
<Bob_Dole>
buttering up your filesystem turned out poorly
lovepon has quit [Ping timeout: 260 seconds]
<SolraBizna>
it only turns out poorly if I actually use it
<SolraBizna>
see? that time I was able to type a command line without the SSH session timing out, this is fine
m4ssi has quit [Remote host closed the connection]
m4ssi has joined ##openfpga
sunxi_fan has joined ##openfpga
m4ssi has quit [Remote host closed the connection]
kuldeep has quit [Read error: Connection reset by peer]
kuldeep has joined ##openfpga
Bike has joined ##openfpga
Bike is now known as Bicyclidine
<sorear>
1800-2012 and 1364-2005 do not make particularly clear how `define is supposed to be handled in edge cases
<sorear>
is there a practical application for high-efficiency ecp5 bitstream compression (i.e. better than what can be done in hardware, but still decompressable with a small amount of C code)?
<sensille>
sorear: if you have a small mcu on board that feeds the ecp5, it might even save an external flash
<sensille>
i was thinking of building just that
<sensille>
but how much better can you compress?
<daveshah>
Might be interesting for folknology's BlackEdge board using an stm32 for programming
<daveshah>
The USB 1.1 link will saturate before SPI
<sorear>
how high does the stm32 clock?
<sorear>
sensille: hard to say this early / without doing any research on the capabilities of the vendor compression
<sorear>
sensille: for the purposes of this analysis let's say half size. dunno if that's remotely achievable
<sorear>
actually hmm. does anyone have a reasonably large corpus of uncompressed bitstreams for various devices
azonenberg_work has quit [Ping timeout: 240 seconds]
<felix_>
balrog: the linux version of ise 14.7 works fine on windows 10 using the windows subsystem for linux. installed the debian variant, installed xorg in there and installed mobaxterm on the windows side and things just worked. well, i don't remember if it was ise where i had to patch the 32/64bit check out of the installationscript, but that was all i needed to patch for some of the xilinx software
<balrog>
lol...
<Flux42>
I tried that a while back. But ISE was just too painful to go back to.
<Bob_Dole>
I was wondering if it was a differentiation in application, if they were perhaps not power-efficient but fast(er) while bigger. Is there any indication Lattice might put out some bigger ECP5s?
<gruetzkopf>
i'd love a ECP5 with more SERDES blocks
<travis-ci>
whitequark/Glasgow#132 (master - e20ad9a : whitequark): The build has errored.
<Bob_Dole>
Oh. I was looking on Mouser, only a couple ecp3's seemed to have NRND status. (and was asking if they were similar under the hood for if adding support would end up being Easy.)
<Bob_Dole>
I was thinking of, I think one of the MAX cplds had an older variant using the same bitstream format?
<sorear>
yes
<sorear>
MAX II versus MAX V i think? rqou's bailiwick
rohitksingh has quit [Ping timeout: 240 seconds]
<sorear>
can probably reuse most of trellis to do ecp3 if you wanted to
<sorear>
my priorities are generally based on "how likely is someone to need this", which for a NRND chip is probably just s6 and similar due to the sheer number of s6 devboards random people have
<Bob_Dole>
yeah, I didn't realize they were all NRND because of mouser only labeling a few as it.
ZipCPU has quit [Ping timeout: 252 seconds]
<daveshah>
Yes, most of the bitstream parser and fuzzers could be reused. But you'd have to rerun the fuzzers and modify tile names as needed, etc
<daveshah>
But as I say 7 series is probably a better use of time than ecp3
<daveshah>
the main advantage of ECP5 over 7 series is cost
<daveshah>
Which the ecp3 doesn't really have
<sorear>
unclear how compelling the intel and microchip lines are
ZipCPU has joined ##openfpga
<daveshah>
Larger intel parts might be trouble because of the redundancy stuff and the stories about silicon bugs being worked around in their software. Don't know how true that is
<daveshah>
Microsemi is more interesting imo
<daveshah>
With their non volatile and low power stuff
<daveshah>
Also the cortex m3 SoCs
<sorear>
the microsemi stuff is true nonvolatile, which could be useful for some stuff (900 µs from POR to output drivers active according to the polarfire DS)
<Bob_Dole>
prior to learning what I have, I hoped they'd end up being faster, like older ice40s are faster than the new UPs, and they'd be a relatively-simple thing to add until 7 series can be done, as a stop-gap... but slower and more expensive. blah.
<sorear>
still surprisingly high, and might be in a tight race with the more optimized CPLDs
<sorear>
well this is more like comparing an ice40 to an ice65
<daveshah>
I do wonder if Lattice's next parts will have a hard RISC V core
<daveshah>
As they have joined RISC-V
<sorear>
I don't even have downloaded any Intel/Altera datasheets, they must be useful for something though
<daveshah>
Hard floating point is quite cool imo
<sorear>
does microsemi have designer programmable hard M3s on any parts?
<Bob_Dole>
I know Altera's Cyclones used to be popular around the time of Xilinx's 3 series Spartans being everywhere
<Bob_Dole>
and then they dropped off the map at some point and until the foss toolchain I was only seeing xilinx used for anything at all
<Bob_Dole>
excluding intel talking about putting them on cpus
<rqou>
cyclone arch is nice if you're obsessed with micro-optimizations
<daveshah>
sorear: SmartFusion
<rqou>
altera product naming is kinda a disaster though
<rqou>
as are family trees
<Bob_Dole>
the first time I heard anyone using altera again.. was that surprise move by arduino, using the intel parts, not just with a proprietary toolchain, but cloud-based proprietary.
<daveshah>
With things like the Cyclone 10 LP, Cyclone IV E and Cyclone III all sharing the same IDCODE iirc
<rqou>
yeah, they're allegedly bitstream compatible too
<rqou>
meanwhile cyclone 10 gx is a completely different arch
<daveshah>
Not timing compatible though
<daveshah>
At least III and IV E are different processes
<Bob_Dole>
I was kinda interested in the Intel stuff for a bit because of the OpenCL integration in their toolchains. Interested because of mining and the first discussion of FPGA-offloading of algo parts being able to make such a big improvement
<rqou>
oh yeah, timing is different
<rqou>
max ii/v are also bitstream compatible
<Bob_Dole>
speaking of the max ii/v, what state is the toolchain in for them, and.. what reasons might they be considered over the ice40s, for instance? >.>
<daveshah>
Non volatile is probably a big advantage
<Bob_Dole>
is it not OTP so can be reprogrammed later? because oh yes
<Bob_Dole>
that would be a huge advantage
<daveshah>
It's flash I believe unlike the ice40
<Bob_Dole>
if the toolchain is working, then SolraBizna may find some interest in that
<sorear>
my understanding is that MAX uses a flash separate from the SRAM config, but they're connected by a very wide bus so the load takes a few 100 cycles
<daveshah>
Yes, similar to MachXO I think
<daveshah>
Whereas ice40 NVCM loads at SPI type speeds
<sorear>
contrast ice40, where configuration from NVCM is bit-serial exactly as external config, and polarfire etc which don't have config SRAM (the flash transistors drive the control signals directly)
<prpplague>
azonenberg_work: softiron is who i work for
<prpplague>
azonenberg_work: we actually have "chain of custody" for parts
marshallh has joined ##openfpga
<azonenberg_work>
prpplague: verified how? say you have a xilinx fpga that was fabbed at tsmc, the wafer went to a wirebonding factory on the other side of taiwan
<azonenberg_work>
chinese intelligence comes in and intercepts the package, does a few fib cuts on the wafer, then forwards it
<azonenberg_work>
can you detect that?
m4ssi has joined ##openfpga
<marshallh>
wouldn't they be working at TSMC already
<prpplague>
azonenberg_work: yea, i can't got into details, but it was more to point out that we work and target specific customers who have "concerns" about the types of hardware attacks we have discussed
<azonenberg_work>
marshallh: that too, i was just pointing out an example
m4ssi has quit [Remote host closed the connection]
prpplague has quit [Ping timeout: 264 seconds]
prpplague has joined ##openfpga
xdeller has quit [Remote host closed the connection]
xdeller has joined ##openfpga
ayjay_t has quit [Ping timeout: 245 seconds]
ayjay_t has joined ##openfpga
wbraun has quit [Quit: wbraun]
azonenberg_work has quit [Ping timeout: 268 seconds]