<openfpga-github>
[openfpga] azonenberg pushed 19 new commits to master: https://git.io/vdNKt
<openfpga-github>
openfpga/master b58ba64 Robert Ou: hidapi-linux: Delete numbered report logic...
<openfpga-github>
openfpga/master 0dc1be1 Robert Ou: hidapi-linux: Delete kernel version check...
<openfpga-github>
openfpga/master 66ff720 Robert Ou: hidapi-linux: Remove a bunch of parsing we do not use...
<azonenberg>
Happy? :p
<azonenberg>
Seems to work
<rqou>
nice
* azonenberg
goes back to improving his eye analysis code
<rqou>
how much did you test more-systemd-infected distros
<azonenberg>
i tested on debian
<azonenberg>
if i get reports of it breaking elsewhere, we'll deal with it then
<azonenberg>
:p
<rqou>
lol ok :P
<azonenberg>
anyway i'm going to have quite the nice SI analysis toolkit by the end of this project
<cyrozap>
Hey look I made a thing that's somewhat relevant to this channel (but unfortunately not PSoC related) that I've been working on for the past few weeks: https://github.com/cyrozap/mediatek-lte-baseband-re
<cyrozap>
Part of it involves reverse-engineering a publicly-unknown DSP ISA.
<cyrozap>
Which I've only just barely scratched the surface of.
Hootch has joined ##openfpga
nrossi has joined ##openfpga
digshadow has quit [Ping timeout: 240 seconds]
eduardo_ has joined ##openfpga
eduardo has quit [Ping timeout: 260 seconds]
<openfpga-github>
[openfpga] rqou opened issue #116: build system somehow finds yosys not on path, and then fails to run it https://git.io/vdN1X
<rqou>
which is a bit suboptimal if you have lots of cores
<rqou>
hmm, actually i still have a problem despite that
<rqou>
something i changed is now hitting a very RAM-intensive code path
<rqou>
even -j1 ooms my machine
<whitequark>
rqou: LTO?
<whitequark>
maybe?
<rqou>
i've always had LTO enabled
<rqou>
(for release builds)
<whitequark>
well yeah
<whitequark>
LTO can be very hungry
<whitequark>
ThinLTO will fix that but it has not shipped yet
<rqou>
but i used to be able to build this code with LTO
<whitequark>
re cmake: use --verbose and then grep for that path
<rqou>
without OOMing
<whitequark>
sure, rust changes what it feeds to LLVM occasionally
<whitequark>
you happened to be not lucky
<rqou>
no, i downgraded my rust install, still no good
<rqou>
hmm, i actually have a thought
<rqou>
nope
<rqou>
i downgraded to the "no external deps" revision and it still ooms
<rqou>
wtf
<rqou>
i wonder if i should "fix" this problem by paying a visit to Amazon.com :P
Bike has joined ##openfpga
<openfpga-github>
[openfpga] rqou commented on issue #116: `--verbose` did not give any extra information, but this issue only happens if I try to cross compile using `cmake --verbose .. -DBUILD_RUST=off -DBUILD_DOC=off -DCMAKE_INSTALL_PREFIX=/opt/fpgatools -DCMAKE_C_COMPILER=x86_64-linux-musl-gcc -DCMAKE_CXX_COMPILER=x86_64-linux-musl-g++ -DCMAKE_SYSTEM_NAME=Linux -DCMAKE_C_FLAGS="--sysroot=/home/rqou/code/fpgatools-build-system/tools-x86_64/sysroot" -DC
<openfpga-github>
[openfpga] rqou commented on issue #116: `--verbose` did not give any extra information, but this issue only happens if I try to cross compile using `cmake --verbose .. -DBUILD_RUST=off -DBUILD_DOC=off -DCMAKE_INSTALL_PREFIX=/opt/fpgatools -DCMAKE_C_COMPILER=x86_64-linux-musl-gcc -DCMAKE_CXX_COMPILER=x86_64-linux-musl-g++ -DCMAKE_SYSTEM_NAME=Linux -DCMAKE_C_FLAGS="--sysroot=/home/rqou/code/fpgatools-build-system/tools-x86_64/sysroot" -DC
<azonenberg>
rqou: if you have lots of cores and not a lot of ram
<azonenberg>
buy more ram :p
<azonenberg>
whats the point of having lots of cpu if you cant actually keep them fed?
<cr1901_modern>
azonenberg: B/c plenty of software is still written in C which doesn't take 1GB per compiler instance to generate an object file :P
<cr1901_modern>
Well actually, some C++ doesn't take that much either. Def a bit telling that my most powerful machine can build gcc in 3 minutes but takes 20 minutes to build LLVM (-j8)
nrossi has quit [Quit: Connection closed for inactivity]
<azonenberg>
cr1901_modern: i meant in general
<azonenberg>
i typically try to allocate 4G per CPU minimum, maybe more if i can
<cr1901_modern>
I'd need 32GB then
<azonenberg>
That's what i have
<azonenberg>
and i get things oomkilled and/or VMs failign to start about once or twice a week
<azonenberg>
My next workstation is going to be 128GB of RAM and 20 or 24 cores (dual xeons) which is 6.4 or 5.3 GB per core respectively
<zino>
My compile and experiment machine has 9GiB per core which is a bit on the lower end, but the motherboard doesn't take more than 144GiB RAM
<cr1901_modern>
Yea I had access to one of those until about 2 years ago. I don't have $5000+ to blow though. Maybe I could pay someone for a timeslice tho if I needed it
<cr1901_modern>
How difficult is it to compile a binary/library on one Linux distro that runs on another?
gnufan1 has joined ##openfpga
gnufan has quit [Ping timeout: 248 seconds]
<azonenberg>
cr1901_modern: i had about a $3K budget for my planned upgrade iirc
<azonenberg>
i forget, i've been shoving updated hardware into a newegg wishlist every few months when i get bored
<azonenberg>
and havent yet made the plunge
<jn__>
cr1901_modern: if the glibcs match up correctly, or you statically link against musl, it should work
<awygle>
I am fairly sure my current desktop mobo won't go over 16GB
<azonenberg>
awygle: mine goes to 32 but thats generally a bit of a limit
<azonenberg>
much beyond there you're looking at server boards
<azonenberg>
But i wanted to go dual CPU anyway and that means a server board
<awygle>
But I spent less than 500 on the whole thing so *shrug*
<azonenberg>
LOL
<azonenberg>
achievement unlocked
<azonenberg>
segfaulted my 100base-TX protocol decoder
<azonenberg>
with malformed PHY-layer traffic
<awygle>
Lol
<awygle>
Shoulda used rust :-P
<azonenberg>
Rust would crash/abort on an out-of-bounds read too, no?
<azonenberg>
just not corrupt everything on a bad write
<awygle>
uh iirc rust would panic! on an array-out-of-bounds access rather than returning garbage
<awygle>
depending on the actual problem rust may not have let you write the code at all
<awygle>
i was 80% meme-ing with the suggestion though, rqou or whitequark can probably actually defend it a lot better than me, a rust dilettante
<azonenberg>
This is an asan abort
<azonenberg>
so close enough :p
pointfree1 has quit [Ping timeout: 240 seconds]
digshadow has joined ##openfpga
pointfree1 has joined ##openfpga
<rqou>
azonenberg: most of the time i _do_ have enough ram
<rqou>
4GB per core would be 80GB, which is kinda expensive
<openfpga-github>
[openfpga] rqou commented on issue #116: I updated my cmake command based on that, and it still falsely detects yosys (command is `cmake --verbose .. -DBUILD_RUST=off -DBUILD_DOC=off -DCMAKE_INSTALL_PREFIX=/opt/fpgatools -DCMAKE_C_COMPILER=x86_64-linux-musl-gcc -DCMAKE_CXX_COMPILER=x86_64-linux-musl-g++ -DCMAKE_SYSTEM_NAME=Linux -DCMAKE_SYSROOT="/home/rqou/code/fpgatools-build-system/tools_x86_64/sysroot" -DCMAKE_EXE_LINKER_FLAGS="-stati
<openfpga-github>
[openfpga] rqou commented on issue #116: I updated my cmake command based on that, and it still falsely detects yosys (command is `cmake --verbose .. -DBUILD_RUST=off -DBUILD_DOC=off -DCMAKE_INSTALL_PREFIX=/opt/fpgatools -DCMAKE_C_COMPILER=x86_64-linux-musl-gcc -DCMAKE_CXX_COMPILER=x86_64-linux-musl-g++ -DCMAKE_SYSTEM_NAME=Linux -DCMAKE_SYSROOT="/home/rqou/code/fpgatools-build-system/tools_x86_64/sysroot" -DCMAKE_EXE_LINKER_FLAGS="-stati
<zino>
cr1901_modern: The general rule is that glibc compilations are forward compatible, but things compiled on a newer glibc is not guaranteed to link with an older one.
<zino>
I recommend VMs, or docker/singularity to have the complete OS you want to compile against instead.
<rqou>
or just use musl? :P
<azonenberg>
rqou: you have 20 cores?
<azonenberg>
and how much ram?
<rqou>
20 threads (10 cores)
<rqou>
and 32GB
<azonenberg>
oh
<azonenberg>
thats a nice xeon or something?
<azonenberg>
i'd want 40GB minimum and 80 ideally for that much
<rqou>
it's a xeon on a consumer-grade mobo :P
<azonenberg>
There is no reason to have that much CPU without enough ram to keep it fed
<rqou>
there's plenty of ram most of the time
<rqou>
meanwhile my NAS machine has a hilarious opposite setup - Pentium-branded cpu on a server mobo
<rqou>
anyways, I'm still trying to bisect why my code used to be compilable but now ooms every time
<nmz787>
i have a 22 core xeon
<nmz787>
with consumer grade X99 mobo
<nmz787>
2 sticks of 32GB ECC RAM
<nmz787>
50M cache :O
<nmz787>
I had been running it in my livingroom until my wife got annoyed by the whine of the fans a month or so ago
<nmz787>
I should try moving it to the closet, I had been trying to train the deepspeech neural network
<nmz787>
but also wanted to do some SAT/SMT solving on it
<zino>
I have a box with 3TB ECC RAM that was saved from being thrown away. Unfortunately it's a cardboard moving box, and it's 4GiB sticks...
<jn__>
that's a LOT of memory sticks
<zino>
Yea. I was decomissioning and old cluster a year ago.
<Zorix>
ebay that stuff
<azonenberg>
nmz787: don't you have a FIB in the basement?
<azonenberg>
right now forced 100M mode, no autonegotiation or 10M mode
<azonenberg>
and no RX either, it only TX's
<azonenberg>
i'm testing in loopback and havent brought up a link with any other hardware yet
<azonenberg>
Plan is to implement a MII bus
<azonenberg>
then use this as an excuse to add 10/100 support to my old TriModeEthernetMAC core that is not really tri-mode :p
<nmz787>
azonenberg: garage, yes. SEM in the den.
<azonenberg>
once i have it sending frames with valid CRCs to the broadcast address i'll see if i can talk to my laptop and wireshark a frame
<nmz787>
looking into buying an atomic layer deposition ssytem, too
* azonenberg
wants to build a sputtering rig after moving
<azonenberg>
Buying a house in a few months
<azonenberg>
First step will be insulation, climate control, and dehumidification in the garage
<azonenberg>
Targeting a 2-car sized space
<azonenberg>
as most houses here seem to be equipped with one
<azonenberg>
Plan is to run overhead cable trays in a figure-8 pattern around the perimeter of the garage and then down the middle
<azonenberg>
benches and storage around the perimeter of the lab then more benches in an "island" down the middle
<azonenberg>
overhead trays will have 120VAC, cat5e, and multimode fiber
<azonenberg>
Along the walls, and possibly routed to the central benches as well (TBD): hot and cold water, waste water return, low pressure (~40 PSI) nitrogen, and suction for dust/particle collection from drilling/machining operations
<azonenberg>
The garage will not be a cleanroom or anything, although i'll try and keep any kind of "shop" activities on the far end of the room from the microscopes
<azonenberg>
any fun thin-film stuff i build will be inside a glovebox with a separate HEPA-filtered supply
<rqou>
so azonenberg, how are we going to approach homecmos next? RIE or ghetto HF?
<rqou>
how do you even go about doing a DIY RIE setup?
<rqou>
hrm, we should start with a not-shit hotplate stirrer :P
dingbat has quit [Quit: Connection closed for inactivity]
<azonenberg>
I was planning to do DIY RIE actually
<azonenberg>
start with sputtering, then make a very similar tool and configure as a parallel plate RIE
<nmz787>
azonenberg: yeah I just have a dessicant dehumidifier... I need to pipe it out soon, had been dumping containers last year, and just plugged it in yesterday with all the rain we've been getting the last few days
<nmz787>
I have walls with drywall in the garage, no insulation I believe
<nmz787>
was planning to at least build a wall around the FIB to close it off from the rest of the garage
<nmz787>
insulation and maybe even a HEPA in the ceiling would be nice, for positive pressure
<nmz787>
but... hard to do with the FIB in the garage now... and the rain having started (and no where else the FIB will really fit)
<nmz787>
I'd like to have an RIE or DRIE :O
<openfpga-github>
[openfpga] rqou opened issue #117: this commit makes rust release builds suddenly take 20+GB https://git.io/vdA8g
<rqou>
finally bisected, but no idea wtf is happening
gnufan1 has quit [Quit: Leaving.]
enriq has joined ##openfpga
teepee has quit [Ping timeout: 252 seconds]
teepee has joined ##openfpga
<azonenberg>
nmz787: yeah i have a bucket based one
<azonenberg>
when i move i'm goign to add a permanent sewer i can drain into
<nmz787>
at this point I'm thinking just tube jammed through the corner of the garage door weatherstripping