06:53 <marmrt[m]> <JamesMunns[m]> "did anyone ever make a tool that..." <- https://github.com/s7rul/pico-test-bench Erik's thesis is on this. I don't remember the details, and can't find a copy of the thesis atm.

07:38 <Farooq> so now I've used a 1k ohm resister to analog input of my stm32. Do you have any idea to make the noise more drastic?

07:38 <Farooq> as in, bigger changes

07:39 <Farooq> the ADC is 12bit. I get values in range of 4050-4096. Currently to make the changes have high influence in the seed, I read from ADC 5 times and then multiply the numbers after subbing 4000

07:40 <Farooq> I'm interesting in hardware/electronic solutions to learn more on that port. Writing a code which makes these small changes or differences so big, is not much of a big deal for me. I already can code well. But I don't know electronics

07:40 <Farooq> * know electronics(I'm learning)

07:41 <Farooq> oh and I'm using these values to seed the RNG

07:42 <whitequark[cis]> ducks run a tesla coil nearby

07:42 <whitequark[cis]> that will generate enough noise alright

07:42 <Farooq> maybe if I use a resistor with lower wattage, the influence of heat would be bigger?

07:42 <Farooq> @whitequark:matrix.org wdym?

07:43 <whitequark[cis]> uh... arc discharge produces broadband RF noise that couples to basically all electronics around it

07:43 <whitequark[cis]> whatever noise figures you're seeing now, wherever they are, i guarantee you that if you start an arc discharge somewhere close they'll get an order of magnitude worse

07:44 <whitequark[cis]> this isn't a very serious suggestino

07:44 <whitequark[cis]> s/suggestino/suggestion/

07:45 <Ralph[m]> whitequark[cis]: i'd take a radiation source and measure that 😜

07:45 <Ralph[m]> (to be fair you can buy such entropy sources, but i don't think that they're particularly cheap, so this also isn't a very serious suggestion 😅)

07:45 <whitequark[cis]> you can get a radiation source for like ten dollars

07:45 <whitequark[cis]> search amazon for "negative ion"

07:45 <Farooq> ten dollars? too much!

07:46 <whitequark[cis]> a geiger counter will run you more than that, but you can approximate it with a CMOS camera that is looking at a black rectangle

07:46 <Farooq> hmm alternatively, I could use a photocell

07:47 <Ralph[m]> <JamesMunns[m]> "🍴" <- there was no need for that - once i had dirbaio's confirmation that `DelayNs` should be clone-able the maintainer was awesome and finished up his PR and merged & released it last night! now `adafruit-seesaw` can be used with stable rust (for most modules, anyway) and with `postcard-rpc` 🥳

07:47 <Ralph[m]> btw, dirbaio: would it make sense to update the `DelayNs` docs to state that it should support `Clone`? i don't feel confident to suggest that edit myself as i'm not familiar enough with the intricacies around the implementations

08:00 <JamesMunns[m]> <Farooq> "oh and I'm using these values to..." <- Usually stm32 parts have an internal temperature sensor as well as internal voltage monitor, these tend to have some noise in the low bits, one technique I've used is reading them many times (100-1000x), and feed all the readings into a hasher like ChaCha8, then use that output as the seed of a chacha8 rngs

08:02 <mkj[m]> another option, instead of sampling voltage, sample time. https://github.com/mkj/caprand

08:03 <mkj[m]> * time. https://github.com/mkj/caprand (that also uses what James said)

08:09 <JamesMunns[m]> Hmm, I thought someone had made a chacha8/20 hasher impl, but I can't find it. The idea is to just use a fixed initial seed, feed the data into it like you're encrypting it, then use the state after consuming all data as the seed to an rng. I can't find it offhand tho

08:11 <vollbrecht[m]> while a chaotic system is "not random" its nature makes it really hard to predict. It's also something nice to use as one source. Keyword is "Chua-Circuit" :D

08:20 <Farooq> <JamesMunns[m]> "Usually stm32 parts have an..." <- ty

08:21 <Farooq> I think for the first project I will just use the resistor

08:21 <Farooq> also I was supposed to make the code understandable even for people who don't know Rust

08:22 <Farooq> I would like to send it to my friend with extensive C experience in embedded. And ask him what does he think about Rust

08:22 <Farooq> Maybe he could become Rusty xD

08:26 <Alex[m]> Hi 👋🏻

08:26 <Alex[m]> What is the best maintained crate today with `no-std` std io traits?

08:26 <Alex[m]> Currently i'm trying both - `portable-io` and `no_std_io2`, but `no_std_io2` looks better because it's kinda transparent - there is feature-gated re-export of `std::io` traits if feature `"std"` is enabled, but there is unsolved issues like [that](https://github.com/wcampbell0x2a/no-std-io2/issues/5). So IDK 🤷🏻‍♂️

08:28 <JamesMunns[m]> https://docs.rs/embedded-io/latest/embedded_io/ is probably the most commonly used in the embedded space, it's mostly oriented around serial ports. There's also an `embedded-io-async` version.

10:34 <jason-kairos[m]> Is it possible to exclude a single function from LTO / optimizations?

10:35 <diondokter[m]> jason-kairos[m]: You can prevent inlining with:... (full message at <https://catircservices.org/_irc/v1/media/download/AX801MYOPGMCeg1kJ4fuydVUq1wbzk-3oiUBdZKqGLG7QNv4ZhVgjsJFwb9d1BpO6zy2iiAHnMley7Udc4dVW7C_8AAAAAAAAGNhdGlyY3NlcnZpY2VzLm9yZy9WUmdBc2ZHc0pVWUVwaUFTTkJRT3NoZGs>)

10:35 <diondokter[m]> Other than that, I don't think you can do much

10:50 <Alex[m]> <JamesMunns[m]> "https://docs.rs/embedded-io/..."; <- I see, thanks. I just need traits like Read, Write, BufRead…

11:13 <mkj[m]> n

15:23 <MitchellDScott[m> Does anyone have recommendations on a Blas/lapack crate for small/medium matrices. I’d like to use something like https://github.com/giaf/blasfeo but would rather not depend on importing a c library. I would like to target cortex-m mcus, but I don’t know enough about hardware for this to be a hard requirement.

15:45 <jason-kairos[m]> rust really needs the per function optimizations flags that are available in C / clang / GCC

15:46 <JamesMunns[m]> https://github.com/rust-lang/rust/issues/54882

15:46 <JamesMunns[m]> Currently, you can't generally set optimizations on a resolution smaller than crates.

21:26 <thejpster> evening all

21:27 <thejpster> kind of extraordinary this still works

21:44 <JamesMunns[m]> IRC?

21:44 <JamesMunns[m]> Nice :D

21:44 <thejpster> from a PA-RISC machine