02:10 <re_irc> <Julia> I am working on an extremely memory constrained application and I was wondering if it is possible to deallocate parts of a slice or vec. It seems I can truncate a vector for instance, but that does not actually reduce the memory allocated to it. Any ideas?

02:12 <re_irc> < (@jamesmunns:beeper.com)> Not generally - arrays are stored on the stack, and you can't generally "shrink" a stack frame without returning

02:12 <re_irc> < (@jamesmunns:beeper.com)> if you are using the standard library, you can shrink a vec with something like "shrink_to_capacity"

02:13 <re_irc> < (@jamesmunns:beeper.com)> but if you are not using the standard library (or some other heap allocator), it's not generally possible to shrink a static array or stack frame.

02:13 <re_irc> <Julia> I am using a heap allocator, but it is no-std.

02:14 <re_irc> < (@jamesmunns:beeper.com)> Ah! Then yeah, if you are using standard alloc types, https://doc.rust-lang.org/stable/std/vec/struct.Vec.html#method.shrink_to or https://doc.rust-lang.org/stable/std/vec/struct.Vec.html#method.shrink_to_fit are probably what you want.

02:16 <re_irc> < (@jamesmunns:beeper.com)> I think you need to drop the items off the tail first, then call truncate to free the space. Depending on your allocator (?), this may end up doing another allocation. I don't think Rust's alloc guarantees allocation resizings are done in place.

02:17 <re_irc> < (@jamesmunns:beeper.com)> e.g. if you have a vec with capacity 32, and len 8, and you shrink it to 8, it will probably (I think?) allocate a second, capacity 8 vector, copy the data there, then free the original vector

02:33 <re_irc> < (@jamesmunns:beeper.com)> followed up privately, but tl;dr:

02:33 <re_irc> - The Allocator trait has "resize", and GlobalAlloc has "realloc", which _could_ do resizing in place (vec calls realloc on "shrink_to*"), BUT

02:33 <re_irc> - CortexMHeap, and the underlying linked-list-allocator don't support realloc, which means realloc gets the default behavior of "make new alloc, copy to dest, free old alloc"

02:33 <re_irc> < (@jamesmunns:beeper.com)> which means the heap usage is going to get bigger before it gets smaller.

02:36 <re_irc> < (@yatekii:matrix.org)> hmm how do I best make a linkerscript in a library customizable? :/

02:36 <re_irc> < (@jamesmunns:beeper.com)> honestly? "build.rs" probably.

02:37 <re_irc> < (@yatekii:matrix.org)> hmm oki :)

02:37 <re_irc> < (@jamesmunns:beeper.com)> generate a linker fragment like "mylib.x", ask users to include "-Tmylib.x" or whatever in their ".cargo/config.toml"

02:38 <re_irc> < (@yatekii:matrix.org)> hmm do you have an example for that? :/ linkerscript is total blackmagic to me ...

02:38 <re_irc> < (@jamesmunns:beeper.com)> I think defmt does something like this?

02:38 <re_irc> < (@yatekii:matrix.org)> I know that cortex-m does it like this somehow

02:38 <re_irc> https://github.com/knurling-rs/defmt/blob/main/defmt/build.rs

02:39 <re_irc> < (@yatekii:matrix.org)> hmmm

02:39 <re_irc> < (@yatekii:matrix.org)> thanks!

02:40 <re_irc> < (@jamesmunns:beeper.com)> that's not doing any modification or template filling or anything, but it's doing the read/write/add "rustc-link-search" thing you will need to do from your lib

02:40 <re_irc> < (@jamesmunns:beeper.com)> you'd mutate the contents of the "linker_script" var for your use case.

02:42 <re_irc> <Julia> I managed to find a very _ugly_ workaround for my problem.

02:42 <re_irc> <Julia> let mut output = vec![0; 27000];

02:42 <re_irc> SliceReader::new(include_bytes!("../inputs/d01c.txt")),

02:42 <re_irc> let result = MyLzss::decompress(

02:42 <re_irc> SliceWriter::new(&mut output),

02:42 <re_irc> ... long message truncated: https://psion.agg.io/_matrix/media/r0/download/psion.agg.io/IH7lbMbr7CpDDyHSOlFE4YplampTWnqf (12 lines)

02:43 <re_irc> < (@jamesmunns:beeper.com)> :D

02:43 <re_irc> <Julia> Read in 27kb, check what the actual size was, drop it, reallocate it, read again.

02:43 <re_irc> < (@jamesmunns:beeper.com)> you could make an "empty writer" to avoid the first heap allocation

02:44 <re_irc> < (@jamesmunns:beeper.com)> I do that same trick in postcard: https://github.com/jamesmunns/postcard/blob/56d0a58bcfebe8e80db9920410ea6036c3cab32f/src/ser/serializer.rs#L345-L362

02:44 <re_irc> < (@jamesmunns:beeper.com)> still just as wasteful, CPU wise, but at least you'll save yourself the first allocation?

02:45 <re_irc> <Julia> I don't know if that'll work. I'm not really directly reading here. I am actually decompressing :P.

02:46 <re_irc> <Julia> Or maybe I misunderstand what you just said.

02:47 <re_irc> < (@jamesmunns:beeper.com)> the "decompress" takes something that implements "lzss::Write" (instead of core::fmt::Write): https://docs.rs/lzss/latest/lzss/trait.Write.html

02:48 <re_irc> < (@jamesmunns:beeper.com)> so you could do the same trick, implement the "write" function that just increments the count by 1, instead of actually pushing a byte.

02:48 <re_irc> < (@jamesmunns:beeper.com)> Or just does nothing, because I think "compress" also counts the size :D

03:05 <re_irc> < (@yatekii:matrix.org)> was real easy in the end thank you!!

05:32 <re_irc> < (@korken89:matrix.org)> : Cool! Unfortunate that it's still not an MCU :(

05:33 <re_irc> < (@korken89:matrix.org)> I'd think that the killer market for riscv should be like cortex M territory

05:33 <re_irc> < (@korken89:matrix.org)> Extreme volumes and low margins

05:33 <re_irc> < (@korken89:matrix.org)> So if you don't pay license you should have a good advantage

05:34 <re_irc> < (@korken89:matrix.org)> It's probably coming like a tsunami though :D

05:34 <re_irc> < (@korken89:matrix.org)> The only sane ones I've found is the CH32 ones like CH32F307

05:35 <re_irc> < (@korken89:matrix.org)> STM32 copies with riscv core

05:35 <re_irc> < (@9names:matrix.org)> e907 is an mcu. on bl808 it has a app processor strapped on (not well integrated tbh), but bl616 is basically the same with just the e907.

05:35 <re_irc> < (@korken89:matrix.org)> Oh

05:35 <re_irc> < (@9names:matrix.org)> yeah if they're shipping i'm sure we'll see a ch with a similar config soon enough

05:35 <re_irc> < (@jamesmunns:beeper.com)> esp32-cx series are also riscv mcus

05:36 <re_irc> < (@korken89:matrix.org)> I have too look closer i found like an raspberry pi 4

05:36 <re_irc> < (@jamesmunns:beeper.com)> not cortex-m flavored at all, afaik :D

05:36 <re_irc> < (@9names:matrix.org)> yeah but they don't have CLIC afaik?

05:36 <re_irc> < (@korken89:matrix.org)> They don't :(

05:36 <re_irc> < (@korken89:matrix.org)> I already checked esp32

05:36 <re_irc> < (@jamesmunns:beeper.com)> ah, checking for clic + mcu format

05:36 <re_irc> < (@jamesmunns:beeper.com)> (now I get it)

05:37 <re_irc> < (@korken89:matrix.org)> I really want to try the new multi arch support for RTIC, but for that i need an MCU to test against 😅

05:37 <re_irc> < (@korken89:matrix.org)> (the codegen is now made hardware agnostic, in theory)

05:38 <re_irc> < (@jamesmunns:beeper.com)> aw, the wch569 doesn't use the CLIC either

05:38 <re_irc> < (@korken89:matrix.org)> I ordered CH32F307 as it has a good enough interrupt controller

05:38 <re_irc> < (@jamesmunns:beeper.com)> used the "PFIC"

05:38 <re_irc> < (@korken89:matrix.org)> It has its own CLIC like controller

05:38 <re_irc> < (@korken89:matrix.org)> Same features at least

05:39 <re_irc> < (@korken89:matrix.org)> Then comes the fun issue of having one bindings feature per riscv core we support until CLIC is standard 😅

05:39 <re_irc> < (@jamesmunns:beeper.com)> (the ch569 is a neat MCU with a USB3.0 host/device controller)

05:39 <re_irc> < (@korken89:matrix.org)> It's cool

05:39 <re_irc> < (@9names:matrix.org)> nice nice nice nice nice

05:40 <re_irc> < (@korken89:matrix.org)> CH32F307 has 2.0 HS and GigE Ethernet

05:40 <re_irc> < (@korken89:matrix.org)> But so far I'm only interested in getting RTIC running :D

05:40 <re_irc> < (@korken89:matrix.org)> I want it to be ready for when riscv becomes mainstream

05:42 <re_irc> < (@korken89:matrix.org)> Maybe i should try on esp32 as well, see how much we can support

05:45 <re_irc> < (@korken89:matrix.org)> I'm trying to find the docs on their interrupt controller :P

05:48 <re_irc> < (@korken89:matrix.org)> Seems too be preemptive

05:49 <re_irc> < (@korken89:matrix.org)> And it has basepri like functionality

05:51 <re_irc> < (@korken89:matrix.org)> I'll order a dev board for that as well:)

05:51 <re_irc> < (@korken89:matrix.org)> Worth experimenting

05:51 <re_irc> < (@korken89:matrix.org)> Does anyone know once that has SWD or JTAG port?

05:52 <re_irc> < (@korken89:matrix.org)> * one

06:08 <re_irc> < (@9names:matrix.org)> the riscv ESP's have builtin usb-jtag with probe-rs support, if that's what you're referring to. it's super convenient

06:10 <re_irc> < (@korken89:matrix.org)> Oh nice

06:10 <re_irc> < (@korken89:matrix.org)> No need for a probe then :D

07:08 <re_irc> < (@matoushybl:matrix.org)> They do have the builtin jtag but on many devkits they still use usb/uart converter for programming.

07:08 <re_irc> This one has the native jtag: https://github.com/esp-rs/esp-rust-board

07:09 <re_irc> < (@matoushybl:matrix.org)> They do have the builtin jtag but on many devkits they still use usb/uart converter for programming.
This one has the native jtag: https://github.com/esp-rs/esp-rust-board (can be bought on aliexpress or mouser)

07:17 <re_irc> < (@korken89:matrix.org)> Nice

16:21 <re_irc> < (@luojia65:matrix.org)> Happy new year! It's year of the rabbit now

17:46 <re_irc> < (@yatekii:matrix.org)> : how is the animal determined?

17:47 <re_irc> <Félix the Newbie> I think it's a cycle

17:47 <re_irc> < (@korken89:matrix.org)> Does anyone know is a static site generator that has a template for the rust website kind of look?

17:47 <re_irc> < (@korken89:matrix.org)> * of

19:31 <re_irc> < (@CyReVolt:matrix.org)> Félix the Newbie: yep, 12 years with one animal each, combined with 5 elements - so a 60-year cycle in total :-)

19:31 <re_irc> Now the year of the 🐅 (my year) just ended, and this is now the year of the water rabbit.

19:31 <re_irc> We are featuring a special version of the logo on https://coreboot.org accordingly - a hare being the coreboot mascot :-)

20:28 <re_irc> <dngrs (spookyvision@{github,cohost})> that's a very pretty logo

23:26 <re_irc> <Hexafox> Hey all o/

23:27 <re_irc> <Hexafox> I'm currently on a long term project to try and get rust on the nintendo ds.

23:30 <re_irc> <dngrs (spookyvision@{github,cohost})> I assume you've already found https://github.com/rrohrer/rs-nds ?