Picolibc 1.7.5 Released For Embedded Systems With Limited RAM

[phoronix]

Phoronix: Picolibc 1.7.5 Released For Embedded Systems With Limited RAM

Keith Packard continues developing Picolibc as his C standard library alternative to the likes of Musl and uClibc for a libc implementation that runs well on embedded hardware, especially for platforms with limited amounts of RAM...

https://www.phoronix.com/scan.php?page=news_item&px=Picolibc-1.7.5-Released

[uid313]

It's too bad Keith Packard picked GPL for his programming language Snek, which makes it impossible to embed the interpreter in apps. I wish he had picked the MIT or BSD license, or even the LGPL.

[camel_case]

Problem with newlib is that it is difficult to create stack-only applications if you want to use the comfort of *ptrintf and so on. On very small systems thats actually sometimes an requirement. People implement a lot of standard lib functionality themself. Could picolib an better effort? (licence will suck anyway ..)

[binarybanana]

Some numbers would be interesting. I know from personal experiece that swapping out the libc can lead to significant differences in memory usage. For example, htop with glibc uses ~5MB memory (RSS) on amd64/arm64, but only ~2.5MB with musl. Haven't looked deeper than that and larger programs likely benefit less, but for embedded devices running a bunch of busybox programs the difference is probably as substantial overall as with htop, if not even more so.

[onlyLinuxLuvUBack]

Some numbers would be interesting. I know from personal experiece that swapping out the libc can lead to significant differences in memory usage. For example, htop with glibc uses ~5MB memory (RSS) on amd64/arm64, but only ~2.5MB with musl. Haven't looked deeper than that and larger programs likely benefit less, but for embedded devices running a bunch of busybox programs the difference is probably as substantial overall as with htop, if not even more so.

1-2-3-4-5-1.jpg

but i prefer a phoronix benchmark and graph probably zig+musl vs picolib when it works

[kretikus]

There are some other small libc implementations out there dietlibc for example:

https://www.fefe.de/dietlibc/

A comparison of the different libc implementations might be fun

[AHSauge]

Problem with newlib is that it is difficult to create stack-only applications if you want to use the comfort of *ptrintf and so on. On very small systems thats actually sometimes an requirement. People implement a lot of standard lib functionality themself. Could picolib an better effort? (licence will suck anyway ..)

For more safety critical things, stack-only is quite beneficial, if not even a hard-requirement, so it would be interesting to see if picolib solves that in some way.

Regarding licensing though: What's the problem? It seems to be pretty much BSD licensed code, which wouldn't cause me much trouble at least, and I would be using this in a fully commercial setting.