Segfault when using `new_random()` on larger matrices

[paulstey]

Segfault when using `new_random()` on larger matrices

[paulstey]

Thanks so much for your help, @Andlon!

Your intuition is correct, this seems to be a stack overflow. The output below is what I get with RUST_BACKTRACE=1. And yes, you're also correct that I get the same result when using repeat() rather than new_random().

I will use DMatrix as you suggest. Thanks again for your help, @Andlon! And thanks also for this phenomenal crate!!

thread 'main' has overflowed its stack
fatal runtime error: stack overflow
Aborted

[paulstey]

Oh, and here is my Cargo.toml file.

[package]
name = "nalgebra_mat_mul"
version = "0.1.0"
edition = "2021"

[dependencies]
nalgebra = { version = "0.32.2", features = ["rand"] }

[Andlon]

Judging from your example, this is probably a stack overflow. Can you give the output with debug symbols turned on? It would be great to confirm. If that's the case, you'd probably get the same problem if you don't use get_random but instead just Matrix::repeat(2.0) (avoiding 0.0 since compilers/allocators etc. sometimes go alternate routes for zeroed memory).

A 512x512 f64 matrix requires 16 MB stack space, which is likely larger than the stack size your OS uses by default.

For such large matrices, there are usually no benefits to storing them on the stack. I'd recommend to use DMatrix instead, if you can.