I found a helper function to allocate boxed arrays onto the heap, but it overflows at line 8.
What could possibly cause this?
#vec macro causes stack overflow
pallid lintel
shut laurel
Can you try doing
let vec = [val].into_iter().cycle().take(N).collect();
```I thought they fixed the vec macro to not put the whole thing on the stack, but maybe not.
steep crag
it never put anything on the stack
wispy bison
std::iter::repeat(val) would be better than [val].into_iter().cycle().
pallid lintel
Can you try doing
```rs
let vec = [val].into_iter().cycle().take(N).collect();
`...
Now it overflows at line 7.
steep crag
can you post the error?
pallid lintel
After line 7 the call stack only has __chkstk
steep crag
then maybe you're calling it recursively?
steep crag
what is an Option of?
pallid lintel
Some large data for the GPU
steep crag
well then this alone could overflow
steep crag
that's a lot
a couple of them on the stack at once could blow up
vec![val; N] doesn't allocate the whole thing on the stack, but it could have like 2 at once
pallid lintel
Is there another way to get Box<[Option<gpu::Chunk>; SIZE]> initialized to none?
shut laurel
The solution here is probably to make it Vec<Brick> and figure out which chunk each brick is in manually.
steep crag
the problem with Option<gpu::Chunk> is that to write it you need to have the whole value
which is big
steep crag
why do you need Box<[Option<gpu::Chunk>; SIZE]>?
shut laurel
You can still get chunk-sized slices out of it
pallid lintel
It's a staging storage of a 2D grid. I write x elements to it where x < SIZE and then send it all to the GPU.
I was hoping I could just send all the visible chunks every frame, but maybe I need to only send mutated chunks or something.
shut laurel
Is there a function to directly copy the contents of a slice to the vram without putting it on the stack?
pallid lintel
The framework I use asks for a &[u8]. I haven't gotten that far to see if it crashes or not.
Looks like they do it this way ptr::copy_nonoverlapping(data.as_ptr(), staging_buffer_ptr, data.len());
And then flush the staging buffer.
