elemental_01.py test seems to fail something

[Thirumalai-Shaktivel]

I think the issue is related to the array initialisation,

lpython/integration_tests/elemental_01.py

Lines 106 to 107 in 37eca1a

	arraynd: f64[256, 64, 16] = empty((256, 64, 16), dtype=float64)
	sinnd: f64[256, 64, 16] = empty((256, 64, 16), dtype=float64)

[certik]

These arrays are 2MB each, which doesn't seem too big, although on the Apple M1 it seems the stack is only 8MB, so maybe it runs out of stack?

[certik]

A quick check is to try to make them smaller and see if the tests passes. If it does, it's running out of stack.

[Thirumalai-Shaktivel]

from lpython import i32, f32, f64
from numpy import float32, empty, float64, sin

def elemental_sin():
    i: i32
    j: i32
    k: i32

    array1d: f32[256] = empty(256, dtype=float32)
    sin1d: f32[256] = empty(256, dtype=float32)

    for i in range(256):
        array1d[i] = f32(i)

    sin1d = sin(sin(array1d))

    arraynd: f64[256, 64, 16] = empty((256, 64, 16), dtype=float64)
    sinnd: f64[256, 64, 16] = empty((256, 64, 16), dtype=float64)

elemental_sin()

$ lpython examples/expr2.py
$ ./expr2.out              
[1]    18897 segmentation fault  ./expr2.out

Replacing array with allocatable works

    arraynd: Allocatable[f64[:,:,:]] = empty((256, 64, 16), dtype=float64)
    sinnd: Allocatable[f64[:,:,:]] = empty((256, 64, 16), dtype=float64)

[certik]

That seems like a bug, as you need to allocate it first.

[ubaidsk]

@Thirumalai-Shaktivel which system do you use and are you able to reproduce it?

[Thirumalai-Shaktivel]

MacBook Pro M1 Pro.
Yup, I'm able to reproduce it.
Did you try in your Linux machine?

[ubaidsk]

MacBook Pro M1 Pro.
Yup, I'm able to reproduce it.

Just tried it. It seems like running the elemental_01.py test alone (without using ./integration_tests/run_tests.py) reproduces it.

[ubaidsk]

I wonder why the CI does not catch it.