iqmc_cooper2 regression test fails in numba mode on linux OS

[spasmann]

My branch iqmc/source_tilt is failing the iqmc_cooper2 regression test in Numba mode when I push to GitHub. You can see the test results here.
What's odd is it passes in pure python and numba mode on Windows and Mac.

This seems to be a Linux issue because it also passes in Python but fails in Numba on Quartz. I've tried messing with some of the input parameters, increasing/decreasing particles, but still can't get the test to pass.

Mac results:

PC results:

Linux (Quartz) results:

[spasmann]

Interestingly, the test is passing when I double the number of cells in each dimension of the tally mesh from 20 to 40. Here is the original input deck

import numpy as np
import mcdc


# =============================================================================
# Set model
# =============================================================================
# A shielding problem based on Problem 2 of [Coper NSE 2001]
# https://ans.tandfonline.com/action/showCitFormats?doi=10.13182/NSE00-34

# Set materials
SigmaT = 5.0
c = 0.8
m_barrier = mcdc.material(capture=np.array([SigmaT]), scatter=np.array([[SigmaT * c]]))
SigmaT = 1.0
m_room = mcdc.material(capture=np.array([SigmaT]), scatter=np.array([[SigmaT * c]]))

# Set surfaces
sx1 = mcdc.surface("plane-x", x=0.0, bc="reflective")
sx2 = mcdc.surface("plane-x", x=2.0)
sx3 = mcdc.surface("plane-x", x=2.4)
sx4 = mcdc.surface("plane-x", x=4.0, bc="vacuum")
sy1 = mcdc.surface("plane-y", y=0.0, bc="reflective")
sy2 = mcdc.surface("plane-y", y=2.0)
sy3 = mcdc.surface("plane-y", y=4.0, bc="vacuum")

# Set cells
mcdc.cell([+sx1, -sx2, +sy1, -sy2], m_room)
mcdc.cell([+sx1, -sx4, +sy2, -sy3], m_room)
mcdc.cell([+sx3, -sx4, +sy1, -sy2], m_room)
mcdc.cell([+sx2, -sx3, +sy1, -sy2], m_barrier)

# =============================================================================
# iQMC Parameters
# =============================================================================
N = 2e2
Nx = Ny = 20
maxit = 2
tol = 1e-3
x = np.linspace(0, 4, num=Nx + 1)
y = np.linspace(0, 4, num=Ny + 1)
generator = "halton"

# fixed source in lower left corner
fixed_source = np.zeros((Nx, Ny))
fixed_source[0 : int(0.25 * Nx), 0 : int(0.25 * Nx)] = 1

phi0 = np.ones((Nx, Ny))

mcdc.iQMC(
    x=x,
    y=y,
    fixed_source=fixed_source,
    phi0=phi0,
    maxitt=maxit,
    tol=tol,
    generator=generator,
    score=["tilt-x", "tilt-y", "tilt-xy"],
)

# =============================================================================
# Set tally, setting, and run mcdc
# =============================================================================
# Setting
mcdc.setting(N_particle=N)
# Run
mcdc.run()

[jpmorgan98]

Could we change the test to do that? I know that's an unsatisfying result but this would allow us to move on

[spasmann]

Sorry forgot to follow up on this, but that's what I did. The test is passing now.