Vulkan floor_divide gives incorrect outputs with integer divisor

[GregoryComer]

🐛 Describe the bug

Floor divide ops delegated to Vulkan give incorrect outputs when the divisor (or dividend) are not floats. When both are floats, output appears valid.

import torch
from executorch.backends.vulkan.partitioner.vulkan_partitioner import VulkanPartitioner
from executorch.exir import to_edge_transform_and_lower, EdgeCompileConfig, to_edge
from executorch.extension.pybindings.portable_lib import _load_for_executorch_from_buffer
from typing import Callable, List, Optional, Tuple, Union

class FloorDivideModel(torch.nn.Module):
    def __init__(self):
        super().__init__()
        
    def forward(self, x, y):
        return torch.floor_divide(x, y)
        
model = FloorDivideModel()
inputs = (
    torch.randint(-100, 100, (10, 10)).to(torch.float),
    torch.full((10, 10), 2).clone() # Divisor of 2 
)
eager_outputs = model(*inputs)

ep = torch.export.export(model.eval(), inputs)
print(ep)
lowered = to_edge_transform_and_lower(
    ep,
    partitioner=[VulkanPartitioner()],
    compile_config=EdgeCompileConfig(_check_ir_validity=False)
).to_executorch()
print(lowered.exported_program())

et_model = _load_for_executorch_from_buffer(lowered.buffer)
et_outputs = et_model([*inputs])[0]

print(f"Inputs: {inputs}")
print(f"Eager: {eager_outputs}")
print(f"ET:    {et_outputs}")

Outputs:

Eager: tensor([[ 14., -23.,  21.,   4., -26.,  22., -17.,  14., -21., -20.],
        [ 23.,  23.,  17., -24.,  40., -27.,  -6., -36., -44.,  11.],
        [-23.,  30., -45., -22.,  23.,  45., -15., -42.,  49.,   3.],
        [ -6.,  38.,  28.,   6., -29., -36.,  46.,   4.,   8., -26.],
        [ 49.,  -7., -45.,  18.,  36.,  18., -40.,  25.,  15.,   0.],
        [ 20., -34.,  43., -28.,  13., -48.,  20.,  18.,   0., -14.],
        [-40.,   9.,  38.,  49., -40.,  43.,  -3., -48.,  30., -10.],
        [ 13.,  32.,  -1.,  22.,  27.,  38., -35.,  35.,  15., -17.],
        [ 13., -36.,  20., -37.,  43.,  44., -34., -26.,   0.,  -8.],
        [-16.,  15.,  33., -10.,  21., -21.,  14.,  25., -31.,  26.]])
ET:    tensor([[inf, -inf, inf, inf, -inf, inf, -inf, inf, -inf, -inf],
        [inf, inf, inf, -inf, inf, -inf, -inf, -inf, -inf, inf],
        [-inf, inf, -inf, -inf, inf, inf, -inf, -inf, inf, inf],
        [-inf, inf, inf, inf, -inf, -inf, inf, inf, inf, -inf],
        [inf, -inf, -inf, inf, inf, inf, -inf, inf, inf, nan],
        [inf, -inf, inf, -inf, inf, -inf, inf, inf, nan, -inf],
        [-inf, inf, inf, inf, -inf, inf, -inf, -inf, inf, -inf],
        [inf, inf, -inf, inf, inf, inf, -inf, inf, inf, -inf],
        [inf, -inf, inf, -inf, inf, inf, -inf, -inf, nan, -inf],
        [-inf, inf, inf, -inf, inf, -inf, inf, inf, -inf, inf]])

Versions

Run on Meta internal master, Jul 3, fbcode/SwiftShader

cc @SS-JIA @manuelcandales @cbilgin