Pass XSMM runner lib to MLIR execution engine

[adam-smnk]

Manually finds and passes XSMM runner library to the MLIR JIT engine to resolve missing TPP xsmm_* symbols when executing from Python.

Works only for Linux currently.

Tested with:

import torch
import torch.nn as nn
import openvino as ov

# Define a synthetic model
class LinearModel(nn.Module):
    def __init__(self, input_size, output_size):
        super(LinearModel, self).__init__()
        self.model = nn.Sequential(
            nn.Linear(input_size, output_size),
            nn.ReLU(),
        )
    def forward(self, a):
        return self.model(a)

# Create an instance of the model
input_size = 1024
output_size = 128
model = LinearModel(input_size, output_size)
# Generate random weights
model.model[0].weight.data.normal_(0, 0.01)
model.model[0].bias.data.fill_(0.01)

input_data = torch.tensor(range(1, input_size*output_size+1)).to(torch.float32)\
    .view(output_size, input_size)

with torch.no_grad():
    reference = model(input_data)
    print('Reference:\n', reference)

# Dynamic shapes
# ov_model = ov.convert_model(model, example_input=input_data)
# Static shapes
ov_model = ov.convert_model(model, input=(ov.Shape([output_size, input_size]), ov.Type.f32))

# expect lots of debug output with MLIR fragments before and after lowering in this call:
ov_compiled = ov.compile_model(ov_model, 'CPU')

tested = ov_compiled(input_data)[0]
print('Tested:\n', tested)
diff = reference - tested
print('Reference - Tested:\n', diff)

[adam-smnk]

Still a bit hacky but the same lib finding mechanism seems to be used for OV plugins.
Long term there must be a better solution but it should work "out of the box" with Python for now.

[rengolin]

Later this could be moved to some sys library, but it should be fine for now, thanks!

[slyalin]

Fails in both C++ and Python with:

LLVM ERROR: out of memory
Allocation failed
Aborted (core dumped)

[slyalin]

@adam-smnk, it fails to compile without TPP_MLIR. Should be under #ifdef TPP_MLIR as well.

[adam-smnk]

Good point