[mlir][sparse][python] add an "exhaustive" sparse test using python

[aartbik]

Summary

Using the python API to easily set up sparse kernels, this test
exhaustively builds, compilers, and runs SpMM for all annotations
on a sparse tensor, making sure every version generates the correct
result. This test also illustrates using the python API to set up
a sparse kernel and sparse compilation.

[bixia]

mlir/test/python/dialects/sparse_tensor/test_SpMM.py

17–20

I don't see any use of this routine?

[aartbik]

mlir/test/python/dialects/sparse_tensor/test_SpMM.py

17–20

It is an idiom that is mandatory for the MLIR python tests (see other files).
The use is very hidden at L130 ;-)

mlir/test/python/dialects/sparse_tensor/test_SpMM.py

+# RUN: SUPPORT_LIB=%mlir_runner_utils_dir/libmlir_c_runner_utils%shlibext %PYTHON %s | FileCheck %s
+
+import os
+import ctypes
+import mlir.all_passes_registration
+import numpy as np
+
+from mlir.dialects import builtin
+from mlir.dialects.linalg.opdsl.lang import *
+from mlir.dialects.sparse_tensor import *
+from mlir.execution_engine import *
+from mlir.ir import *
+from mlir.passmanager import *
+from mlir.runtime import *
+
+
+def run(f):
+  print('\nTEST:', f.__name__)
+  f()
+  return f
+
+
+@linalg_structured_op
+def matmul_dsl(
+    A=TensorDef(T, S.M, S.K),
+    B=TensorDef(T, S.K, S.N),
+    C=TensorDef(T, S.M, S.N, output=True)):
+  C[D.m, D.n] += A[D.m, D.k] * B[D.k, D.n]
+
+
+def build_SpMM(attr: EncodingAttr):
+  """Build SpMM kernel.
+
+  This method generates a linalg op with for matrix multiplication using
+  just the Python API. Effectively, a generic linalg op is constructed
+  that computes C(i,j) += A(i,k) * B(k,j) for annotated matrix A.
+  """
+  module = Module.create()
+  f64 = ir.F64Type.get()
+  a = RankedTensorType.get([3, 4], f64, attr)
+  b = RankedTensorType.get([4, 2], f64)
+  c = RankedTensorType.get([3, 2], f64)
+  arguments = [a, b, c]
+  with InsertionPoint(module.body):
+
+    @builtin.FuncOp.from_py_func(*arguments)
+    def spMxM(*args):
+      return matmul_dsl(args[0], args[1], outs=[args[2]])
+
+  return module
+
+
+def boilerplate(attr: EncodingAttr):
+  """Returns boilerplate main method.
+
+  This method sets up a boilerplate main method that calls the generated
+  sparse kernel. For convenience, this part is purely done as string input.
+  """
+  return f"""
+func @main(%c: tensor<3x2xf64>) -> tensor<3x2xf64>
+  attributes {{ llvm.emit_c_interface }} {{
+  %0 = constant dense<[ [ 1.1,  0.0,  0.0,  1.4 ],
+                        [ 0.0,  0.0,  0.0,  0.0 ],
+                        [ 0.0,  0.0,  3.3,  0.0 ]]> : tensor<3x4xf64>
+  %a = sparse_tensor.convert %0 : tensor<3x4xf64> to tensor<3x4xf64, {attr}>
+  %b = constant dense<[ [ 1.0,  2.0 ],
+                        [ 4.0,  3.0 ],
+                        [ 5.0,  6.0 ],
+                        [ 8.0,  7.0 ]]> : tensor<4x2xf64>
+  %1 = call @spMxM(%a, %b, %c) : (tensor<3x4xf64, {attr}>,
+                                  tensor<4x2xf64>,
+                                  tensor<3x2xf64>) -> tensor<3x2xf64>
+  return %1 : tensor<3x2xf64>
+}}
+"""
+
+
+def build_compile_and_run_SpMM(attr: EncodingAttr, support_lib: str, compiler):
+  # Build.
+  module = build_SpMM(attr)
+  func = str(module.operation.regions[0].blocks[0].operations[0].operation)
+  module = Module.parse(func + boilerplate(attr))
+  # Compile.
+  compiler(module)
+  execution_engine = ExecutionEngine(
+      module, opt_level=0, shared_libs=[support_lib])
+  # Set up numpy input, invoke the kernel, and get numpy output.
+  # Built-in bufferization uses in-out buffers.
+  # TODO: replace with inplace comprehensive bufferization.
+  Cin = np.zeros((3, 2), np.double)
+  Cout = np.zeros((3, 2), np.double)
+  Cin_memref_ptr = ctypes.pointer(
+      ctypes.pointer(get_ranked_memref_descriptor(Cin)))
+  Cout_memref_ptr = ctypes.pointer(
+      ctypes.pointer(get_ranked_memref_descriptor(Cout)))
+  execution_engine.invoke('main', Cout_memref_ptr, Cin_memref_ptr)
+  Cresult = ranked_memref_to_numpy(Cout_memref_ptr[0])
+
+  # Sanity check on computed result.
+  expected = [[12.3, 12.0], [0.0, 0.0], [16.5, 19.8]]
+  if np.allclose(Cresult, expected):
+    pass
+  else:
+    quit(f'FAILURE')
+
+
+class SparseCompiler:
+  """Sparse compiler passes."""
+
+  def __init__(self, options: str):
+    pipeline = (
+        f'sparsification{{{options}}},'
+        f'sparse-tensor-conversion,'
+        f'builtin.func(convert-linalg-to-loops,convert-vector-to-scf),'
+        f'convert-scf-to-std,'
+        f'func-bufferize,'
+        f'tensor-constant-bufferize,'
+        f'builtin.func(tensor-bufferize,std-bufferize,finalizing-bufferize),'
+        f'convert-vector-to-llvm{{reassociate-fp-reductions=1 enable-index-optimizations=1}},'
+        f'convert-memref-to-llvm,'
+        f'convert-std-to-llvm')
+    self.pipeline = pipeline
+
+  def __call__(self, module: Module):
+    PassManager.parse(self.pipeline).run(module)
+
+
+# CHECK-LABEL: TEST: testSpMM
+# CHECK: Passed 72 tests
+@run
+def testSpMM():
+  support_lib = os.getenv('SUPPORT_LIB')
+  with Context() as ctx, Location.unknown():
+    count = 0
+    # Fixed compiler optimization strategy.
+    # TODO: explore state space here too
+    par = 0
+    vec = 0
+    vl = 1
+    e = False
+    opt = (f'parallelization-strategy={par} '
+           f'vectorization-strategy={vec} '
+           f'vl={vl} enable-simd-index32={e}')
+    # Exhaustive loop over various ways to annotate a kernel with
+    # a *single* sparse tensor. Even this subset already gives
+    # quite a large state space!
+    levels = [[DimLevelType.dense, DimLevelType.dense],
+              [DimLevelType.dense, DimLevelType.compressed],
+              [DimLevelType.compressed, DimLevelType.dense],
+              [DimLevelType.compressed, DimLevelType.compressed]]
+    orderings = [
+        AffineMap.get_permutation([0, 1]),
+        AffineMap.get_permutation([1, 0])
+    ]
+    bitwidths = [0, 8, 32]
+    for levels in levels:
+      for ordering in orderings:
+        for pwidth in bitwidths:
+          for iwidth in bitwidths:
+            attr = EncodingAttr.get(levels, ordering, pwidth, iwidth)
+            compiler = SparseCompiler(options=opt)
+            build_compile_and_run_SpMM(attr, support_lib, compiler)
+            count = count + 1
+    print('Passed ', count, 'tests')