llvm
/
torch-mlir
mirror of https://github.com/llvm/torch-mlir
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
torch-mlir/pytest/NumpyCompiler/ndarray_inference.py

27 lines
1.0 KiB
Python
Raw Normal View History

Add a trivial copy elision canonicalization on ndarray->tensor. * This elides the very common code the compiler adds for chaining otherwise tensor-related numpy ops together. * More aggressive canonicalizations would require more advanced analysis.
2020-07-06 09:09:43 +08:00
# RUN: %PYTHON %s | npcomp-opt -split-input-file -npcomp-cpa-type-inference -canonicalize | FileCheck %s --dump-input=fail
Add NdArray type inference conversion.
2020-07-04 07:38:10 +08:00
import numpy as np
from npcomp.compiler import test_config
from npcomp.compiler.frontend import EmittedError
import_global = test_config.create_import_dump_decorator()
global_data = (np.zeros((2, 3)) + [1.0, 2.0, 3.0] * np.reshape([1.0, 2.0],
(2, 1)))
a = np.asarray([1.0, 2.0])
b = np.asarray([3.0, 4.0])
# Test the basic flow of invoking a ufunc call with constants captured from
# a global using explicit function syntax (np.add(a, b)).
# CHECK-LABEL: func @global_add
Allow the ndarray type to carry a shape.
2020-07-06 08:34:03 +08:00
# CHECK-SAME: -> !numpy.ndarray<*:f64>
Add NdArray type inference conversion.
2020-07-04 07:38:10 +08:00
@import_global
def global_add():
Complete the basic spike to perform dtype inference. * Correctly infers the unknown dtypes that emit as part of compilation for the simple ufunc case. * Significant more testing needs to be done on the details now that the pass is minimally functional. * The actual pass itself is still too hacky/not general, but the underlying analysis is further along.
2020-07-06 07:09:16 +08:00
# CHECK-NOT: UnknownType
Add a trivial copy elision canonicalization on ndarray->tensor. * This elides the very common code the compiler adds for chaining otherwise tensor-related numpy ops together. * More aggressive canonicalizations would require more advanced analysis.
2020-07-06 09:09:43 +08:00
# CHECK: numpy.builtin_ufunc_call<"numpy.multiply"> ({{.*}}, {{.*}}) : (tensor<2xf64>, tensor<2xf64>) -> tensor<*xf64>
# CHECK: numpy.builtin_ufunc_call<"numpy.add"> ({{.*}}, {{.*}}) : (tensor<2xf64>, tensor<*xf64>) -> tensor<*xf64>
Complete the basic spike to perform dtype inference. * Correctly infers the unknown dtypes that emit as part of compilation for the simple ufunc case. * Significant more testing needs to be done on the details now that the pass is minimally functional. * The actual pass itself is still too hacky/not general, but the underlying analysis is further along.
2020-07-06 07:09:16 +08:00
# CHECK-NOT: UnknownType
Add a trivial copy elision canonicalization on ndarray->tensor. * This elides the very common code the compiler adds for chaining otherwise tensor-related numpy ops together. * More aggressive canonicalizations would require more advanced analysis.
2020-07-06 09:09:43 +08:00
return np.add(a, np.multiply(a, b))