I would like to summarize the target of the project for my own understanding, kindly guide me if I am deviating from the goal here.
Let's say we want to use a trigo functions from numpy like sin(x). For a snippet like -

from numpy import array, empty, float64,

def f(xi32: f64[:]) -> f64:
    xi32[0] = 0.1
    xi32[1] = 0.2
    xi32[2] = 0.3
    print(sin(xi32))

def d():
    ai32: f64[3] = empty(3, dtype=float64)
    f(ai32)

d()

Right now, this is handled here and due to the@vectorizedecorator we are able to handle array inputs.

lpython/src/runtime/lpython_intrinsic_numpy.py

Lines 9 to 12 in 6098157

    
           @overload 
        
           @vectorize 
        
           def sin(x: f64) -> f64: 
        
               return _lfortran_dsin(x)

So we already are able to generate the ASR, from lpython_instrinsic_numpy.py file but we want this to be handled by the instantiate_Sin/ create_Sin functions from intrinsic_function_registry.h.

	@overload
	@vectorize
	def sin(x: f64) -> f64:
	return _lfortran_dsin(x)

Implement all NumPy intrinsics via ASR #2474

Description

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions