numba CUDA中的递归函数

Question

numba 的文档指出：

Recursion support in numba is currently limited to self-recursion with explicit type annotation for the function.

我做了这个简单的设备功能：

@cu.jit(numba.i4(numba.i4), device=True)
def mutate(val: int) -> int:
    if(val < 1):
        return val
    else:
        return mutate(val-1)

这是一个相当简单的递归测试。现在从我的内核代码调用这个函数我得到 Untyped global name 'mutate': cannot determine Numba type of <class 'numba.ir.UndefinedType'> 错误。我还应该如何指定函数的类型？我该如何解决这个问题？

Answer 1

Recursion support in numba is currently limited to self-recursion with explicit type annotation for the function.

首先要说明的是，numba 文档不包含您引用的文本。那来自 Numba Enhancement Proposal 6。一项扩展 numba 工作方式的提议。不是 language/compiler.

的功能

也就是说，这个：

import numba

@numba.jit(numba.i4(numba.i4))
def mutate(val: int) -> int:
    if(val < 1):
        return val
    else:
        return mutate(val-1)

如您所料的那样工作：

In [12]: %run recursion.py

In [13]: mutate??
Signature:       mutate(val:int) -> int
Call signature:  mutate(*args, **kwargs)
Type:            CPUDispatcher
String form:     CPUDispatcher(<function mutate at 0x7f77f787f840>)
File:            ~/SO/recursion.py
Source:         
@numba.jit(numba.i4(numba.i4))
def mutate(val: int) -> int:
    if(val < 1):
        return val
    else:
        return mutate(val-1)
Class docstring:
Implementation of user-facing dispatcher objects (i.e. created using
the @jit decorator).
This is an abstract base class. Subclasses should define the targetdescr
class attribute.
Init docstring: 
Parameters
----------
py_func: function object to be compiled
locals: dict, optional
    Mapping of local variable names to Numba types.  Used to override
    the types deduced by the type inference engine.
targetoptions: dict, optional
    Target-specific config options.
impl_kind: str
    Select the compiler mode for `@jit` and `@generated_jit`
pipeline_class: type numba.compiler.BasePipeline
    The compiler pipeline type.

In [14]: print(mutate(10))
0

但是 Numba CUDA 编译器（我猜是在 nopython 模式下编译）不会编译等效代码，正如您所发现的那样。由此，加上 Numba CUDA 文档中没有提到递归这一事实，我得出结论，Numba CUDA 编译器不支持递归。

numba CUDA中的递归函数

Recursive function in numba CUDA

python

cuda

numba