triton.experimental.gluon.language.shared_memory_descriptor

class triton.experimental.gluon.language.shared_memory_descriptor(self, handle, element_ty, shape, layout, alloc_shape)

Represents a handle to a shared memory allocation in Gluon IR.

Methods

`__init__`(self, handle, element_ty, shape, ...)
`gather`(self, indices, axis[, _semantic])	Gather elements from shared memory along a specified axis using an indices tensor.
`index`(self, index[, _semantic])	Create a subview of shared memory by indexing along the first dimension.
`load`(self, layout[, _semantic])	Load a tensor from shared memory.
`permute`(self, order[, _semantic])	Permute the dimensions of the shared memory descriptor.
`reshape`(self, shape[, _semantic])	Reshape the shared memory descriptor to a new shape and layout.
`scatter`(self, values, indices, axis[, _semantic])	Scatter elements to shared memory along a specified axis using an indices tensor.
`slice`(self, start, length[, dim, _semantic])	Create a subview of shared memory by slicing along a given dimension.
`store`(self, value[, _semantic])	Store a tensor into shared memory.

Attributes