For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /max/get-started.md).

Mojo function

update_w_tile_2d

def update_w_tile_2d[micro_kernel_height: Int, micro_kernel_width: Int, simd_size: Int, effected_by_padding: Bool, has_residual: Bool, last_c_tile: Bool, output_dt: DType, input_dt: DType, filter_dt: DType](output: Pointer[Scalar[output_dt], _safe=False], input: Pointer[Scalar[input_dt], _safe=False], filter: Pointer[Scalar[filter_dt], _safe=False], _init_output: Bool, c_tile_size: Int, f_tile_offset: Int, f_tile_size: Int, conv_shape: ConvShape[Int(2)], n: Int, hw: IndexList[Int(2)])

Updates one output tile of a 2D transposed convolution using register tiling.

Iterates over the filter window for a single output point and accumulates the input and packed filter contributions into the output tile, skipping neighbors that fall inside the padding region.

Parameters:

micro_kernel_height (Int): Height of the micro kernel in the output spatial dimension.
micro_kernel_width (Int): Width of the micro kernel along the output channel dimension.
simd_size (Int): SIMD vector width for the output element type.
effected_by_padding (Bool): Whether the current tile is affected by spatial padding.
has_residual (Bool): Whether the tile contains a residual not aligned to the micro kernel size.
last_c_tile (Bool): Whether this is the last channel tile for epilogue fusion.
output_dt (DType): Element type of the output tensor.
input_dt (DType): Element type of the input tensor.
filter_dt (DType): Element type of the filter tensor.

Args:

output (Pointer[Scalar[output_dt], _safe=False]): Pointer to the output tile to accumulate into.
input (Pointer[Scalar[input_dt], _safe=False]): Pointer to the input tile to read from.
filter (Pointer[Scalar[filter_dt], _safe=False]): Pointer to the packed filter tile to read from.
_init_output (Bool): Unused flag retained for API compatibility.
c_tile_size (Int): Size of the channel tile.
f_tile_offset (Int): Offset of the feature tile within the output channel dimension.
f_tile_size (Int): Size of the feature tile.
conv_shape (ConvShape[Int(2)]): Shape descriptor of the 2D transposed convolution.
n (Int): Batch index of the current sample.
hw (IndexList[Int(2)]): Spatial coordinates of the current output point.