Skip to main content
Log in

Mojo function

get_conv_tile_shape

get_conv_tile_shape[type: DType](c: Int, filter_window_size: Int, micro_kernel_width: Int) -> Index[2]

Compute the (c, f) tile shape in L2. Assume NHWC layout, the tile shape is (R, S, c_tile, f_tile). R and S are by default fully covered. The heuristic tried to block in C as much as possible. If C is small, it would start to block F.