Mojo function

gemm_kernel_amd

gemm_kernel_amd[c_type: DType, c_layout: Layout, a_type: DType, a_layout: Layout, b_type: DType, b_layout: Layout, transpose_b: Bool, c_layout_int_type: DType, a_layout_int_type: DType, b_layout_int_type: DType, c_linear_idx_type: DType, a_linear_idx_type: DType, b_linear_idx_type: DType, config: MatmulConfig[a_type, b_type, c_type, transpose_b], elementwise_lambda_fn: OptionalReg[fn[DType, Int, Int](IndexList[2], SIMD[$0, $1]) capturing -> None] = OptionalReg[fn[DType, Int, Int](IndexList[2], SIMD[$0, $1]) capturing -> None]({:i1 0, 1})](c: LayoutTensor[c_type, c_layout, MutableAnyOrigin, layout_int_type=c_layout_int_type, linear_idx_type=c_linear_idx_type], a: LayoutTensor[a_type, a_layout, MutableAnyOrigin, layout_int_type=a_layout_int_type, linear_idx_type=a_linear_idx_type], b: LayoutTensor[b_type, b_layout, MutableAnyOrigin, layout_int_type=b_layout_int_type, linear_idx_type=b_linear_idx_type])

AMD-optimized GEMM kernel for matrix multiplication C = A * B.

This kernel implements an efficient matrix multiplication algorithm optimized for AMD GPUs, with hierarchical tiling and structured memory access patterns.

Parameters:

c_type (DType): Data type for the output matrix C.
c_layout (Layout): Memory layout for matrix C.
a_type (DType): Data type for the input matrix A.
a_layout (Layout): Memory layout for matrix A.
b_type (DType): Data type for the input matrix B.
b_layout (Layout): Memory layout for matrix B.
transpose_b (Bool): Whether matrix B should be transposed.
c_layout_int_type (DType): Data type for the integer part of matrix C.
a_layout_int_type (DType): Data type for the integer part of matrix A.
b_layout_int_type (DType): Data type for the integer part of matrix B.
c_linear_idx_type (DType): Data type for the linear index of matrix C.
a_linear_idx_type (DType): Data type for the linear index of matrix A.
b_linear_idx_type (DType): Data type for the linear index of matrix B.
config (MatmulConfig[a_type, b_type, c_type, transpose_b]): GEMM configuration parameters (tile sizes, etc.).
elementwise_lambda_fn (OptionalReg[fn[DType, Int, Int](IndexList[2], SIMD[$0, $1]) capturing -> None]): Optional function to apply to output elements.

Args:

c (LayoutTensor[c_type, c_layout, MutableAnyOrigin, layout_int_type=c_layout_int_type, linear_idx_type=c_linear_idx_type]): Output matrix C (result).
a (LayoutTensor[a_type, a_layout, MutableAnyOrigin, layout_int_type=a_layout_int_type, linear_idx_type=a_linear_idx_type]): Input matrix A.
b (LayoutTensor[b_type, b_layout, MutableAnyOrigin, layout_int_type=b_layout_int_type, linear_idx_type=b_linear_idx_type]): Input matrix B (must be transposed).