For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /get-started.md).
Mojo function
load_b_nt
def load_b_nt[mma_shape: IndexList[Int(3)], swizzle: Optional[Swizzle] = Optional()](tile: LayoutTensor[address_space=AddressSpace.SHARED, element_layout=tile.element_layout, layout_int_type=tile.layout_int_type, linear_idx_type=tile.linear_idx_type, masked=tile.masked, alignment=tile.alignment]) -> SIMD[tile.dtype, SIMDLength(8)]
Loads the b operand tile for AMD tensor core MFMA from (N, K) storage.
This function supports double-rate MFMA shapes (32x32x16, 16x16x32) with bfloat16 input. Unlike load_b_tr which expects (K, N) storage, this function works with (N, K) storage which is common when transpose_b=True and B is stored row-major.
The input tile (shape = (mma_shape[1], mma_shape[2])) is split along the K dimension into
two halves of shape (MMA_N, MMA_K//2). Each half is loaded using _load_tr16_b64_warp,
which performs a transposed (column-major) load from shared memory. The hardware transpose
effectively converts the (N, K) storage to (K, N) format needed by MMA.
Example: For 16x16x32 MMA with B stored as (N, K) = (16, 32) in LDS:
# B tile in LDS: shape (16, 32) = (MMA_N, MMA_K)
var b_tile = smem_b.tile[16, 32](n_idx, k_idx)
var b_reg = load_b_nt[IndexList[3](16, 16, 32)](b_tile)
# b_reg now contains 8 bf16 values ready for MFMAParameters:
- mma_shape (
IndexList[Int(3)]): The MMA instruction tile shape (only 32x32x16 or 16x16x32 supported). - swizzle (
Optional[Swizzle]): Optional swizzle pattern for bank-conflict-free LDS access.
Args:
- tile (
LayoutTensor[address_space=AddressSpace.SHARED, element_layout=tile.element_layout, layout_int_type=tile.layout_int_type, linear_idx_type=tile.linear_idx_type, masked=tile.masked, alignment=tile.alignment]): ALayoutTensor, residing in shared memory, with shape (mma_shape[1], mma_shape[2]) and dtypeDType.bfloat16. This is (N, K) storage order.
Returns:
SIMD[tile.dtype, SIMDLength(8)]: SIMD[tile.dtype, 8]: Concatenated transposed SIMD loads from both halves of the tile.