xai-org/x-algorithm · error · ValueError

SWIZZLE_128B_BASE32B is invalid for Major-K

Error message

SWIZZLE_128B_BASE32B is invalid for Major-K

What it means

The 128B_BASE32B swizzle mode (Swizzle<2,5,2>, used for 32-bit-base fp8/mxfp8 tcgen05 operands on Blackwell) has no descriptor encoding for K-major operands in this builder — the PTX smem descriptor format only defines it for MN-major layouts. make_smem_desc_base rejects the combination up front.

Source

Thrown at phoenix/xrex/cutedsl/ranker_fa4/mma_sm100_desc.py:262

    if major is Major.MN:
        swizzle_atom_k_size = 4 if layout_type is LayoutType.SWIZZLE_128B_BASE32B else 8
        canonical_layout = cute.logical_divide(layout, (swizzle_atom_mn_size, swizzle_atom_k_size))
        if not cute.is_congruent(canonical_layout, ((1, 1), (1, 1))):
            raise ValueError("Not a canonical UMMA_MN Layout: Expected profile failure.")
        stride_00 = canonical_layout.stride[0][0]
        if layout_type is not LayoutType.SWIZZLE_NONE and stride_00 != 1:
            raise ValueError("Not a canonical UMMA_MN Layout: Expected stride failure.")
        stride_10 = canonical_layout.stride[1][0]
        if stride_10 != swizzle_atom_mn_size:
            raise ValueError("Not a canonical UMMA_MN Layout: Expected stride failure.")
        stride_01, stride_11 = canonical_layout.stride[0][1], canonical_layout.stride[1][1]
        if layout_type is LayoutType.SWIZZLE_NONE:
            stride_byte_offset, leading_byte_offset = stride_01, stride_11
        else:
            stride_byte_offset, leading_byte_offset = stride_11, stride_01
    else:
        if layout_type == LayoutType.SWIZZLE_128B_BASE32B:
            raise ValueError("SWIZZLE_128B_BASE32B is invalid for Major-K")
        if not cute.size(layout.shape[0]) % 8 == 0:
            raise ValueError("Not a canonical UMMA_K Layout: Expected MN-size multiple of 8.")
        canonical_layout = cute.logical_divide(layout, (8, 2))
        if not cute.is_congruent(canonical_layout, ((1, 1), (1, 1))):
            raise ValueError("Not a canonical UMMA_K Layout: Expected profile failure.")
        stride_00 = canonical_layout.stride[0][0]
        if stride_00 != swizzle_atom_mn_size:
            raise ValueError("Not a canonical UMMA_K Layout: Expected stride failure.")
        stride_10 = canonical_layout.stride[1][0]
        if layout_type is not LayoutType.SWIZZLE_NONE and stride_10 != 1:
            raise ValueError("Not a canonical UMMA_K Layout: Expected stride failure.")
        stride_01 = canonical_layout.stride[0][1]
        stride_byte_offset, leading_byte_offset = stride_01, stride_10

    desc = 0
    desc |= (leading_byte_offset & 0x3FFF) << 16
    desc |= (stride_byte_offset & 0x3FFF) << 32
    desc |= (VERSION & 0x3) << 46

View on GitHub (pinned to 24c60942c5)

Solutions

1. Store the operand MN-major (swap to an MN-major smem layout) so Swizzle<2,5,2> is legal
2. If K-major is required, switch to a standard swizzle (Swizzle<3,4,3> 128B or <2,4,3> 64B) instead of 128B_BASE32B
3. Set the MMA operand major mode to MN in the MmaOp so mma_op_to_idesc and the smem descriptor agree

Example fix

# before
desc = make_smem_desc_base(kmajor_layout, cute.Swizzle(2,5,2), Major.K)
# after
desc = make_smem_desc_base(mnmajor_layout, cute.Swizzle(2,5,2), Major.MN)

if sw.num_base == 5:  # Swizzle<2,5,2> = 128B_BASE32B
    assert major is Major.MN, '128B_BASE32B requires MN-major operand'

def base32b_ok(sw: cute.Swizzle, major: Major) -> bool:
    return not (sw.num_base == 5 and major is Major.K)

When it happens

Trigger: Calling make_smem_desc_base (directly or via gemm_ptx*/smem_desc_base_from_tensor) with swizzle Swizzle<2,5,2> and major=Major.K, e.g. an fp8 B-operand stored K-major in shared memory.

Common situations: Using mxfp8/fp8 tensors where the B operand arrives K-major from a transpose-free GEMM epilogue; porting a CUTLASS CollectiveBuilder config with K-major B and reusing the 128B_BASE32B smem atom.

Related errors

• Unsupported swizzle triple for UMMA smem descriptor
• Not a canonical UMMA_MN Layout: Expected profile failure.
• Not a canonical UMMA_MN Layout: Expected stride failure.
• Not a canonical UMMA_K Layout: Expected MN-size multiple of
• Not a canonical UMMA_K Layout: Expected profile failure.

AI-assisted analysis of xai-org/x-algorithm@24c60942c5 (2026-08-28). Data as JSON: /api/errors/0d7f09fc808f6765. Report an issue: GitHub.