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

cdna4_block_scaled_mfma

def cdna4_block_scaled_mfma[a_scale_byte_index: Int32, b_scale_byte_index: Int32, a_matrix_format: CDNA4F8F6F4MatrixFormat, b_matrix_format: CDNA4F8F6F4MatrixFormat](mut d: SIMD[DType.float32], a: SIMD[DType.uint8], b: SIMD[DType.uint8], packed_scale_word_a: Int32, packed_scale_word_b: Int32)

Executes a CDNA4 f8f6f4 block-scaled MFMA, inferring the MMA shape from the accumulator width (16 lanes -> 32x32x64, 4 lanes -> 16x16x128).

a_scale_byte_index and b_scale_byte_index select byte 0..3 from the packed 32-bit E8M0 packed_scale_word_a and packed_scale_word_b inputs.

Was this page helpful?

Thank you! We'll create more content like this.

Thank you for helping us improve!

😔 What went wrong?

Some code doesn’t work

It includes inaccurate information

It's missing information I need

It was difficult to understand

Other

Submit