triton.experimental.gluon.language.amd.cdna4.load_shared_fp4_repacked

triton.experimental.gluon.language.amd.cdna4.load_shared_fp4_repacked(mem_desc, layout, _semantic=None)

Load M/N-packed fp4 bytes from shared memory into a K-packed MFMA dot operand layout.

The source shared memory descriptor must contain int8 or uint8 packed fp4 values. The destination shape is inferred from the source shape and dot operand index in layout.