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

tcgen05_cp

def tcgen05_cp[*, cta_group: Int32, datapaths: Int, bits: Int, src_fmt: String = "", dst_fmt: String = "", multicast: String = ""](tmem_addr: UInt32, s_desc: MMASmemDescriptor)

Copies data from shared memory described by the matrix descriptor s_desc to tensor memory tmem_addr.

Note: This function is only available on NVIDIA Blackwell GPUs (SM 100+).

Parameters:

cta_group (Int32): The cooperative thread array (CTA) group ID.
datapaths (Int): The first dimension of the shape.
bits (Int): The second dimension of the shape.
src_fmt (String): Source format string.
dst_fmt (String): Destination format string.
multicast (String): Multicast string.

Args:

tmem_addr (UInt32): Address of the tensor memory.
s_desc (MMASmemDescriptor): Matrix descriptor for the copy operation.