cp.async.bulk

PTX ISA: cp.async.bulk

Implementation notes

NOTE. Both srcMem and dstMem must be 16-byte aligned, and size must be a multiple of 16.

Changelog

In earlier versions, cp_async_bulk_multicast was enabled for SM_90. This has been changed to SM_90a.

Unicast

cp.async.bulk.shared::cluster.global.mbarrier::complete_tx::bytes

// cp.async.bulk.dst.src.mbarrier::complete_tx::bytes [dstMem], [srcMem], size, [smem_bar]; // 1a. unicast PTX ISA 80, SM_90
// .dst       = { .shared::cluster }
// .src       = { .global }
template <typename=void>
__device__ static inline void cp_async_bulk(
  cuda::ptx::space_cluster_t,
  cuda::ptx::space_global_t,
  void* dstMem,
  const void* srcMem,
  const uint32_t& size,
  uint64_t* smem_bar);

cp.async.bulk.shared::cluster.shared::cta.mbarrier::complete_tx::bytes

// cp.async.bulk.dst.src.mbarrier::complete_tx::bytes [dstMem], [srcMem], size, [rdsmem_bar]; // 2.  PTX ISA 80, SM_90
// .dst       = { .shared::cluster }
// .src       = { .shared::cta }
template <typename=void>
__device__ static inline void cp_async_bulk(
  cuda::ptx::space_cluster_t,
  cuda::ptx::space_shared_t,
  void* dstMem,
  const void* srcMem,
  const uint32_t& size,
  uint64_t* rdsmem_bar);

cp.async.bulk.global.shared::cta.bulk_group

// cp.async.bulk.dst.src.bulk_group [dstMem], [srcMem], size; // 3.  PTX ISA 80, SM_90
// .dst       = { .global }
// .src       = { .shared::cta }
template <typename=void>
__device__ static inline void cp_async_bulk(
  cuda::ptx::space_global_t,
  cuda::ptx::space_shared_t,
  void* dstMem,
  const void* srcMem,
  const uint32_t& size);

Multicast

cp.async.bulk.shared::cluster.global.mbarrier::complete_tx::bytes.multicast::cluster

// cp.async.bulk{.dst}{.src}.mbarrier::complete_tx::bytes.multicast::cluster [dstMem], [srcMem], size, [smem_bar], ctaMask; // 1.  PTX ISA 80, SM_90a
// .dst       = { .shared::cluster }
// .src       = { .global }
template <typename=void>
__device__ static inline void cp_async_bulk(
  cuda::ptx::space_cluster_t,
  cuda::ptx::space_global_t,
  void* dstMem,
  const void* srcMem,
  const uint32_t& size,
  uint64_t* smem_bar,
  const uint16_t& ctaMask);