AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

GDA get* APIs (#243)

Browse Source 
* feat(GDA): add `get*` and `get*_nbi` APIs for mlx and bnxt NICs
   - implemented thread, wave and wg variants of `get*` and `get_nbi`.

* test(GDA): enable functional tests for `get*` and `get*_nbi` APIs
This commit is contained in:
Avinash Kethineedi
2025-09-10 12:24:53 -04:00
committed by GitHub
parent 4a760d741a
commit 671f8187f4
5 changed files with 91 additions and 44 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/gda/queue_pair.hpp
+11
View File
@@ -61,6 +61,7 @@ extern "C" {
#elif defined(GDA_MLX5)
#define GDA_MAX_ATOMIC 1
#define GDA_OP_RDMA_WRITE MLX5_OPCODE_RDMA_WRITE
#define GDA_OP_RDMA_READ MLX5_OPCODE_RDMA_READ
#define GDA_OP_ATOMIC_FA MLX5_OPCODE_ATOMIC_FA
#define GDA_OP_ATOMIC_CS MLX5_OPCODE_ATOMIC_CS
#endif
@@ -102,6 +103,16 @@ class QueuePair {
*/
__device__ void put_nbi(void *dest, const void *source, size_t nelems, int pe);
/**
* @brief Create and enqueue a non-blocking get work queue entry (wqe).
*
* @param[in] dest Destination address for data transmission.
* @param[in] source Source address for data transmission.
* @param[in] nelems Size in bytes of data transmission.
* @param[in] pe Destination processing element of data transmission.
*/
__device__ void get_nbi(void *dest, const void *source, size_t nelems, int pe);
/**
* @brief Empty all completions from the completion queue.
*/