iw_cxgb4: Support FW write completion WR
To optimize NVME-oF READ IOPs, use a specialized WQE that combines the RDMA WRITE and SEND_INV WR chain submitted by the NVME-oF target driver. This reduces uP overhead per NVME-oF IO, and results in over 10% improvement in NVME-oF 4K READ IOPs. Signed-off-by:Potnuri Bharat Teja <bharat@chelsio.com> Signed-off-by:
Steve Wise <swise@opengridcomputing.com> Signed-off-by:
Jason Gunthorpe <jgg@mellanox.com>
Showing
- drivers/infiniband/hw/cxgb4/device.c 1 addition, 0 deletionsdrivers/infiniband/hw/cxgb4/device.c
- drivers/infiniband/hw/cxgb4/qp.c 146 additions, 1 deletiondrivers/infiniband/hw/cxgb4/qp.c
- drivers/infiniband/hw/cxgb4/t4.h 5 additions, 1 deletiondrivers/infiniband/hw/cxgb4/t4.h
- drivers/infiniband/hw/cxgb4/t4fw_ri_api.h 31 additions, 0 deletionsdrivers/infiniband/hw/cxgb4/t4fw_ri_api.h
Loading
Please register or sign in to comment