Glossaries¶
Keyword |
Description |
---|---|
ACK |
Acknowledge |
AM |
Active Message |
AMO |
Atomic Memory Operation |
ANL |
Argonne National Laboratory |
AZP |
Azure Pipeline |
bcopy |
Buffered copy |
Bistro |
Binary Instrumentation |
BTL |
Open MPI’s Byte Transfer Layer |
CM |
Connection Manager |
CMA |
Linux Kernel Cross Memory Attach |
CQ |
Completion Queue (InfiniBand) |
CQE |
Completion Queue Entry (InfiniBand) |
csmock |
static analysis tools |
CUDA |
Compute Unified Device Architecture (NVIDIA) |
DC |
Dynamically Connected transport (Mellanox/Nvidia InfiniBand) |
DPM |
Direct Packet Mode |
EP |
EndPoint |
FC |
Flow Control |
FD |
File Descriptor |
GDR |
GPUDirect RDMA |
gtest |
Google Test |
HPC |
High Performance Computing |
HWTM |
HardWare Tag Matching |
IB |
InfiniBand |
iface |
Interaface |
IPC |
Inter Process Communication |
JUCX |
Java API over UCP |
KLM |
A new sophisticated way of creating memory regions (Mellanox) |
KNEM |
Inria’s Kernel module for process to process zero copy (https://knem.gitlabpages.inria.fr) |
LDPM |
Legacy DPM |
LLNL |
Lawrence Livermore National Laboratory |
madvise |
give advice about use of memory. See madvise(2) |
MD |
Memory Domain |
MEMH |
Memory Handle |
MLX |
Mellanox Technologies |
mlx5 |
Mellanox-5 generation of Mellanox device driver |
MM |
Memory Mapper |
MP-XRQ |
Multi-Packet XRQ |
MPI |
Message Passing INterface |
MPICH |
ANL MPI Implementation |
MTT |
The MPI Testing Tool |
NAK |
Negative Acknowledge |
NB |
Non blocking operations |
NBI |
Non blocking implicit |
NBX |
Non blocking extended |
ODP |
OnDemand Paging |
OFA |
OpenFabrics Alliance |
OMPI |
OpenMPI |
OOB |
Out of band / Out of box |
OOO |
Out of Order |
OPA |
Omni-Path Architecture |
Open MPI |
A MPI Implementation |
ORNL |
Oak Ridge National Laboratory |
PCIe |
PCI Express |
PFN |
Physical Frame Number |
PGAS |
Partitioned Global Address Space |
POSIX |
Portable operating system interface |
ppn |
processes per node |
PR |
Pull Request |
QP |
Queue Pair (InfiniBand) |
RC |
Reliable Connection (InfiniBand) |
rcache |
Registration Cache |
RDMA |
Remote Direct Memory Access |
REQ |
Request |
rkey |
Remote KEY |
RMA |
Remote Memory Access |
RNDV |
Rendezvous protocol |
RnR |
Receiver Not Ready |
RoCE |
RDMA over Converged Ethernet |
ROCm |
Radeon Open Compute platform (AMD) |
RTE |
Run Time Environment |
RX |
Receive |
skb |
Socket Buffer |
SM |
Shared Memory |
SM |
Subnet Manager (InfiniBand) |
SockCM |
Socket Connection Manager |
SRQ |
Shared Receive Queue |
SysV |
UNIX System V |
TL |
Transport Layer |
TLS |
Transpot LayerS |
TM |
Tag Matching |
TX |
Transmit |
UC |
Unreliable Connection (InfiniBand) |
UCC |
Unified Collective Communications |
UCD |
Unified Communication Datatypes |
UCF |
Unified Communication Framework |
UCM |
Unified Communication Memory |
UCP |
Unified Communication Protocols Higher level API |
UCS |
Unified Communication Service Common utilities |
UCT |
Unified Communication Transport Lower level API |
UCX |
Unified Communication X |
UD |
Unreliable Datagram (InfiniBand) |
uGNI |
user level generic network interface (Cray) |
UMR |
User mode memory registration |
VPI |
Virtual Protocol Interconnect |
WFE |
Wait for event |
WIP-DNM |
Work-in-Progress-Do-Not-Merge |
WQ |
Work Queue (InfiniBand) |
WQE |
Work Queue Elements (pronounce WOOKIE) |
WR |
Work Request |
XPMEM |
XPMEM Linux kernel module originally developed by SGI implementing cross process memory mapping with zero copy |
XRC |
eXtended Reliable Connection (InfiniBand) |
XRCD |
XRC Domains |
XRQ |
Mellanox Shared Receive Queue |
Zcopy |
Zero Copy |