Please enable JavaScript.
Coggle requires JavaScript to display documents.
Parallel Processors from Client to Cloud ("Vector" vs Scalar…
Parallel Processors from Client to Cloud
Data Level Parallelism
SISD(Single Instruction, Single Data)
MIMD
SIMD
SPMD(Single Program, Multi Data)
"Vector" vs Scalar
great deal of work
do not have to check for data hazards within a vector instruction
reputation => easy (MIMD)
reduce hazard checking => save energy&time
pattern
behavior predetermined => nonexistence of control hazard
savings in instruction
Hardware Multithreading
fine-grained multithreading
switching between threads a er every instruction
coarse-grainde multithreading
switching between threads only after significant events, such as a last-level cache miss
SMT(simultaneous multithreading)
utilizing the resources needed for multiple issue, dynamically scheduled microarchitecture
Multicore and Other Shared Memory
Multiprocessors
uniform memory access (UMA)
atency to any word in main memory is about the same no matter which processor requests the access
nonuniform memory access (NUMA)
some memory accesses are much faster than others depending on which processor asks for which word
Graphics Processing Units
NVDIA...
Message-Passing
Multiprocessors
Multiprocessor Network Topologies
network
bandwidth
bisection bandwidth
fully connected network
multistage network
crossbar network