VLIW

E904642

VLIW (Very Long Instruction Word) is a computer processor architecture that executes multiple operations in parallel by encoding them into a single long instruction word, relying heavily on compiler optimization for performance.

Jump to: Statements Referenced by

Statements (49)

Predicate Object
instanceOf computer processor architecture
instruction set architecture paradigm
abbreviationOf Very Long Instruction Word NERFINISHED
advantage lower power consumption potential
potentially high performance on regular workloads
reduced hardware complexity compared to superscalar
simpler control logic
contrastsWith dynamic scheduling in hardware
out-of-order execution
superscalar architecture
designGoal exploit parallelism at compile time
shift scheduling complexity to the compiler
simplify hardware complexity
disadvantage binary compatibility issues across implementations
code size increase due to long instruction words
difficulty handling irregular control flow
heavy dependence on compiler quality
sensitivity to pipeline stalls
enables instruction-level parallelism
encodes multiple operations into a single long instruction word
executes multiple operations in parallel
fullName Very Long Instruction Word NERFINISHED
hasCharacteristic exposes functional units directly to the compiler
fixed-width instruction bundles
limited hardware hazard detection
multiple operation fields per instruction word
simple instruction dispatch hardware
statically scheduled instruction issue
historicalContext developed by Josh Fisher and colleagues at Yale and Multiflow
proposed in the 1980s
influenced EPIC architecture NERFINISHED
Itanium architecture NERFINISHED
parallelismType instruction-level parallelism
relatedTo EPIC (Explicitly Parallel Instruction Computing) NERFINISHED
Very Long Instruction Word DSPs NERFINISHED
reliesOn compiler optimization
static scheduling by the compiler
requires advanced compiler technology
aggressive instruction scheduling
software-based dependency analysis
software-based speculation and predication support
schedulingType static scheduling
typicalFeature multiple functional units per core
predicated execution support
software pipelining support
usedIn digital signal processors
embedded systems
media processing processors
some general-purpose CPUs

Referenced by (1)

Full triples — surface form annotated when it differs from this entity's canonical label.

Crusoe architecture VLIW