AVX
E640423
AVX is an x86 instruction set extension that introduces advanced vector processing capabilities to accelerate floating-point and parallel computations.
Statements (46)
| Predicate | Object |
|---|---|
| instanceOf |
SIMD instruction set
ⓘ
instruction set extension ⓘ |
| abbreviation | AVX NERFINISHED ⓘ |
| alsoSupportedBy | AMD NERFINISHED ⓘ |
| architectureFamily |
x86
ⓘ
x86-64 ⓘ |
| backwardCompatibleWith | SSE instructions NERFINISHED ⓘ |
| belongsTo | Intel AVX family NERFINISHED ⓘ |
| category | CPU instruction set ⓘ |
| contextSwitchRequirement | extended state management via XSAVE/XRSTOR ⓘ |
| dataTypeSupport |
double-precision floating-point
ⓘ
single-precision floating-point ⓘ |
| designGoal |
enable more efficient compiler vectorization
ⓘ
increase vector width without increasing instruction length excessively ⓘ |
| encoding | VEX prefix ⓘ |
| extendsRegister | XMM registers to 256-bit YMM registers ⓘ |
| feature | 256-bit wide vector registers ⓘ |
| firstSupportedInMicroarchitecture |
AMD Bulldozer
NERFINISHED
ⓘ
Intel Ivy Bridge NERFINISHED ⓘ Intel Sandy Bridge NERFINISHED ⓘ |
| fullName | Advanced Vector Extensions NERFINISHED ⓘ |
| improves |
data-level parallelism
ⓘ
floating-point throughput ⓘ |
| improvesOver | SSE NERFINISHED ⓘ |
| introducedBy | Intel NERFINISHED ⓘ |
| keyFeature |
VEX prefix encoding
ⓘ
non-destructive destination operand ⓘ three-operand instruction format ⓘ |
| purpose |
accelerate floating-point computations
ⓘ
accelerate parallel computations ⓘ support vector processing ⓘ |
| registerWidth | 256 bits ⓘ |
| requires | operating system support for YMM state saving ⓘ |
| standardizedIn | Intel 64 and IA-32 Architectures Software Developer’s Manual NERFINISHED ⓘ |
| successor |
AVX-512
NERFINISHED
ⓘ
AVX2 NERFINISHED ⓘ |
| supports |
SIMD operations
ⓘ
vectorized arithmetic ⓘ vectorized data movement ⓘ vectorized logical operations ⓘ |
| typicalUseCase |
3D graphics
ⓘ
high-performance computing ⓘ media processing ⓘ scientific computing ⓘ signal processing ⓘ |
| vendorFeatureFlagName | AVX ⓘ |
Referenced by (5)
Full triples — surface form annotated when it differs from this entity's canonical label.