# ARM Contex-A9 NEON/VFP
Introduction
Xilinx Zynq-7000 AP SoC internal integration of a dual-core processor ARM Contex-A9. ARM Contex-A9 ARMv7 processor architecture can be seen from the figure, the architecture built SIMD (Single Instruction Multiple Data) and floating point unit NEON engine VFPv3.
NEON data types supported are:
- Unsigned integer U8 U16 U32 U64
- Signed integer S8 S16 S32 S64
- Integer of unspecified type I8 I16 I32 I64
- Floating-point number F16 F32
- Polynomial over {0,1} P8
VFPv3 support for single precision floating point and double-precision floating point
NEON and VFPv3 common register group, and the instruction set is integrated, are beginning V: V {<mod>} <op> {<shape>} {<cond>} (<dest> {<dt>.} }, src1, src2
NEON vs VFP
NEON works only in vector mode, but you can not VFP SIMD, despite the "vector" in its name.
For floating-point, NEON only supports single-precision floating-point, and VFPv3 supports both single-precision floating-point, and also supports double precision floating point. BTW, ARMv8 of NEON also support double precision floating point.
VFP NEON before the appearance earlier, mainly in order to complete the floating point arithmetic, although termed vector, but if actually SIMD operation can not be completed, the data is sequentially processed. BTW, Contex-A9 in VFP realized the pipeline.
VFP is a legacy technology, changes with the version, follow-up may be discarded (nonsense).