Half-precision floating-point format
In computing, half precision (sometimes called FP16) is a binary floating-point computer number format that occupies 16 bits (two bytes in modern computers) in computer memory. It is intended for storage of floating-point values in applications where higher precision is not essential, in particular image processing and neural networks. Almost all modern uses follow the IEEE 754-2008 standard, where the 16-bit base-2 format is referred to as binary16, and the exponent uses 5 bits. This can express values in the range ±65,504, with the minimum value above 1 being 1 + 1/1024.
- Comment
- enIn computing, half precision (sometimes called FP16) is a binary floating-point computer number format that occupies 16 bits (two bytes in modern computers) in computer memory. It is intended for storage of floating-point values in applications where higher precision is not essential, in particular image processing and neural networks. Almost all modern uses follow the IEEE 754-2008 standard, where the 16-bit base-2 format is referred to as binary16, and the exponent uses 5 bits. This can express values in the range ±65,504, with the minimum value above 1 being 1 + 1/1024.
- Date
- enJanuary 2021
- Depiction
- Find
- enHalf-precision floating-point
- Has abstract
- enIn computing, half precision (sometimes called FP16) is a binary floating-point computer number format that occupies 16 bits (two bytes in modern computers) in computer memory. It is intended for storage of floating-point values in applications where higher precision is not essential, in particular image processing and neural networks. Almost all modern uses follow the IEEE 754-2008 standard, where the 16-bit base-2 format is referred to as binary16, and the exponent uses 5 bits. This can express values in the range ±65,504, with the minimum value above 1 being 1 + 1/1024. Depending on the computer, half-precision can be over an order of magnitude faster than double precision, e.g. 550 PFLOPS for half-precision vs 37 PFLOPS for double precision on one cloud provider.
- Hypernym
- Format
- Is primary topic of
- Half-precision floating-point format
- Label
- enHalf-precision floating-point format
- Link from a Wikipage to an external page
- www.mathworks.com/matlabcentral/fileexchange/23173
- stackoverflow.com/a/6162687/237321
- web.archive.org/web/20090419063321/http:/www.analog.com/static/imported-files/processor_manuals/ADSP_2136x_PGR_rev1-1.pdf
- technet.microsoft.com/en-us/library/bb147247(v=vs.85).aspx
- www.khronos.org/registry/DataFormat/specs/1.2/dataformat.1.2.html%2316bitfp
- www.fox-toolkit.org/ftp/fasthalffloatconversion.pdf
- www.mrob.com/pub/math/floatformats.html%23minifloat
- web.archive.org/web/20170531074746/http:/oss.sgi.com/projects/ogl-sample/registry/ARB/half_float_pixel.txt
- gcc.gnu.org/onlinedocs/gcc/Half-Precision.html
- www.openexr.org/
- Link from a Wikipage to another Wikipage
- −0
- 0 (number)
- 16 bit
- 3dfx Interactive
- ARM architecture family
- AVX-512
- Bandwidth (computers)
- Bfloat16 floating-point format
- Binary (computing)
- Category:Binary arithmetic
- Category:Floating point types
- Cg (programming language)
- Computer graphics
- Computer memory
- Computer number format
- Computing
- Control register
- D3DX
- Datatype
- Direct3D
- Double precision
- Dynamic range
- Exponent
- Exponent bias
- F16C
- File:IEEE 754r Half Floating Point Format.svg
- Floating-point
- GeForce FX
- GIMP
- IEC 10967
- IEEE 754
- IEEE 754-2008
- Image processing
- Industrial Light & Magic
- Infinity
- Instruction set architecture
- JPEG XR
- Machine learning
- MATLAB
- Microsoft
- Minifloat
- NaN
- Neural network
- Neural networks
- Nvidia
- Offset-binary
- OpenEXR
- OpenGL
- Power Management Bus
- Precision (arithmetic)
- Primitive data type
- Radix
- RGBE image format
- Sapphire Rapids
- Shadow
- SIGGRAPH
- Sign bit
- Significand
- Silicon Graphics
- SIMD
- Single-precision
- Subnormal numbers
- Unit in the last place
- X86
- SameAs
- m.06w17nf
- Mitja precisió
- Q1994657
- ucQw
- Число половинной точности
- Число половинної точності
- 半精度浮動小数点数
- 半精度浮点数
- Subject
- Category:Binary arithmetic
- Category:Floating point types
- Thumbnail
- WasDerivedFrom
- Half-precision floating-point format?oldid=1103245465&ns=0
- WikiPageLength
- 16160
- Wikipage page ID
- 23552810
- Wikipage revision ID
- 1103245465
- WikiPageUsesTemplate
- Template:Citation needed
- Template:Data types
- Template:Distinguish
- Template:External links
- Template:Floating-point
- Template:Math
- Template:Mono
- Template:Reflist
- Template:Sfrac
- Template:Short description
- Template:Unreferenced section
