Skip to content

[RFC/WIP] unify the concepts of Fixed and UFixed #32

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 4 commits into from
Closed

[RFC/WIP] unify the concepts of Fixed and UFixed #32

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
8636937
unify the concepts of Fixed and UFixed
vchuravy Oct 15, 2015
80f986f
special case some methods for Unsigned
vchuravy Oct 19, 2015
de63ce3
rename reinterpret to getindex
vchuravy Oct 22, 2015
ee87107
rounding fixes
vchuravy Oct 22, 2015
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
250 changes: 228 additions & 22 deletions src/FixedPointNumbers.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
VERSION >= v"0.4.0-dev+6521" && __precompile__()
# ERSION >= v"0.4.0-dev+6521" && __precompile__()
Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

oops :)


module FixedPointNumbers

Expand All @@ -9,13 +9,10 @@ using Base: IdFun, AddFun, MulFun, reducedim_initarray
import Base: ==, <, <=, -, +, *, /, ~,
convert, promote_rule, show, showcompact, isinteger, abs, decompose,
isnan, isinf, isfinite,
zero, one, typemin, typemax, realmin, realmax, eps, sizeof, reinterpret,
zero, one, typemin, typemax, realmin, realmax, eps, sizeof, reinterpret, getindex,
trunc, round, floor, ceil, bswap,
div, fld, rem, mod, mod1, rem1, fld1, min, max,
start, next, done, r_promote, reducedim_init
# T => BaseType
# f => Number of Bytes reserved for fractional part
abstract FixedPoint{T <: Integer, f} <: Real

export
FixedPoint,
Expand All @@ -33,6 +30,7 @@ export
ufixed12,
ufixed14,
ufixed16,
fixed16,
# literal constructor constants
uf8,
uf10,
Expand All @@ -42,25 +40,213 @@ export
# Functions
scaledual

reinterpret(x::FixedPoint) = x.i
# T => BaseType
# f => Number of Bytes reserved for fractional part
immutable FixedPoint{T <: Integer, f} <: Real
i::T

# constructor for manipulating the representation;
# selected by passing an extra dummy argument
FixedPoint(i::T, _) = new(i)
FixedPoint(i::Integer,_) = new(i % T)
Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

i % T is a new change and maybe it is better to throw a DomainError, at least we should add a test for this


FixedPoint(x) = convert(FixedPoint{T,f}, x)
end

# comparison
=={T <: FixedPoint}(x::T, y::T) = x.i == y.i
<{T <: FixedPoint}(x::T, y::T) = x.i < y.i
<={T <: FixedPoint}(x::T, y::T) = x.i <= y.i
# basic typealiases for signed and unsigned
typealias Fixed{T <: Signed, f} FixedPoint{T, f}
typealias UFixed{T <: Unsigned, f} FixedPoint{T, f}

# default provided typealiases
typealias Fixed16 Fixed{Int32, 16}
typealias UFixed8 UFixed{UInt8,8}
typealias UFixed10 UFixed{UInt16,10}
typealias UFixed12 UFixed{UInt16,12}
typealias UFixed14 UFixed{UInt16,14}
typealias UFixed16 UFixed{UInt16,16}

getindex(x::FixedPoint) = x.i
reinterpret{T,f}(::Type{T}, x::FixedPoint{T,f}) = x[]
reinterpret{T <: Integer,f}(::Type{FixedPoint{T,f}}, x::T) = FixedPoint{T,f}(x, 0)

rawtype{T,f}(::Type{FixedPoint{T,f}}) = T
nbitsfrac{T,f}(::Type{FixedPoint{T,f}}) = f
rawone(v) = one(v)[]

# comparisons
=={T <: FixedPoint}(x::T, y::T) = x[] == y[]
<{T <: FixedPoint}(x::T, y::T) = x[] < y[]
<={T <: FixedPoint}(x::T, y::T) = x[] <= y[]

# predicates
isinteger{T,f}(x::FixedPoint{T,f}) = (x.i&(1<<f-1)) == 0
isinteger{T,f}(x::FixedPoint{T,f}) = (x[] & (1<<f-1)) == 0
isfinite(x::FixedPoint) = true
isnan(x::FixedPoint) = false
isinf(x::FixedPoint) = false

# traits
typemax{T<: FixedPoint}(::Type{T}) = T(typemax(rawtype(T)), 0)
typemin{T<: FixedPoint}(::Type{T}) = T(typemin(rawtype(T)), 0)
realmin{T<: FixedPoint}(::Type{T}) = typemin(T)
realmax{T<: FixedPoint}(::Type{T}) = typemax(T)
eps{T<:FixedPoint}(::Type{T}) = T(one(rawtype(T)),0)
eps{T<:FixedPoint}(::T) = eps(T)
sizeof{T<:FixedPoint}(::Type{T}) = sizeof(rawtype(T))

include("fixed.jl")
include("ufixed.jl")
include("deprecations.jl")
zero{T <: FixedPoint}(::Type{T}) = T(zero(rawtype(T)),0)
one{T <: Unsigned, f}(::Type{FixedPoint{T, f}}) = FixedPoint{T, f}(2^f-1,0)
function one{T <: Signed,f}(::Type{FixedPoint{T, f}})
if sizeof(T) * 8 > f
return FixedPoint{T, f}(2^f-1,0)
else
throw(DomainError())
end
end

# basic operators & arithmetics
(-){T<:FixedPoint}(x::T) = T(-x[], 0)
(~){T<:FixedPoint}(x::T) = T(~x[], 0)
abs{T<:FixedPoint}(x::T) = T(abs(x[]),0)

+{T<:FixedPoint}(x::T, y::T) = T(x[] + y[],0)
-{T<:FixedPoint}(x::T, y::T) = T(x[] - y[],0)

# with truncation:
# *{T<:FixedPoint}(x::T, y::T) =
# T(Base.widemul(x[],y[]) >> nbitsfrac(T),0)
# with rounding up:
function *{T<:FixedPoint}(x::T, y::T)
f = nbitsfrac(T)
Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

A was wondering if this is getting optimized out or if we rather should use @generated for functions depending on f = nbitsfrac(...) and T = rawtype(...)

i = Base.widemul(x[],y[])
T((i + convert(widen(rawtype(T)), 1) << (f-1) )>>f,0)
end

function /{T<:FixedPoint}(x::T, y::T)
f = nbitsfrac(T)
T(div(convert(widen(rawtype(T)), x[]) << f, y.i), 0)
end

# Conversions to FixedPoint
convert{T,f}(::Type{FixedPoint{T,f}}, x::Integer) =
FixedPoint{T,f}(convert(T,x)<<f,0)
convert{T,f}(::Type{FixedPoint{T,f}}, x::AbstractFloat) =
FixedPoint{T,f}(trunc(T,x) << f + round(T, rem(x,1)*(1<<f)),0)
convert{T,f}(::Type{Fixed{T,f}}, x::Rational) =
FixedPoint{T,f}(x.num)/FixedPoint{T,f}(x.den)

# Conversions from FixedPoint
convert{T <: Signed,f}(::Type{BigFloat}, x::FixedPoint{T,f}) =
convert(BigFloat,x[]>>f) + convert(BigFloat,x[]&(1<<f - 1))/convert(BigFloat,1<<f)
convert{T <: Unsigned, f}(::Type{BigFloat}, x::FixedPoint{T, f}) =
x[]*(1/BigFloat(rawone(x)))

convert{TF<:AbstractFloat,T <: Signed,f}(::Type{TF}, x::FixedPoint{T,f}) =
convert(TF,x[]>>f) + convert(TF,x[]&(1<<f - 1))/convert(TF,1<<f)
convert{TF<:AbstractFloat,T <: Unsigned,f}(::Type{TF}, x::FixedPoint{T,f}) =
x[]*(1/convert(T, rawone(x)))

convert{T,f}(::Type{Bool}, x::FixedPoint{T,f}) = x[] != zero(x)

function convert{TI<:Integer, T,f}(::Type{TI}, x::FixedPoint{T,f})
isinteger(x) || throw(InexactError())
convert(TI, x[]>>f)
end

convert{TR<:Rational,T <: Signed,f}(::Type{TR}, x::FixedPoint{T,f}) =
convert(TR, x[] >> f + (x[] & (1 << f - 1)) // (1 << f))
convert{Ti<:Integer}(::Type{Rational{Ti}}, x::UFixed) = convert(Ti, x[])//convert(Ti, rawone(x))
convert(::Type{Rational}, x::UFixed) = x[]//rawone(x)

# Special conversions for constructors
convert{T<:FixedPoint}(::Type{T}, x::T) = x
convert{T1<:FixedPoint}(::Type{T1}, x::FixedPoint) = reinterpret(T1, round(rawtype(T1), (rawone(T1)/rawone(x))*x[]))
convert(::Type{UFixed16}, x::UFixed8) = reinterpret(UFixed16, convert(UInt16, 0x0101*x[]))
convert{T<:FixedPoint}(::Type{T}, x::Real) = T(round(rawtype(T), rawone(T)*x),0)

# Constructors
ufixed8(x) = convert(UFixed8, x)
ufixed10(x) = convert(UFixed10, x)
ufixed12(x) = convert(UFixed12, x)
ufixed14(x) = convert(UFixed14, x)
ufixed16(x) = convert(UFixed16, x)
fixed16(x) = convert(Fixed16, x)

@vectorize_1arg Real ufixed8
@vectorize_1arg Real ufixed10
@vectorize_1arg Real ufixed12
@vectorize_1arg Real ufixed14
@vectorize_1arg Real ufixed16
@vectorize_1arg Real fixed16

# Promote rules
promote_rule{T <: FixedPoint,TI<:Integer}(::Type{T}, ::Type{TI}) = T
promote_rule{T <: FixedPoint,TF<:AbstractFloat}(::Type{T}, ::Type{TF}) = TF
promote_rule{T <: FixedPoint,TR <: Rational}(::Type{T}, ::Type{TR}) = TR

# for T in UF
# for Ti in (Int8, UInt8, Int16, UInt16, Int32, UInt32, Int64, UInt64)
# Tp = eps(convert(Float32, typemax(Ti))) > eps(T) ? Float64 : Float32
# @eval begin
# promote_rule(::Type{$T}, ::Type{$Ti}) = $Tp
Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Are these promote_rules needed or do the three on top cover them?

# end
# end
# end

# Math functions
# Round towards negative infinity
trunc{T<:FixedPoint}(x::T) = T(x[] & ~(1 << nbitsfrac(T) - 1), 0)
# Round towards negative infinity
floor{T<:FixedPoint}(x::T) = trunc(x)
# Round towards positive infinity
ceil{T<:FixedPoint}(x::T) = trunc(T(x[] + 1 << (nbitsfrac(T)-1), 0))
# Round towards even
function round{T<:FixedPoint}(x::T)
even = x[] & (1 << nbitsfrac(T)) == 0
if even
return floor(x)
else
return ceil(x)
end
end

trunc{TI<:Integer, T <: FixedPoint}(::Type{TI}, x::T) =
convert(TI, x[] >> nbitsfrac(T))
floor{T<:Integer}(::Type{T}, x::FixedPoint) = trunc(T, x)
ceil{T<:Integer}(::Type{T}, x::FixedPoint) = trunc(T, ceil(x))
round{T<:Integer}(::Type{T}, x::FixedPoint) = trunc(T, round(x))

# for T in UF
# f = nbitsfrac(T)
# R = rawtype(T)
# roundmask = convert(R, 1<<(f-1))
# k = 8*sizeof(R)-f
# ceilmask = (typemax(R)<<k)>>k
# @eval begin
# round(x::$T) = (y = trunc(x); return convert(rawtype($T), x[]-y[])&$roundmask>0 ? $T(y+one($T)) : y)
# ceil(x::$T) = (y = trunc(x); return convert(rawtype($T), x[]-y[])&$ceilmask >0 ? $T(y+one($T)) : y)
# end
# end

for f in (:div, :fld, :rem, :mod, :mod1, :rem1, :fld1, :min, :max)
@eval begin
$f{T<:FixedPoint}(x::T, y::T) = T($f(x[],y[]),0)
end
end

function minmax{T<:UFixed}(x::T, y::T)
a, b = minmax(x[], y[])
T(a,0), T(b,0)
end

# Special function
decompose{T,f}(x::FixedPoint{T,f}) = x[], -f, 1
# function decompose(x::UFixed)
# g = gcd(x[], rawone(x))
# div(x[],g), 0, div(rawone(x),g)
# end

bswap{T <: Union{UInt8, Int8}, f}(x::FixedPoint{T,f}) = x
bswap{T <: FixedPoint}(x::T) = T(bswap(x[]),0)

# Promotions for reductions
const Treduce = Float64
Expand All @@ -75,14 +261,6 @@ reducedim_init{T<:FixedPoint}(f::IdFun, op::MulFun,
A::AbstractArray{T}, region) =
reducedim_initarray(A, region, one(Treduce))

# TODO: rewrite this by @generated
for T in tuple(Fixed16, UF...)
R = rawtype(T)
@eval begin
reinterpret(::Type{$R}, x::$T) = x.i
end
end

# When multiplying by a float, reduce two multiplies to one.
# Particularly useful for arrays.
scaledual(Tdual::Type, x) = one(Tdual), x
Expand All @@ -92,4 +270,32 @@ scaledual{Tdual<:Number}(b::Tdual, x) = b, x
@compat scaledual{Tdual<:Number, T<:FixedPoint}(b::Tdual, x::Union{T,AbstractArray{T}}) =
convert(Tdual, b/one(T)), reinterpret(rawtype(T), x)

# Show
function show(io::IO, x::FixedPoint)
print(io, typeof(x))
print(io, "(")
showcompact(io, x)
print(io, ")")
end

const _log2_10 = 3.321928094887362
showcompact{T,f}(io::IO, x::FixedPoint{T,f}) = show(io, round(convert(Float64,x), ceil(Int,f/_log2_10)))

# Iteration
# The main subtlety here is that iterating over 0x00uf8:0xffuf8 will wrap around
# unless we iterate using a wider type
start{T<:FixedPoint}(r::StepRange{T}) = convert(typeof(r.start[] + r.step[]), r.start[])
next{T<:FixedPoint}(r::StepRange{T}, i::Integer) = (T(i,0), i+r.step[])
done{T<:FixedPoint}(r::StepRange{T}, i::Integer) = isempty(r) || (i > r.stop[])

immutable UFixedConstructor{T,f} end
*{T,f}(n::Integer, ::UFixedConstructor{T,f}) = UFixed{T,f}(n,0)
const uf8 = UFixedConstructor{UInt8,8}()
const uf10 = UFixedConstructor{UInt16,10}()
const uf12 = UFixedConstructor{UInt16,12}()
const uf14 = UFixedConstructor{UInt16,14}()
const uf16 = UFixedConstructor{UInt16,16}()

include("deprecations.jl")

end # module
3 changes: 2 additions & 1 deletion src/deprecations.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,4 +10,5 @@ import Base.@deprecate_binding
@deprecate_binding Ufixed16 UFixed16

@deprecate_binding Fixed32 Fixed16
@deprecate Fixed(x::Real) convert(Fixed{Int32, 16}, x)

@deprecate reinterpret(x::FixedPoint) getindex(x)
66 changes: 0 additions & 66 deletions src/fixed.jl
View file
Open in desktop

This file was deleted.

Loading