eliminate intrinsics

NEWS.md

@@ -215,6 +215,12 @@ Library improvements
 Compiler/Runtime improvements
 -----------------------------
 
+* `ccall` is now implemented as a macro, removing the need for special code-generator support for Intrinsics.
+
+* `ccall` gained limited support for a `llvmcall` calling-convention. This can replace many uses of `llvmcall` with a simpler, shorter declaration.
+
+* All Intrinsics are now Builtin functions instead and have proper error checking and fall-back static compilation support.
+
 Deprecated or removed
 ---------------------
 

base/Enums.jl

@@ -2,15 +2,15 @@
 
 module Enums
 
-import Core.Intrinsics.box
+import Core.Intrinsics.bitcast
 export Enum, @enum
 
 function basetype end
 
 abstract Enum{T<:Integer}
 
-Base.convert{T<:Integer}(::Type{Integer}, x::Enum{T}) = box(T, x)
-Base.convert{T<:Integer,T2<:Integer}(::Type{T}, x::Enum{T2}) = convert(T, box(T2, x))
+Base.convert{T<:Integer}(::Type{Integer}, x::Enum{T}) = bitcast(T, x)
+Base.convert{T<:Integer,T2<:Integer}(::Type{T}, x::Enum{T2}) = convert(T, bitcast(T2, x))
 Base.write{T<:Integer}(io::IO, x::Enum{T}) = write(io, T(x))
 Base.read{T<:Enum}(io::IO, ::Type{T}) = T(read(io, Enums.basetype(T)))
 
@@ -106,7 +106,7 @@ macro enum(T,syms...)
         Base.@__doc__(bitstype $(sizeof(basetype) * 8) $(esc(typename)) <: Enum{$(basetype)})
         function Base.convert(::Type{$(esc(typename))}, x::Integer)
             $(membershiptest(:x, values)) || enum_argument_error($(Expr(:quote, typename)), x)
-            box($(esc(typename)), convert($(basetype), x))
+            return bitcast($(esc(typename)), convert($(basetype), x))
         end
         Enums.basetype(::Type{$(esc(typename))}) = $(esc(basetype))
         Base.typemin(x::Type{$(esc(typename))}) = $(esc(typename))($lo)

base/base.jl

@@ -145,8 +145,8 @@ end
 finalize(o::ANY) = ccall(:jl_finalize_th, Void, (Ptr{Void}, Any,),
                          Core.getptls(), o)
 
-gc(full::Bool=true) = ccall(:jl_gc_collect, Void, (Cint,), full)
-gc_enable(on::Bool) = ccall(:jl_gc_enable, Cint, (Cint,), on)!=0
+gc(full::Bool=true) = ccall(:jl_gc_collect, Void, (Int32,), full)
+gc_enable(on::Bool) = ccall(:jl_gc_enable, Int32, (Int32,), on) != 0
 
 immutable Nullable{T}
     hasvalue::Bool

base/bool.jl

@@ -34,12 +34,12 @@ julia> ![true false true]
 function !(x::Bool)
     ## We need a better heuristic to detect this automatically
     @_pure_meta
-    return box(Bool,not_int(unbox(Bool,x)))
+    return not_int(x)
 end
 
 (~)(x::Bool) = !x
-(&)(x::Bool, y::Bool) = box(Bool,and_int(unbox(Bool,x),unbox(Bool,y)))
-(|)(x::Bool, y::Bool) = box(Bool,or_int(unbox(Bool,x),unbox(Bool,y)))
+(&)(x::Bool, y::Bool) = and_int(x, y)
+(|)(x::Bool, y::Bool) = or_int(x, y)
 
 """
     xor(x, y)
@@ -58,7 +58,7 @@ julia> [true; true; false] ⊻ [true; false; false]
  false
 ```
 """
-xor(x::Bool, y::Bool) = (x!=y)
+xor(x::Bool, y::Bool) = (x != y)
 
 >>(x::Bool, c::Unsigned) = Int(x) >> c
 <<(x::Bool, c::Unsigned) = Int(x) << c

base/boot.jl

@@ -116,8 +116,6 @@
 #     runnable::Bool
 # end
 
-import Core.Intrinsics.ccall
-
 export
     # key types
     Any, DataType, Vararg, ANY, NTuple,

base/c.jl

@@ -2,7 +2,7 @@
 
 # definitions related to C interface
 
-import Core.Intrinsics: cglobal, box
+import Core.Intrinsics: cglobal, bitcast
 
 cfunction(f::Function, r, a) = ccall(:jl_function_ptr, Ptr{Void}, (Any, Any, Any), f, r, a)
 
@@ -27,18 +27,21 @@ else
     typealias Culong UInt
     typealias Cwchar_t Int32
 end
+
 """
     Clong
 
 Equivalent to the native `signed long` c-type.
 """
 Clong
+
 """
     Culong
 
 Equivalent to the native `unsigned long` c-type.
 """
 Culong
+
 """
     Cwchar_t
 
@@ -58,13 +61,13 @@ if !is_windows()
 end
 
 # construction from typed pointers
-convert{T<:Union{Int8,UInt8}}(::Type{Cstring}, p::Ptr{T}) = box(Cstring, p)
-convert(::Type{Cwstring}, p::Ptr{Cwchar_t}) = box(Cwstring, p)
-convert{T<:Union{Int8,UInt8}}(::Type{Ptr{T}}, p::Cstring) = box(Ptr{T}, p)
-convert(::Type{Ptr{Cwchar_t}}, p::Cwstring) = box(Ptr{Cwchar_t}, p)
+convert{T<:Union{Int8,UInt8}}(::Type{Cstring}, p::Ptr{T}) = bitcast(Cstring, p)
+convert(::Type{Cwstring}, p::Ptr{Cwchar_t}) = bitcast(Cwstring, p)
+convert{T<:Union{Int8,UInt8}}(::Type{Ptr{T}}, p::Cstring) = bitcast(Ptr{T}, p)
+convert(::Type{Ptr{Cwchar_t}}, p::Cwstring) = bitcast(Ptr{Cwchar_t}, p)
 
 # construction from untyped pointers
-convert{T<:Union{Cstring,Cwstring}}(::Type{T}, p::Ptr{Void}) = box(T, p)
+convert{T<:Union{Cstring,Cwstring}}(::Type{T}, p::Ptr{Void}) = bitcast(T, p)
 
 pointer(p::Cstring) = convert(Ptr{UInt8}, p)
 pointer(p::Cwstring) = convert(Ptr{Cwchar_t}, p)

base/checked.jl

@@ -8,7 +8,7 @@ export checked_neg, checked_abs, checked_add, checked_sub, checked_mul,
        checked_div, checked_rem, checked_fld, checked_mod, checked_cld,
        add_with_overflow, sub_with_overflow, mul_with_overflow
 
-import Core.Intrinsics: box, unbox,
+import Core.Intrinsics:
        checked_sadd_int, checked_ssub_int, checked_smul_int, checked_sdiv_int,
        checked_srem_int,
        checked_uadd_int, checked_usub_int, checked_umul_int, checked_udiv_int,

base/ctypes.jl

@@ -9,78 +9,104 @@
 Equivalent to the native `unsigned char` c-type (`UInt8`).
 """
 typealias Cuchar UInt8
+
+
 """
     Cshort
 
 Equivalent to the native `signed short` c-type (`Int16`).
 """
 typealias Cshort Int16
+
+
 """
     Cushort
 
 Equivalent to the native `unsigned short` c-type (`UInt16`).
 """
 typealias Cushort UInt16
+
+
 """
     Cint
 
 Equivalent to the native `signed int` c-type (`Int32`).
 """
 typealias Cint Int32
+
+
 """
     Cuint
 
 Equivalent to the native `unsigned int` c-type (`UInt32`).
 """
 typealias Cuint UInt32
+
+
 """
     Cptrdiff_t
 
 Equivalent to the native `ptrdiff_t` c-type (`Int`).
 """
 typealias Cptrdiff_t Int
+
+
 """
     Csize_t
 
 Equivalent to the native `size_t` c-type (`UInt`).
 """
 typealias Csize_t UInt
+
+
 """
     Cssize_t
 
 Equivalent to the native `ssize_t` c-type.
 """
 typealias Cssize_t Int
+
+
 """
     Cintmax_t
 
 Equivalent to the native `intmax_t` c-type (`Int64`).
 """
 typealias Cintmax_t Int64
+
+
 """
     Cuintmax_t
 
 Equivalent to the native `uintmax_t` c-type (`UInt64`).
 """
 typealias Cuintmax_t UInt64
+
+
 """
     Clonglong
 
 Equivalent to the native `signed long long` c-type (`Int64`).
 """
 typealias Clonglong Int64
+
+
 """
     Culonglong
 
 Equivalent to the native `unsigned long long` c-type (`UInt64`).
 """
 typealias Culonglong UInt64
+
+
 """
     Cfloat
 
 Equivalent to the native `float` c-type (`Float32`).
 """
 typealias Cfloat Float32
+
+
 """
     Cdouble
 

base/deprecated.jl

@@ -1758,4 +1758,9 @@ end)
 @deprecate(SharedArray{T}(filename::AbstractString, ::Type{T}, dims::NTuple, offset; kwargs...),
            SharedArray{T,length(dims)}(filename, dims, offset; kwargs...))
 
+@noinline function is_intrinsic_expr(x::ANY)
+    Base.depwarn("is_intrinsic_expr is deprecated. There are no intrinsic functions anymore.", :is_intrinsic_expr)
+    return false
+end
+
 # End deprecations scheduled for 0.6

base/essentials.jl

@@ -115,7 +115,7 @@ unsafe_convert{T}(::Type{T}, x::T) = x # unsafe_convert (like convert) defaults
 unsafe_convert{T<:Ptr}(::Type{T}, x::T) = x  # to resolve ambiguity with the next method
 unsafe_convert{P<:Ptr}(::Type{P}, x::Ptr) = convert(P, x)
 
-reinterpret{T}(::Type{T}, x) = box(T, x)
+reinterpret{T}(::Type{T}, x) = bitcast(T, x)
 reinterpret(::Type{Unsigned}, x::Float16) = reinterpret(UInt16,x)
 reinterpret(::Type{Signed}, x::Float16) = reinterpret(Int16,x)
 
@@ -148,11 +148,11 @@ setindex!(A::Array{Any}, x::ANY, i::Int) = Core.arrayset(A, x, i)
 map(f::Function, a::Array{Any,1}) = Any[ f(a[i]) for i=1:length(a) ]
 
 function precompile(f::ANY, args::Tuple)
-    ccall(:jl_compile_hint, Cint, (Any,), Tuple{Core.Typeof(f), args...}) != 0
+    ccall(:jl_compile_hint, Int32, (Any,), Tuple{Core.Typeof(f), args...}) != 0
 end
 
 function precompile(argt::Type)
-    ccall(:jl_compile_hint, Cint, (Any,), argt) != 0
+    ccall(:jl_compile_hint, Int32, (Any,), argt) != 0
 end
 
 """

base/fastmath.jl

@@ -23,7 +23,9 @@ module FastMath
 
 export @fastmath
 
-import Core.Intrinsics: box, unbox, powi_llvm, sqrt_llvm_fast
+import Core.Intrinsics: powi_llvm, sqrt_llvm_fast, neg_float_fast,
+    add_float_fast, sub_float_fast, mul_float_fast, div_float_fast, rem_float_fast,
+    eq_float_fast, ne_float_fast, lt_float_fast, le_float_fast
 
 const fast_op =
     Dict(# basic arithmetic
@@ -131,18 +133,13 @@ end
 
 FloatTypes = Union{Float32, Float64}
 
-sub_fast{T<:FloatTypes}(x::T) = box(T,Base.neg_float_fast(unbox(T,x)))
+sub_fast{T<:FloatTypes}(x::T) = neg_float_fast(x)
 
-add_fast{T<:FloatTypes}(x::T, y::T) =
-    box(T,Base.add_float_fast(unbox(T,x), unbox(T,y)))
-sub_fast{T<:FloatTypes}(x::T, y::T) =
-    box(T,Base.sub_float_fast(unbox(T,x), unbox(T,y)))
-mul_fast{T<:FloatTypes}(x::T, y::T) =
-    box(T,Base.mul_float_fast(unbox(T,x), unbox(T,y)))
-div_fast{T<:FloatTypes}(x::T, y::T) =
-    box(T,Base.div_float_fast(unbox(T,x), unbox(T,y)))
-rem_fast{T<:FloatTypes}(x::T, y::T) =
-    box(T,Base.rem_float_fast(unbox(T,x), unbox(T,y)))
+add_fast{T<:FloatTypes}(x::T, y::T) = add_float_fast(x, y)
+sub_fast{T<:FloatTypes}(x::T, y::T) = sub_float_fast(x, y)
+mul_fast{T<:FloatTypes}(x::T, y::T) = mul_float_fast(x, y)
+div_fast{T<:FloatTypes}(x::T, y::T) = div_float_fast(x, y)
+rem_fast{T<:FloatTypes}(x::T, y::T) = rem_float_fast(x, y)
 
 add_fast{T<:FloatTypes}(x::T, y::T, zs::T...) =
     add_fast(add_fast(x, y), zs...)
@@ -157,14 +154,10 @@ mul_fast{T<:FloatTypes}(x::T, y::T, zs::T...) =
     end
 end
 
-eq_fast{T<:FloatTypes}(x::T, y::T) =
-    Base.eq_float_fast(unbox(T,x),unbox(T,y))
-ne_fast{T<:FloatTypes}(x::T, y::T) =
-    Base.ne_float_fast(unbox(T,x),unbox(T,y))
-lt_fast{T<:FloatTypes}(x::T, y::T) =
-    Base.lt_float_fast(unbox(T,x),unbox(T,y))
-le_fast{T<:FloatTypes}(x::T, y::T) =
-    Base.le_float_fast(unbox(T,x),unbox(T,y))
+eq_fast{T<:FloatTypes}(x::T, y::T) = eq_float_fast(x, y)
+ne_fast{T<:FloatTypes}(x::T, y::T) = ne_float_fast(x, y)
+lt_fast{T<:FloatTypes}(x::T, y::T) = lt_float_fast(x, y)
+le_fast{T<:FloatTypes}(x::T, y::T) = le_float_fast(x, y)
 
 isinf_fast(x) = false
 isfinite_fast(x) = true
@@ -251,12 +244,11 @@ end
 # builtins
 
 pow_fast{T<:FloatTypes}(x::T, y::Integer) = pow_fast(x, Int32(y))
-pow_fast{T<:FloatTypes}(x::T, y::Int32) =
-    box(T, Base.powi_llvm(unbox(T,x), unbox(Int32,y)))
+pow_fast{T<:FloatTypes}(x::T, y::Int32) = Base.powi_llvm(x, y)
 
 # TODO: Change sqrt_llvm intrinsic to avoid nan checking; add nan
 # checking to sqrt in math.jl; remove sqrt_llvm_fast intrinsic
-sqrt_fast{T<:FloatTypes}(x::T) = box(T, Base.sqrt_llvm_fast(unbox(T,x)))
+sqrt_fast{T<:FloatTypes}(x::T) = sqrt_llvm_fast(x)
 
 # libm