1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
|
# This file is a part of Julia. License is MIT: https://julialang.org/license
make_value(::Type{T}, i::Integer) where {T<:Integer} = 3*i%T
make_value(::Type{T},i::Integer) where {T<:AbstractFloat} = T(3*i)
Vec{N,T} = NTuple{N,Base.VecElement{T}}
# Crash report for #15244 motivated this test.
@generated function thrice_iota(::Type{Vec{N,T}}) where {N,T}
:(tuple($([:(Base.VecElement(make_value($T,$i))) for i in 1:N]...)))
end
function call_iota(n::Integer,t::DataType)
x = thrice_iota(Vec{n,t})
@test x[1].value === make_value(t,1)
@test x[n].value === make_value(t,n)
end
# Try various tuple lengths and element types
for i=1:20
for t in [Bool, Int8, Int16, Int32, Int64, Float32, Float64]
call_iota(i,t)
end
end
# Try various large tuple lengths and element types #20961
for i in (34, 36, 48, 64, 72, 80, 96)
for t in [Bool, Int8, Int16, Int32, Int64, Float32, Float64]
call_iota(i,t)
end
end
# Another crash report for #15244 motivated this test.
struct Bunch{N,T}
elts::NTuple{N,Base.VecElement{T}}
end
unpeel(x) = x.elts[1].value
@test unpeel(Bunch{2,Float64}((Base.VecElement(5.0),
Base.VecElement(4.0)))) === 5.0
rewrap(x) = VecElement(x.elts[1].value + 0)
let b = Bunch((VecElement(1.0), VecElement(2.0)))
@test rewrap(b) === VecElement(1.0)
end
struct Herd{N,T}
elts::NTuple{N,Base.VecElement{T}}
Herd{N,T}(elts::NTuple{N,T}) where {N,T} = new(ntuple(i->Base.VecElement{T}(elts[i]), N))
end
function check(x::Herd{N,T}) where {N,T}
for i=1:N
@test x.elts[i].value === N*N+i-1
end
end
check(Herd{1,Int}((1,)))
check(Herd{2,Int}((4,5)))
check(Herd{4,Int}((16,17,18,19)))
struct Gr{N, T}
u::T
v::Bunch{N,T}
w::T
end
let a = Vector{Gr{2,Float64}}(undef, 2)
a[2] = Gr(1.0, Bunch((VecElement(2.0), VecElement(3.0))), 4.0)
a[1] = Gr(5.0, Bunch((VecElement(6.0), VecElement(7.0))), 8.0)
@test a[2] == Gr(1.0, Bunch((VecElement(2.0), VecElement(3.0))), 4.0)
end
@test isa(VecElement((1,2)), VecElement{Tuple{Int,Int}})
# The following test mimic SIMD.jl
const _llvmtypes = Dict{DataType, String}(
Float64 => "double",
Float32 => "float",
Int32 => "i32",
Int64 => "i64"
)
@generated function vecadd(x::Vec{N, T}, y::Vec{N, T}) where {N, T}
llvmT = _llvmtypes[T]
func = T <: AbstractFloat ? "fadd" : "add"
exp = """
%3 = $(func) <$(N) x $(llvmT)> %0, %1
ret <$(N) x $(llvmT)> %3
"""
return quote
Base.@_inline_meta
Core.getfield(Base, :llvmcall)($exp, Vec{$N, $T}, Tuple{Vec{$N, $T}, Vec{$N, $T}}, x, y)
end
end
function f20961(x::Vector{Vec{N, T}}, y::Vector{Vec{N, T}}) where{N, T}
@inbounds begin
a = x[1]
b = y[1]
return vecadd(a, b)
end
end
# Test various SIMD Vectors with known good sizes
for T in (Float64, Float32, Int64, Int32)
for N in 1:36
# For some vectortypes Julia emits llvm arrays instead of vectors
if N % 7 == 0 || N % 11 == 0 || N % 13 == 0 || N % 15 == 0 ||
N % 19 == 0 || N % 23 == 0 || N % 25 == 0 || N % 27 == 0 ||
N % 29 == 0 || N % 31 == 0
continue
end
a = ntuple(i->VecElement(T(i)), N)
result = ntuple(i-> VecElement(T(i+i)), N)
b = vecadd(a, a)
@test b == result
b = f20961([a], [a])
@test b == result
end
end
|
