ArchRobison

julia> function foo( a::NTuple{4,Float32}, b::NTuple{4,Float32} )
           (a[1]+b[1],a[2]+b[2],a[3]+b[3],a[4]+b[4])
       end
foo (generic function with 1 method)

julia> t = NTuple{4,Float32}
(Float32,Float32,Float32,Float32)

julia> code_llvm(foo,(t,t))

ArchRobison

function saxpy( a, x, y )
    @simd for i=1:length(x)
        @inbounds y[i] = y[i]+a*x[i]
    end
end

function flog( n, reps, tolerance )
    x = rand(Float32,n)
    y = rand(Float32,n)
    z = copy(y)

ArchRobison

function sweep( irange, jrange, U, Vx, Vy, A, B )
    for j in jrange
        @simd for i in irange
            @inbounds begin
                u = U[i,j]
                Vx[i,j] += (A[i,j+1]+A[i,j])*(U[i,j+1]-u)
                Vy[i,j] += (A[i+1,j]+A[i,j])*(U[i+1,j]-u)
                U [i,j] = u + B[i,j]*((Vx[i,j]-Vx[i,j-1]) + (Vy[i,j]-Vy[i-1,j]))
            end
        end

ArchRobison

function inner( x, y )
    s = zero(eltype(x))
    @simd for i=1:length(x)
        @inbounds s += x[i]*y[i]
    end
    s
end

function flog( n, reps, tolerance )
    x = rand(Float32,n)

ArchRobison

$ cat foo.jl
function add( a::NTuple{4,Float32}, b::NTuple{4,Float32} )
    (a[1]+b[1],a[2]+b[2],a[3]+b[3],a[4]+b[4])
end

function mul( a::NTuple{4,Float32}, b::NTuple{4,Float32} )
    (a[1]*b[1],a[2]*b[2],a[3]*b[3],a[4]*b[4])
end

function madd( a::NTuple{4,Float32}, b::NTuple{4,Float32}, c::NTuple{4,Float32} )

ArchRobison

$ cat vadd.jl
function vadd( a::NTuple{4,Float64}, b::NTuple{4,Float64} )
    (a[1]+b[1],a[2]+b[2],a[3]+b[3],a[4]+b[4])
end

function vadd_one!(arr::Array{Float64, 1})
    len  = length(arr) # assuming len multiple of 4
    one =  (1.0, 1.0, 1.0, 1.0)
    @inbounds for i = 1:4:len
        inp = (arr[i], arr[i+1], arr[i+2], arr[i+3])

ArchRobison

$ cat v.jl
function vadd_one!(arr::Array{Float32, 1})
    @simd for i = 1:length(arr)
         @inbounds arr[i] += 1
    end
end

code_llvm(vadd_one!,(Array{Float32, 1},))
$ julia v.jl

ArchRobison

L.preheader:                                      ; preds = %top
  %.op = add i64 %0, 1, !dbg !9
  %8 = select i1 %2, i64 %.op, i64 1, !dbg !9
  %9 = sub i64 %8, %5, !dbg !9
  %xtraiter = and i64 %9, 7
  switch i64 %xtraiter, label %L.unr [
    i64 0, label %L.preheader.split
    i64 1, label %L.unr22
    i64 2, label %L.unr18
    i64 3, label %L.unr14

ArchRobison

--- a/src/init.c
+++ b/src/init.c
@@ -64,6 +64,9 @@ extern BOOL (WINAPI *hSymRefreshModuleList)(HANDLE);
 #include <sched.h>   // for setting CPU affinity
 #endif

+#undef SIGSTKSZ
+#define SIGSTKSZ (64*1024)
+
 char *julia_home = NULL;

ArchRobison

::::::::::::::
test1.jl
::::::::::::::
# Example from early discussion.
# Has constant lower bound, expression as upper bound, and no reductions.
function test1( a, x, y )
    @simd for i=1:length(x)
        @inbounds y[i] = y[i]+a*x[i]
    end
end