Attempts to vectorize with two of the variants in https://gist.github.com/simonbyrne/62f25608aaf831de5996 .

gistfile1.txt

$ cat truncloop.jl
# use floating point AND instructions (andps/andpd)
# https://github.com/JuliaLang/julia/issues/9868
function and_float(x::Float64,y::Float64)
       Base.llvmcall("""%av = insertelement<2 x double> undef, double %0, i32 0
         %bv = insertelement<2 x double> undef, double %1, i32 0
         %ai = bitcast <2 x double> %av to <2 x i64>
         %bi = bitcast <2 x double> %bv to <2 x i64>
         %and.i = and <2 x i64> %ai, %bi
         %cf = bitcast <2 x i64> %and.i to <2 x double>
         %cfe = extractelement<2 x double> %cf, i32 0
         ret double %cfe""",Float64,(Float64,Float64),x,y)
end

# 3) AND out bits
t3(x::Float64) = reinterpret(Float64,reinterpret(UInt64,x) & 0xffff_ffff_0000_0000)

# 4) AND out bits using and_float
t4(x::Float64) = and_float(x,reinterpret(Float64, 0xffff_ffff_0000_0000))

function loop3(a,b)
    @inbounds @simd for i = 1:length(a)
        a[i] = t3(b[i])
    end
end

function loop4(a,b)
    @inbounds @simd for i = 1:length(a)
        a[i] = t4(b[i])
    end
end

println("loop3 ----------------------------------------")
code_native(loop3,(Matrix{Float64},Matrix{Float64}))
println()

println("loop4 ----------------------------------------")
code_native(loop4,(Matrix{Float64},Matrix{Float64}))
println()

$ julia truncloop.jl
loop3 ----------------------------------------
        .text
Filename: truncloop.jl
Source line: 0
        push    rbp
        mov     rbp, rsp
Source line: 21
        mov     rax, qword ptr [rdi + 16]
        xor     edx, edx
Source line: 49
        test    rax, rax
        cmovns  rdx, rax
        dec     rdx
        jo      L240
        inc     rdx
        jo      L240
Source line: 50
        test    rdx, rdx
        jle     L230
Source line: 21
        mov     r8, qword ptr [rdi + 8]
        mov     r9, qword ptr [rsi + 8]
        xor     edi, edi
        test    rdx, rdx
        je      L172
        xor     edi, edi
Source line: 56
        mov     r10, rdx
        and     r10, -16
        je      L167
        lea     rdi, qword ptr [r8 + 96]
        lea     rcx, qword ptr [r9 + 96]
        mov     rax, rdx
        and     rax, -16
        movabs  rsi, 140719585243136
Source line: 22
        vbroadcastsd    ymm0, qword ptr [rsi]
        nop     word ptr cs:[rax + rax]
L112:   vandps  ymm1, ymm0, ymmword ptr [rcx - 96]
        vandps  ymm2, ymm0, ymmword ptr [rcx - 64]
        vandps  ymm3, ymm0, ymmword ptr [rcx - 32]
        vandps  ymm4, ymm0, ymmword ptr [rcx]
        vmovups ymmword ptr [rdi - 96], ymm1
        vmovups ymmword ptr [rdi - 64], ymm2
        vmovups ymmword ptr [rdi - 32], ymm3
        vmovups ymmword ptr [rdi], ymm4
Source line: 56
        sub     rdi, -128
        sub     rcx, -128
        add     rax, -16
        jne     L112
        mov     rdi, r10
L167:   cmp     rdx, rdi
        je      L230
L172:   lea     rax, qword ptr [r8 + 8*rdi]
        lea     rcx, qword ptr [r9 + 8*rdi]
        sub     rdx, rdi
        movabs  rsi, -4294967296
        nop     word ptr cs:[rax + rax]
L208:   mov     rdi, qword ptr [rcx]
Source line: 22
        and     rdi, rsi
        mov     qword ptr [rax], rdi
Source line: 57
        add     rax, 8
        add     rcx, 8
        dec     rdx
        jne     L208
L230:   mov     eax, 29730704
Source line: 64
        pop     rbp
        vzeroupper
        ret
Source line: 49
L240:   movabs  rax, jl_overflow_exception
        mov     rdi, qword ptr [rax]
        movabs  rax, jl_throw_with_superfluous_argument
        mov     esi, 49
        call    rax

loop4 ----------------------------------------
        .text
Filename: truncloop.jl
Source line: 0
        push    rbp
        mov     rbp, rsp
Source line: 27
        mov     rcx, qword ptr [rdi + 16]
        xor     eax, eax
Source line: 49
        test    rcx, rcx
        cmovns  rax, rcx
        dec     rax
        jo      L96
        inc     rax
        jo      L96
Source line: 50
        test    rax, rax
        jle     L89
Source line: 27
        mov     rcx, qword ptr [rdi + 8]
        mov     rdx, qword ptr [rsi + 8]
        movabs  rsi, -4294967296
Source line: 28
        vmovq   xmm0, rsi
        nop     word ptr [rax + rax]
L64:    vmovsd  xmm1, qword ptr [rdx]
        vandpd  xmm1, xmm1, xmm0
        vmovlpd qword ptr [rcx], xmm1
Source line: 57
        add     rcx, 8
        add     rdx, 8
        dec     rax
        jne     L64
L89:    mov     eax, 29730704
Source line: 64
        pop     rbp
        ret
Source line: 49
L96:    movabs  rax, jl_overflow_exception
        mov     rdi, qword ptr [rax]
        movabs  rax, jl_throw_with_superfluous_argument
        mov     esi, 49
        call    rax