Fix #1359 - Optimize splatting make usign linear_ramp

[jfalcou]

No description provided.

[DenisYaroshevskiy]

Nice!

[jfalcou]

BEFORE:

wide(int n): 
        stp     x29, x30, [sp, -64]!
        adrp    x2, :got:__stack_chk_guard
        adrp    x1, :got:__stack_chk_guard
        mov     z0.s, w0
        mov     x29, sp
        ldr     x2, [x2, #:got_lo12:__stack_chk_guard]
        ptrue   p0.b, vl32
        sub     sp, sp, #576
        str     p4, [sp]
        ldr     x1, [x1, #:got_lo12:__stack_chk_guard]
        mov     z1.b, #0
        ldr     x0, [x2]
        str     x0, [sp, 632]
        mov     x0, 0
        add     x0, sp, 592
        st1w    z0.d, p0, [x0]
        add     x0, sp, 608
        st1w    z1.d, p0, [x0]
        add     x0, sp, 592
        ld1w    z0.s, p0/z, [x0]
        ldr     x0, [sp, 632]
        ldr     x2, [x1]
        subs    x0, x0, x2
        mov     x2, 0
        bne     .L5
        ldr     p4, [sp]
        add     sp, sp, 576
        ldp     x29, x30, [sp], 64
        ret

AFTER:

wide(int n):
        index   z1.s, #0, #1
        mov     z0.s, w0
        ptrue   p0.b, vl32
        cmplt   p0.s, p0/z, z1.s, #4
        mov     z1.s, p0/z, #-1
        and     z0.d, z0.d, z1.d
        ret

[jfalcou]

Gonna push the one for make logical too, I forgot it.