Using SSE3 vector shuffel with LDC

[KytoDragon]

I have been trying to port some programs to D that heavely use 
SSE instructions.
In particular, i still need _mm_shuffle_epi8, _mm_alignr_epi8 and 
_mm_aesdec_si128.
LDC does not support the core.simd approach and ldc.simd only 
supports a few operations, including a vector shuffel with a 
fixed mask (I need a variable mask).
So how would one go about using theese with LDC?

I need to be able to:
- consistently generate SSE instruction, even in debug builds.
- inline the function.

I have been unable to find a solution using either the simd 
package, inline asm or inline llvm-ir.