I was anticipating a SIMD shuffle/permute instruction like vpermq, it won't allocate more ram-memory per-se. Anyways, here is what google search AI gave me as

I was anticipating a SIMD shuffle/permute instruction like vpermq, it won't allocate more ram-memory per-se. Anyways, here is what google search AI gave me as an example of how that would work (I don't know this stuff well enough myself): ; Assume ymm0 contains [A, B, C, D] (Q0=A, Q1=B, Q2=C, Q3=D) ; The immediate 0xd8 (11011000 in binary) means: ; - Keep Q0 (index 0) ; - Swap Q1 (index 1) with Q2 (index 2) (110, 011 in binary for bits 1,2) ; - Keep Q3 (index 3) vpermq ymm0, ymm0, 0xd8 ; ymm0 now contains [A, C, B, D]