RFR: 8281375: Accelerate bitCount operation for AVX2 and AVX512 target. [v6]

[Jatin Bhateja]

On Tue, 1 Mar 2022 16:34:52 GMT, Sandhya Viswanathan <sviswanathan at openjdk.org> wrote:

>> original vector of integer = [ a3  a2  a1  a0]
>> unpackldq =  [0, a1, 0, a0]
>> unpackhdq = [ 0, a3 , 0 , a2]
>> perform sum of absolute difference and store the result into LSB 16 bits of each quad word.
>> packuswb  packs at lane granularity i.e, 128 bits packed.
>>       128 bit               128 bit
>> [0, sa3, 0, sa2]  [ 0, sa1, 0, sa0 ]  =>   [ sa3, sa2 , sa1, sa0]
>
> The problem is at 512 bit level.

original 512 bit vector holding 16 integers = [ a15 a14 a13 a12, a11 a10 a9 a8 ,a7 a6 a5 a4 ,a3 a2 a1 a0 ]
 unpackdq lower:
                512 bit (vec1) 
   128           128        128      128
 0 a13 0 a12  0 a9 0 a8  0 a5 0 a4  0 a1 0 a0
 unpackdq higher:
                512 bit (vec1) 
   128           128        128      128
 0 a15 0 a14  0 a11 0 a10  0 a7 0 a6  0 a3 0 a2
 Next sum of absolute difference operation followed by pack will squeez each 128 bit lane of two participant vectors
 and interleave them in resulatant vector.
         VEC1_L3     VEC1_L2     VEC1_L1   VEC1_L0              VEC2_L3     VEC2_L2      VEC2_L1    VEC2_L0           
 [ 0 sa13 0 sa12  0 sa9 0 sa8  0 sa5 0 sa4  0 sa1 0 sa0 ] [ 0 sa15 0 sa14  0 sa11 0 sa10  0 sa7 0 sa6  0 sa3 0 sa2 ]
 [ sa15 sa14 sa13 sa12   sa11 sa10 sa9 sa8  sa7 sa6 sa5 sa4  sa3 sa2 sa1 sa0]

-------------

PR: https://git.openjdk.java.net/jdk/pull/7373