[llvm] [AVX-512] make vpternlogq more aggressive for longer chains of bitmanipulations (PR #189971)

[Julian Pokrovsky via llvm-commits]

================
@@ -1124,14 +1166,31 @@ define void @vec256_v2i64(ptr %in.subvec.ptr, ptr %out.subvec.ptr, ptr %out.vec.
 ; SSE2-NEXT:    movdqa %xmm0, 16(%rdx)
 ; SSE2-NEXT:    retq
 ;
-; AVX-LABEL: vec256_v2i64:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
-; AVX-NEXT:    vpxor (%rdi), %xmm0, %xmm0
-; AVX-NEXT:    vmovdqa %xmm0, (%rsi)
-; AVX-NEXT:    vmovdqa %xmm0, (%rdx)
-; AVX-NEXT:    vmovdqa %xmm0, 16(%rdx)
-; AVX-NEXT:    retq
+; AVX1-LABEL: vec256_v2i64:
----------------
raventid wrote:

merged latest main and regenerated (https://github.com/llvm/llvm-project/pull/189971/changes/fa198307d4aefc4e279dd4a7f9240d8972ce0b89)

https://github.com/llvm/llvm-project/pull/189971