[llvm] [RISCV][CG]Use processShuffleMasks for per-register shuffles (PR #120803)

[Philip Reames via llvm-commits]

================
@@ -168,12 +168,11 @@ define <4 x i64> @m2_splat_into_slide_two_source_v2_lo(<4 x i64> %v1, <4 x i64>
 define <4 x i64> @m2_splat_into_slide_two_source(<4 x i64> %v1, <4 x i64> %v2) vscale_range(2,2) {
 ; CHECK-LABEL: m2_splat_into_slide_two_source:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vsetivli zero, 1, e8, mf8, ta, ma
----------------
preames wrote:

You had added a couple tests in 78ab771;  I had missed finding that on my first quick search.

Please add the following:
- A fully quadratic shuffle, possibly at multiple lmuls
- A mostly linear case - i.e. something which *isn't* linear, but you think is still a good example of the generality.
- Whatever the functional bug was.  

https://github.com/llvm/llvm-project/pull/120803