[clang] [llvm] [RISCV] Support for Zvabd fast-track proposal (PR #124239)

[Pengcheng Wang via llvm-commits]

================
@@ -135,6 +183,43 @@ define signext i32 @sad_2block_16xi8_as_i32(ptr %a, ptr %b, i32 signext %stridea
 ; CHECK-NEXT:    vredsum.vs v8, v8, v12
 ; CHECK-NEXT:    vmv.x.s a0, v8
 ; CHECK-NEXT:    ret
+;
+; ZVABD-LABEL: sad_2block_16xi8_as_i32:
+; ZVABD:       # %bb.0: # %entry
+; ZVABD-NEXT:    vsetivli zero, 16, e8, m1, ta, ma
+; ZVABD-NEXT:    vle8.v v8, (a0)
+; ZVABD-NEXT:    vle8.v v9, (a1)
+; ZVABD-NEXT:    add a0, a0, a2
+; ZVABD-NEXT:    add a1, a1, a3
+; ZVABD-NEXT:    vle8.v v10, (a0)
+; ZVABD-NEXT:    vle8.v v11, (a1)
+; ZVABD-NEXT:    add a0, a0, a2
+; ZVABD-NEXT:    add a1, a1, a3
+; ZVABD-NEXT:    vle8.v v14, (a0)
+; ZVABD-NEXT:    vle8.v v15, (a1)
+; ZVABD-NEXT:    add a0, a0, a2
+; ZVABD-NEXT:    add a1, a1, a3
+; ZVABD-NEXT:    vle8.v v16, (a0)
+; ZVABD-NEXT:    vle8.v v17, (a1)
+; ZVABD-NEXT:    vabdu.vv v10, v10, v11
+; ZVABD-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
+; ZVABD-NEXT:    vzext.vf2 v12, v10
+; ZVABD-NEXT:    vsetvli zero, zero, e8, m1, ta, ma
+; ZVABD-NEXT:    vwabdaccu.vv v12, v8, v9
----------------
wangpc-pp wrote:

I will try to add one.

https://github.com/llvm/llvm-project/pull/124239