[llvm] [RISCV] Initial codegen support for zvqdotq extension (PR #137039)

[Craig Topper via llvm-commits]

================
@@ -181,29 +221,77 @@ entry:
 }
 
 define i32 @reduce_of_sext(<16 x i8> %a) {
-; CHECK-LABEL: reduce_of_sext:
-; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    vsetivli zero, 16, e32, m4, ta, ma
-; CHECK-NEXT:    vsext.vf4 v12, v8
-; CHECK-NEXT:    vmv.s.x v8, zero
-; CHECK-NEXT:    vredsum.vs v8, v12, v8
-; CHECK-NEXT:    vmv.x.s a0, v8
-; CHECK-NEXT:    ret
+; NODOT-LABEL: reduce_of_sext:
+; NODOT:       # %bb.0: # %entry
+; NODOT-NEXT:    vsetivli zero, 16, e32, m4, ta, ma
+; NODOT-NEXT:    vsext.vf4 v12, v8
----------------
topperc wrote:

Not related to this patch, but is this a missed opportunity to use vsext.vf2+vwredsum?

https://github.com/llvm/llvm-project/pull/137039