[llvm] [InstCombine] Transform `vector.reduce.add` and `splat` into multiplication (PR #161020)

[Hongyu Chen via llvm-commits]

================
@@ -308,3 +308,93 @@ define i32 @diff_of_sums_type_mismatch2(<8 x i32> %v0, <4 x i32> %v1) {
   %r = sub i32 %r0, %r1
   ret i32 %r
 }
+
----------------
XChy wrote:

Looks like other folds are done earlier, so the fold in this patch is not triggered. You can try other types like `<8 x i2>`, etc.
> But I could restrict that the type width shall be wider than the vectors width (I mean vectors elements summarized width)

Sure, that's necessary.

https://github.com/llvm/llvm-project/pull/161020