[llvm] [NVPTX] Propagate truncate to operands (PR #98666)

[Artem Belevich via llvm-commits]

================
@@ -18,16 +18,14 @@ define i1 @m2and_ri(i1 %a) {
 
 ; CHECK-LABEL: select2or
 define i1 @select2or(i1 %a, i1 %b) {
-; CHECK: or.b16 %rs{{[0-9]+}}, %rs{{[0-9]+}}, %rs{{[0-9]+}}
-; CHECK-NOT: selp
+; CHECK: or.pred %p{{[0-9]+}}, %p{{[0-9]+}}, %p{{[0-9]+}}
----------------
Artem-B wrote:

Being able to generate predicate math is a nice side effect of the change. 

Let's convert this test to suto-generated checks, too, while we're here.

https://github.com/llvm/llvm-project/pull/98666