[llvm] [RISCV][VSETVLI] Prefer VTYPE for immediate known to be less than VLMAX (PR #69759)

[Philip Reames via llvm-commits]

================
@@ -406,13 +406,13 @@ define <16 x i32> @zextload_v16i8_v16i32(ptr %x) {
 ; LMULMAX1-NEXT:    vslidedown.vi v11, v9, 8
 ; LMULMAX1-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
 ; LMULMAX1-NEXT:    vzext.vf4 v10, v11
-; LMULMAX1-NEXT:    vsetivli zero, 4, e8, mf2, ta, ma
+; LMULMAX1-NEXT:    vsetvli zero, zero, e8, mf2, ta, ma
----------------
preames wrote:

You are unfortunately correct here.  This is an absolutely silly choice for the specification to make.  It doesn't disallow an implementation to have the sane semantics here, but you're right that we can't rely on this instruction being valid when VLMAX changes (even if VL doesn't.)

https://github.com/llvm/llvm-project/pull/69759