[clang] [llvm] [NVPTX] Support i256 load/store with 256-bit vector load (PR #155198)

[Artem Belevich via llvm-commits]

================
@@ -1506,3 +1506,98 @@ define void @local_volatile_4xdouble(ptr addrspace(5) %a, ptr addrspace(5) %b) {
   store volatile <4 x double> %a.load, ptr addrspace(5) %b
   ret void
 }
+
+define void @test_i256_global(ptr addrspace(1) %a, ptr addrspace(1) %b) {
----------------
Artem-B wrote:

I'd add a test case for non-global AS to make sure 256 bit loads/stores are not used there, and, maybe a test case for some other 256-bit type (e.g. a <2 x 128> or <4 x double>) to have a reference for what's expected to happen (or not).

https://github.com/llvm/llvm-project/pull/155198