[llvm-branch-commits] [llvm] [SelectionDAG] Widen <n x T> vector types for atomic load (PR #120598)

[Matt Arsenault via llvm-branch-commits]

================
@@ -146,6 +146,54 @@ define <1 x i64> @atomic_vec1_i64_align(ptr %x) nounwind {
   ret <1 x i64> %ret
 }
 
+define <2 x i8> @atomic_vec2_i8(ptr %x) {
+; CHECK3-LABEL: atomic_vec2_i8:
+; CHECK3:       ## %bb.0:
+; CHECK3-NEXT:    movzwl (%rdi), %eax
+; CHECK3-NEXT:    movd %eax, %xmm0
+; CHECK3-NEXT:    retq
+;
+; CHECK0-LABEL: atomic_vec2_i8:
+; CHECK0:       ## %bb.0:
+; CHECK0-NEXT:    movw (%rdi), %cx
+; CHECK0-NEXT:    ## implicit-def: $eax
+; CHECK0-NEXT:    movw %cx, %ax
+; CHECK0-NEXT:    movd %eax, %xmm0
+; CHECK0-NEXT:    retq
+  %ret = load atomic <2 x i8>, ptr %x acquire, align 4
+  ret <2 x i8> %ret
+}
+
+define <2 x i16> @atomic_vec2_i16(ptr %x) {
+; CHECK-LABEL: atomic_vec2_i16:
+; CHECK:       ## %bb.0:
+; CHECK-NEXT:    movl (%rdi), %eax
+; CHECK-NEXT:    movd %eax, %xmm0
----------------
arsenm wrote:

This is broken, this should emit one atomic load 

https://github.com/llvm/llvm-project/pull/120598