; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
; RUN: llc < %s -O0 -mcpu=sm_20 | FileCheck %s
; RUN: %if ptxas %{ llc < %s -O0 -mcpu=sm_20 | %ptxas-verify %}

target triple = "nvptx64-nvidia-cuda"

define i128 @foo(ptr %p, ptr %o) {
; CHECK-LABEL: foo(
; CHECK:       {
; CHECK-NEXT:    .reg .b64 %rd<4>;
; CHECK-EMPTY:
; CHECK-NEXT:  // %bb.0:
; CHECK-NEXT:    ld.param.b64 %rd2, [foo_param_1];
; CHECK-NEXT:    ld.param.b64 %rd1, [foo_param_0];
; CHECK-NEXT:    ld.b8 %rd3, [%rd1];
; CHECK-NEXT:    st.v2.b64 [%rd2], {%rd3, 0};
; CHECK-NEXT:    st.param.v2.b64 [func_retval0], {%rd3, 0};
; CHECK-NEXT:    ret;
  %c = load i8, ptr %p, align 1
  %i = zext i8 %c to i128
  store i128 %i, ptr %o, align 16
  ret i128 %i
}