; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_80 -mattr=+ptx70| FileCheck --check-prefixes=CHECK-PTX64 %s
; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_80 -mattr=+ptx70 --nvptx-short-ptr| FileCheck --check-prefixes=CHECK-PTX-SHARED32 %s
; RUN: %if ptxas-sm_80 && ptxas-isa-7.0 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_80 -mattr=+ptx70| %ptxas-verify -arch=sm_80 %}
; RUN: %if ptxas-sm_80 && ptxas-isa-7.0 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_80 -mattr=+ptx70 --nvptx-short-ptr| %ptxas-verify -arch=sm_80 %}

declare i1 @llvm.nvvm.mbarrier.test.wait.scope.cta.space.cta(ptr addrspace(3), i64)

define void @mbar_test_wait(ptr addrspace(3) %mbar, i64 %state) {
; CHECK-PTX64-LABEL: mbar_test_wait(
; CHECK-PTX64:       {
; CHECK-PTX64-NEXT:    .reg .pred %p<2>;
; CHECK-PTX64-NEXT:    .reg .b64 %rd<3>;
; CHECK-PTX64-EMPTY:
; CHECK-PTX64-NEXT:  // %bb.0:
; CHECK-PTX64-NEXT:    ld.param.b64 %rd1, [mbar_test_wait_param_0];
; CHECK-PTX64-NEXT:    ld.param.b64 %rd2, [mbar_test_wait_param_1];
; CHECK-PTX64-NEXT:    mbarrier.test_wait.shared.b64 %p1, [%rd1], %rd2;
; CHECK-PTX64-NEXT:    ret;
;
; CHECK-PTX-SHARED32-LABEL: mbar_test_wait(
; CHECK-PTX-SHARED32:       {
; CHECK-PTX-SHARED32-NEXT:    .reg .pred %p<2>;
; CHECK-PTX-SHARED32-NEXT:    .reg .b32 %r<2>;
; CHECK-PTX-SHARED32-NEXT:    .reg .b64 %rd<2>;
; CHECK-PTX-SHARED32-EMPTY:
; CHECK-PTX-SHARED32-NEXT:  // %bb.0:
; CHECK-PTX-SHARED32-NEXT:    ld.param.b32 %r1, [mbar_test_wait_param_0];
; CHECK-PTX-SHARED32-NEXT:    ld.param.b64 %rd1, [mbar_test_wait_param_1];
; CHECK-PTX-SHARED32-NEXT:    mbarrier.test_wait.shared.b64 %p1, [%r1], %rd1;
; CHECK-PTX-SHARED32-NEXT:    ret;
  %ret0 = call i1 @llvm.nvvm.mbarrier.test.wait.scope.cta.space.cta(ptr addrspace(3) %mbar, i64 %state)

  ret void
}
