xref: /llvm-project/llvm/test/CodeGen/NVPTX/i1-ext-load.ll (revision 4583f6d3443c8dc6605c868724e3743161954210)
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --function foo --extra_scrub --default-march nvptx64 --filter-out ".*//.*" --filter-out "[\{\}\(\)]" --version 5
2
3; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_50 | FileCheck %s
4; RUN: %if ptxas %{ llc < %s -mtriple=nvptx64 -mcpu=sm_50 | %ptxas-verify %}
5
6target triple = "nvptx-nvidia-cuda"
7
8define ptx_kernel void @foo(ptr noalias readonly %ptr, ptr noalias %retval) {
9; CHECK-LABEL: foo(
10; CHECK:    .reg .b16 %rs<2>;
11; CHECK:    .reg .b32 %r<4>;
12; CHECK:    .reg .b64 %rd<5>;
13; CHECK-EMPTY:
14; CHECK:    ld.param.u64 %rd1, [foo_param_0];
15; CHECK:    ld.param.u64 %rd2, [foo_param_1];
16; CHECK:    cvta.to.global.u64 %rd3, %rd2;
17; CHECK:    cvta.to.global.u64 %rd4, %rd1;
18; CHECK:    ld.global.nc.u8 %rs1, [%rd4];
19; CHECK:    cvt.u32.u8 %r1, %rs1;
20; CHECK:    add.s32 %r2, %r1, 1;
21; CHECK:    and.b32 %r3, %r2, 1;
22; CHECK:    st.global.u32 [%rd3], %r3;
23; CHECK:    ret;
24  %ld = load i1, ptr %ptr, align 1
25  %zext = zext i1 %ld to i32
26  %add = add i32 %zext, 1
27  %and = and i32 %add, 1
28  store i32 %and, ptr %retval
29  ret void
30}
31