xref: /llvm-project/llvm/test/CodeGen/AMDGPU/vgpr_constant_to_sgpr.ll (revision 6548b6354d1d990e1c98736f5e7c3de876bedc8e)
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc -O0 -mcpu=gfx1030 < %s | FileCheck %s
3
4target triple = "amdgcn-amd-amdhsa"
5
6; Unknown functions are conservatively passed all implicit parameters
7declare void @unknown_call()
8; Use the same constant as a sgpr parameter (for the kernel id) and for a vector operation
9define protected amdgpu_kernel void @kern(ptr %addr) !llvm.amdgcn.lds.kernel.id !0 {
10; CHECK-LABEL: kern:
11; CHECK:       ; %bb.0:
12; CHECK-NEXT:    s_mov_b32 s32, 0
13; CHECK-NEXT:    s_add_u32 s12, s12, s17
14; CHECK-NEXT:    s_addc_u32 s13, s13, 0
15; CHECK-NEXT:    s_setreg_b32 hwreg(HW_REG_FLAT_SCR_LO), s12
16; CHECK-NEXT:    s_setreg_b32 hwreg(HW_REG_FLAT_SCR_HI), s13
17; CHECK-NEXT:    s_add_u32 s0, s0, s17
18; CHECK-NEXT:    s_addc_u32 s1, s1, 0
19; CHECK-NEXT:    ; implicit-def: $vgpr40 : SGPR spill to VGPR lane
20; CHECK-NEXT:    v_writelane_b32 v40, s16, 0
21; CHECK-NEXT:    s_mov_b32 s13, s15
22; CHECK-NEXT:    s_mov_b32 s12, s14
23; CHECK-NEXT:    v_readlane_b32 s14, v40, 0
24; CHECK-NEXT:    s_mov_b64 s[16:17], s[8:9]
25; CHECK-NEXT:    s_load_dwordx2 s[8:9], s[16:17], 0x0
26; CHECK-NEXT:    v_mov_b32_e32 v5, 42
27; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
28; CHECK-NEXT:    v_mov_b32_e32 v3, s8
29; CHECK-NEXT:    v_mov_b32_e32 v4, s9
30; CHECK-NEXT:    flat_store_dword v[3:4], v5
31; CHECK-NEXT:    s_mov_b64 s[18:19], 8
32; CHECK-NEXT:    s_mov_b32 s8, s16
33; CHECK-NEXT:    s_mov_b32 s9, s17
34; CHECK-NEXT:    s_mov_b32 s16, s18
35; CHECK-NEXT:    s_mov_b32 s15, s19
36; CHECK-NEXT:    s_add_u32 s8, s8, s16
37; CHECK-NEXT:    s_addc_u32 s15, s9, s15
38; CHECK-NEXT:    ; kill: def $sgpr8 killed $sgpr8 def $sgpr8_sgpr9
39; CHECK-NEXT:    s_mov_b32 s9, s15
40; CHECK-NEXT:    s_getpc_b64 s[16:17]
41; CHECK-NEXT:    s_add_u32 s16, s16, unknown_call@gotpcrel32@lo+4
42; CHECK-NEXT:    s_addc_u32 s17, s17, unknown_call@gotpcrel32@hi+12
43; CHECK-NEXT:    s_load_dwordx2 s[16:17], s[16:17], 0x0
44; CHECK-NEXT:    s_mov_b64 s[22:23], s[2:3]
45; CHECK-NEXT:    s_mov_b64 s[20:21], s[0:1]
46; CHECK-NEXT:    s_mov_b32 s15, 20
47; CHECK-NEXT:    v_lshlrev_b32_e64 v2, s15, v2
48; CHECK-NEXT:    s_mov_b32 s15, 10
49; CHECK-NEXT:    v_lshlrev_b32_e64 v1, s15, v1
50; CHECK-NEXT:    v_or3_b32 v31, v0, v1, v2
51; CHECK-NEXT:    s_mov_b32 s15, 42
52; CHECK-NEXT:    s_mov_b64 s[0:1], s[20:21]
53; CHECK-NEXT:    s_mov_b64 s[2:3], s[22:23]
54; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
55; CHECK-NEXT:    s_swappc_b64 s[30:31], s[16:17]
56; CHECK-NEXT:    s_endpgm
57  store i32 42, ptr %addr
58  call fastcc void @unknown_call()
59  ret void
60}
61
62!llvm.module.flags = !{!1}
63!0 = !{i32 42}
64!1 = !{i32 1, !"amdhsa_code_object_version", i32 500}
65