1; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX9 %s 2; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX10 %s 3 4declare void @extern_func() 5 6define <4 x float> @non_preserved_vgpr_tuple8(<8 x i32> %rsrc, <4 x i32> %samp, float %bias, float %zcompare, float %s, float %t, float %clamp) { 7; The vgpr tuple8 operand in image_gather4_c_b_cl instruction needs not be 8; preserved across the call and should get 8 scratch registers. 9 10; GFX9-LABEL: non_preserved_vgpr_tuple8: 11; GFX9: buffer_store_dword v44, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill 12; GFX9: buffer_store_dword v40, off, s[0:3], s33 offset:12 ; 4-byte Folded Spill 13; GFX9-NEXT: buffer_store_dword v41, off, s[0:3], s33 offset:8 ; 4-byte Folded Spill 14; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill 15; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s33 ; 4-byte Folded Spill 16 17; GFX9: v_mov_b32_e32 v37, v11 18; GFX9-NEXT: v_mov_b32_e32 v38, v10 19; GFX9-NEXT: v_mov_b32_e32 v49, v9 20; GFX9-NEXT: v_writelane_b32 v44, s30, 0 21; GFX9-NEXT: v_mov_b32_e32 v36, v16 22; GFX9-NEXT: v_mov_b32_e32 v35, v15 23; GFX9-NEXT: v_mov_b32_e32 v34, v14 24; GFX9-NEXT: v_mov_b32_e32 v33, v13 25; GFX9-NEXT: v_mov_b32_e32 v32, v12 26 27; GFX9: ;;#ASMSTART 28; GFX9-NEXT: ;;#ASMEND 29 30; GFX9: image_gather4_c_b_cl v[40:43], v[32:39], s[4:11], s[12:15] dmask:0x1 31; GFX9-NEXT: s_getpc_b64 s[4:5] 32; GFX9-NEXT: s_add_u32 s4, s4, extern_func@gotpcrel32@lo+4 33; GFX9-NEXT: s_addc_u32 s5, s5, extern_func@gotpcrel32@hi+4 34; GFX9-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x0 35; GFX9-NEXT: s_waitcnt lgkmcnt(0) 36; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] 37 38; GFX9: buffer_load_dword v43, off, s[0:3], s33 ; 4-byte Folded Reload 39; GFX9-NEXT: buffer_load_dword v42, off, s[0:3], s33 offset:4 ; 4-byte Folded Reload 40; GFX9-NEXT: buffer_load_dword v41, off, s[0:3], s33 offset:8 ; 4-byte Folded Reload 41; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 offset:12 ; 4-byte Folded Reload 42; GFX9: buffer_load_dword v44, off, s[0:3], s32 offset:16 ; 4-byte Folded Reload 43; GFX9: s_setpc_b64 s[4:5] 44; 45; GFX10-LABEL: non_preserved_vgpr_tuple8: 46; GFX10: buffer_store_dword v44, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill 47; GFX10: buffer_store_dword v40, off, s[0:3], s33 offset:12 ; 4-byte Folded Spill 48; GFX10-NEXT: buffer_store_dword v41, off, s[0:3], s33 offset:8 ; 4-byte Folded Spill 49; GFX10-NEXT: buffer_store_dword v42, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill 50; GFX10-NEXT: buffer_store_dword v43, off, s[0:3], s33 ; 4-byte Folded Spill 51 52; GFX10: v_mov_b32_e32 v36, v16 53; GFX10-NEXT: v_mov_b32_e32 v35, v15 54; GFX10-NEXT: v_mov_b32_e32 v34, v14 55; GFX10-NEXT: v_mov_b32_e32 v33, v13 56; GFX10-NEXT: v_mov_b32_e32 v32, v12 57 58; GFX10: ;;#ASMSTART 59; GFX10-NEXT: ;;#ASMEND 60 61; GFX10: image_gather4_c_b_cl v[40:43], v[32:39], s[4:11], s[12:15] dmask:0x1 62; GFX10-NEXT: v_nop 63; GFX10-NEXT: s_getpc_b64 s[4:5] 64; GFX10-NEXT: s_add_u32 s4, s4, extern_func@gotpcrel32@lo+4 65; GFX10-NEXT: s_addc_u32 s5, s5, extern_func@gotpcrel32@hi+4 66; GFX10-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x0 67; GFX10-NEXT: s_waitcnt lgkmcnt(0) 68; GFX10-NEXT: s_swappc_b64 s[30:31], s[4:5] 69 70; GFX10: buffer_load_dword v43, off, s[0:3], s33 71; GFX10-NEXT: buffer_load_dword v42, off, s[0:3], s33 offset:4 72; GFX10-NEXT: buffer_load_dword v41, off, s[0:3], s33 offset:8 73; GFX10-NEXT: buffer_load_dword v40, off, s[0:3], s33 offset:12 74 75; GFX10: buffer_load_dword v44, off, s[0:3], s32 offset:16 ; 4-byte Folded Reload 76; GFX10: s_setpc_b64 s[4:5] 77main_body: 78 call void asm sideeffect "", "~{v0},~{v1},~{v2},~{v3},~{v4},~{v5},~{v6},~{v7}"() #0 79 call void asm sideeffect "", "~{v8},~{v9},~{v10},~{v11},~{v12},~{v13},~{v14},~{v15}"() #0 80 call void asm sideeffect "", "~{v16},~{v17},~{v18},~{v19},~{v20},~{v21},~{v22},~{v23}"() #0 81 call void asm sideeffect "", "~{v24},~{v25},~{v26},~{v27},~{v28},~{v29},~{v30},~{v31}"() #0 82 %v = call <4 x float> @llvm.amdgcn.image.gather4.c.b.cl.2d.v4f32.f32.f32(i32 1, float %bias, float %zcompare, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0) 83 call void @extern_func() 84 ret <4 x float> %v 85} 86 87define <4 x float> @call_preserved_vgpr_tuple8(<8 x i32> %rsrc, <4 x i32> %samp, float %bias, float %zcompare, float %s, float %t, float %clamp) { 88; The vgpr tuple8 operand in image_gather4_c_b_cl instruction needs to be preserved 89; across the call and should get allcoated to 8 CSRs. 90; Only the lower 5 sub-registers of the tuple are preserved. 91; The upper 3 sub-registers are unused. 92 93; GFX9-LABEL: call_preserved_vgpr_tuple8: 94; GFX9: buffer_store_dword v56, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill 95; GFX9: buffer_store_dword v40, off, s[0:3], s33 offset:16 ; 4-byte Folded Spill 96; GFX9-NEXT: buffer_store_dword v41, off, s[0:3], s33 offset:12 ; 4-byte Folded Spill 97; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s33 offset:8 ; 4-byte Folded Spill 98; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill 99; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s33 ; 4-byte Folded Spill 100 101; GFX9: v_mov_b32_e32 v44, v16 102; GFX9-NEXT: v_mov_b32_e32 v43, v15 103; GFX9-NEXT: v_mov_b32_e32 v42, v14 104; GFX9-NEXT: v_mov_b32_e32 v41, v13 105; GFX9-NEXT: v_mov_b32_e32 v40, v12 106 107; GFX9: s_getpc_b64 s[4:5] 108; GFX9-NEXT: s_add_u32 s4, s4, extern_func@gotpcrel32@lo+4 109; GFX9-NEXT: s_addc_u32 s5, s5, extern_func@gotpcrel32@hi+4 110; GFX9-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x0 111; GFX9-NEXT: image_gather4_c_b_cl v[0:3], v[40:47], s[36:43], s[44:47] dmask:0x1 112; GFX9-NEXT: s_waitcnt vmcnt(0) 113; GFX9-NEXT: global_store_dwordx4 v[0:1], v[0:3], off 114; GFX9-NEXT: s_waitcnt lgkmcnt(0) 115; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] 116; GFX9-NEXT: image_gather4_c_b_cl v[0:3], v[40:47], s[36:43], s[44:47] dmask:0x1 117 118; GFX9: buffer_load_dword v44, off, s[0:3], s33 ; 4-byte Folded Reload 119; GFX9-NEXT: buffer_load_dword v43, off, s[0:3], s33 offset:4 ; 4-byte Folded Reload 120; GFX9-NEXT: buffer_load_dword v42, off, s[0:3], s33 offset:8 ; 4-byte Folded Reload 121; GFX9-NEXT: buffer_load_dword v41, off, s[0:3], s33 offset:12 ; 4-byte Folded Reload 122; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 offset:16 ; 4-byte Folded Reload 123 124; GFX9: buffer_load_dword v56, off, s[0:3], s32 offset:20 ; 4-byte Folded Reload 125; GFX9: s_setpc_b64 s[4:5] 126; 127; GFX10-LABEL: call_preserved_vgpr_tuple8: 128; GFX10: buffer_store_dword v45, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill 129; GFX10: buffer_store_dword v40, off, s[0:3], s33 offset:16 ; 4-byte Folded Spill 130; GFX10-NEXT: buffer_store_dword v41, off, s[0:3], s33 offset:12 ; 4-byte Folded Spill 131; GFX10-NEXT: buffer_store_dword v42, off, s[0:3], s33 offset:8 ; 4-byte Folded Spill 132; GFX10-NEXT: buffer_store_dword v43, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill 133; GFX10-NEXT: buffer_store_dword v44, off, s[0:3], s33 ; 4-byte Folded Spill 134 135; GFX10: s_getpc_b64 s[4:5] 136; GFX10-NEXT: s_add_u32 s4, s4, extern_func@gotpcrel32@lo+4 137; GFX10-NEXT: s_addc_u32 s5, s5, extern_func@gotpcrel32@hi+4 138; GFX10-NEXT: v_mov_b32_e32 v40, v16 139; GFX10-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x0 140; GFX10-NEXT: image_gather4_c_b_cl v[0:3], v[12:19], s[36:43], s[44:47] dmask:0x1 141; GFX10-NEXT: v_mov_b32_e32 v41, v15 142; GFX10-NEXT: v_mov_b32_e32 v42, v14 143; GFX10-NEXT: v_mov_b32_e32 v43, v13 144; GFX10-NEXT: v_mov_b32_e32 v44, v12 145; GFX10-NEXT: ; implicit-def: $vcc_hi 146; GFX10-NEXT: s_waitcnt vmcnt(0) 147; GFX10-NEXT: global_store_dwordx4 v[0:1], v[0:3], off 148; GFX10-NEXT: s_waitcnt lgkmcnt(0) 149; GFX10-NEXT: s_swappc_b64 s[30:31], s[4:5] 150; GFX10-NEXT: image_gather4_c_b_cl v[0:3], [v44, v43, v42, v41, v40], s[36:43], s[44:47] dmask:0x1 151 152; GFX10: buffer_load_dword v44, off, s[0:3], s33 153; GFX10-NEXT: buffer_load_dword v43, off, s[0:3], s33 offset:4 154; GFX10-NEXT: buffer_load_dword v42, off, s[0:3], s33 offset:8 155; GFX10-NEXT: buffer_load_dword v41, off, s[0:3], s33 offset:12 156; GFX10-NEXT: buffer_load_dword v40, off, s[0:3], s33 offset:16 157; GFX10: buffer_load_dword v45, off, s[0:3], s32 offset:20 158; GFX10: s_setpc_b64 s[4:5] 159main_body: 160 %v = call <4 x float> @llvm.amdgcn.image.gather4.c.b.cl.2d.v4f32.f32.f32(i32 1, float %bias, float %zcompare, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0) 161 store <4 x float> %v, <4 x float> addrspace(1)* undef 162 call void @extern_func() 163 %v1 = call <4 x float> @llvm.amdgcn.image.gather4.c.b.cl.2d.v4f32.f32.f32(i32 1, float %bias, float %zcompare, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0) 164 ret <4 x float> %v1 165} 166 167declare <4 x float> @llvm.amdgcn.image.gather4.c.b.cl.2d.v4f32.f32.f32(i32 immarg, float, float, float, float, float, <8 x i32>, <4 x i32>, i1 immarg, i32 immarg, i32 immarg) #1 168 169attributes #0 = { nounwind writeonly } 170attributes #1 = { nounwind readonly } 171