1 # RUN: llc -mtriple=amdgcn -mcpu=gfx1030 -greedy-regclass-priority-trumps-globalness=0 -start-before greedy -o - %s | FileCheck %s -check-prefixes=GCN,OLD
2 # RUN: llc -mtriple=amdgcn -mcpu=gfx1030 -greedy-regclass-priority-trumps-globalness=1 -start-before greedy -o - %s | FileCheck %s -check-prefixes=GCN,NEW
4 # At the time of writing -greedy-regclass-priority-trumps-globalness makes a
5 # significant improvement in the total number of vgprs needed to compile this
6 # test, from 11 down to 7.
9 # OLD: NumVgprs: 11{{$}}
10 # NEW: NumVgprs: 7{{$}}
13 tracksRegLiveness: true
16 successors: %bb.1, %bb.2
17 liveins: $vgpr0, $vgpr6
19 %6:vgpr_32 = COPY $vgpr6
20 undef %30.sub0:vreg_128 = COPY $vgpr0
21 undef %27.sub0:vreg_128 = V_MED3_F32_e64 0, 0, 0, 0, 0, 0, 0, 0, implicit $mode, implicit $exec
22 undef %16.sub0:sgpr_256 = S_MOV_B32 0
23 undef %26.sub1:vreg_64 = V_LSHRREV_B32_e32 1, %6, implicit $exec
24 %27.sub1:vreg_128 = COPY %27.sub0
25 %27.sub2:vreg_128 = COPY %27.sub0
26 %27.sub3:vreg_128 = COPY %27.sub0
27 %26.sub0:vreg_64 = V_MOV_B32_e32 1, implicit $exec
28 %16.sub1:sgpr_256 = COPY %16.sub0
29 %16.sub2:sgpr_256 = COPY %16.sub0
30 %16.sub3:sgpr_256 = COPY %16.sub0
31 %16.sub4:sgpr_256 = COPY %16.sub0
32 %16.sub5:sgpr_256 = COPY %16.sub0
33 %16.sub6:sgpr_256 = COPY %16.sub0
34 %16.sub7:sgpr_256 = COPY %16.sub0
35 IMAGE_STORE_V4_V2_gfx10 %27, %26, %16, 0, 1, -1, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), addrspace 7)
36 S_CBRANCH_SCC1 %bb.2, implicit undef $scc
40 %30.sub1:vreg_128 = V_MOV_B32_e32 0, implicit $exec
41 %30.sub2:vreg_128 = COPY %30.sub1
42 %30.sub3:vreg_128 = COPY %30.sub1
43 %26.sub1:vreg_64 = COPY %30.sub1
44 IMAGE_STORE_V4_V2_gfx10 %30, %26, %16, 0, 1, -1, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), addrspace 7)
51 # OLD: NumVgprs: 7{{$}}
52 # NEW: NumVgprs: 11{{$}}
55 tracksRegLiveness: true
58 liveins: $vgpr0, $vgpr2
60 %8:vgpr_32 = COPY $vgpr2
61 %6:vgpr_32 = COPY $vgpr0
64 successors: %bb.2, %bb.3
66 undef %25.sub0:sgpr_256 = S_MOV_B32 0
67 %10:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
68 %25.sub1:sgpr_256 = COPY %25.sub0
69 %25.sub2:sgpr_256 = COPY %25.sub0
70 %25.sub3:sgpr_256 = COPY %25.sub0
71 %19:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN %10, %25.sub0_sub1_sub2_sub3, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32))
72 %36:vreg_64 = DS_READ2_B32_gfx9 %10, 0, 2, 0, implicit $exec :: (load (s32)), (load (s32))
73 %30:vgpr_32 = V_LSHLREV_B32_e32 2, %36.sub0, implicit $exec
74 %3:vgpr_32 = V_ADD_U32_e32 %30, %6, implicit $exec
75 %4:vgpr_32 = V_ADD_U32_e32 %36.sub1, %8, implicit $exec
76 %15:vgpr_32 = V_OR_B32_e32 %4, %3, implicit $exec
77 %21:sreg_32 = V_READFIRSTLANE_B32 %19, implicit $exec
78 %17:sreg_32 = V_CMP_EQ_U32_e64 0, %15, implicit $exec
79 S_CMP_LG_U32 %21, 0, implicit-def $scc
80 %31:sreg_32_xm0_xexec = S_CSELECT_B32 -1, 0, implicit killed $scc
81 %23:sreg_32 = S_AND_B32 %31, %17, implicit-def dead $scc
82 %5:sreg_32 = COPY $exec_lo, implicit-def $exec_lo
83 %37:sreg_32 = S_AND_B32 %5, %23, implicit-def dead $scc
84 $exec_lo = S_MOV_B32_term %37
85 S_CBRANCH_EXECZ %bb.3, implicit $exec
89 undef %28.sub0:vreg_128 = V_MOV_B32_e32 0, implicit $exec
90 %25.sub1:sgpr_256 = COPY %25.sub0
91 %25.sub2:sgpr_256 = COPY %25.sub0
92 %25.sub3:sgpr_256 = COPY %25.sub0
93 %25.sub4:sgpr_256 = COPY %25.sub0
94 %28.sub1:vreg_128 = COPY %28.sub0
95 %28.sub2:vreg_128 = COPY %28.sub0
96 %28.sub3:vreg_128 = COPY %28.sub0
97 %25.sub5:sgpr_256 = COPY %25.sub0
98 %25.sub6:sgpr_256 = COPY %25.sub0
99 %25.sub7:sgpr_256 = COPY %25.sub0
100 IMAGE_STORE_V4_V3_nsa_gfx10 %28, %3, %30, %4, %25, 0, 2, -1, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable store (s32))