1# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
2# RUN: llc -march=amdgcn -mcpu=fiji -run-pass=regbankselect -regbankselect-fast -verify-machineinstrs -o - %s | FileCheck %s
3# RUN: llc -march=amdgcn -mcpu=fiji -run-pass=regbankselect -regbankselect-greedy -verify-machineinstrs -o - %s | FileCheck %s
4
5---
6name: buffer_load_ss
7legalized: true
8tracksRegLiveness: true
9body: |
10  bb.0:
11    liveins: $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4
12
13    ; CHECK-LABEL: name: buffer_load_ss
14    ; CHECK: liveins: $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4
15    ; CHECK: [[COPY:%[0-9]+]]:sgpr(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
16    ; CHECK: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr4
17    ; CHECK: [[AMDGPU_S_BUFFER_LOAD:%[0-9]+]]:sgpr(<4 x s32>) = G_AMDGPU_S_BUFFER_LOAD [[COPY]](<4 x s32>), [[COPY1]](s32), 0
18    %0:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
19    %1:_(s32) = COPY $sgpr4
20    %2:_(<4 x s32>) = G_AMDGPU_S_BUFFER_LOAD %0, %1, 0
21
22...
23
24---
25name: buffer_load_sv
26legalized: true
27tracksRegLiveness: true
28body: |
29  bb.0:
30    liveins: $sgpr0_sgpr1_sgpr2_sgpr3, $vgpr0
31
32    ; CHECK-LABEL: name: buffer_load_sv
33    ; CHECK: liveins: $sgpr0_sgpr1_sgpr2_sgpr3, $vgpr0
34    ; CHECK: [[COPY:%[0-9]+]]:sgpr(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
35    ; CHECK: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
36    ; CHECK: [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0
37    ; CHECK: [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0
38    ; CHECK: [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[COPY]](<4 x s32>), [[C1]](s32), [[COPY1]], [[C]], 0, 0, 0 :: (dereferenceable invariant load (s128), align 4)
39    %0:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
40    %1:_(s32) = COPY $vgpr0
41    %2:_(<4 x s32>) = G_AMDGPU_S_BUFFER_LOAD %0, %1, 0
42
43...
44
45---
46name: buffer_load_vs
47legalized: true
48tracksRegLiveness: true
49body: |
50  bb.0:
51    liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $sgpr0
52
53    ; CHECK-LABEL: name: buffer_load_vs
54    ; CHECK: successors: %bb.1(0x80000000)
55    ; CHECK: liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $sgpr0
56    ; CHECK: [[COPY:%[0-9]+]]:vgpr(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
57    ; CHECK: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr0
58    ; CHECK: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32)
59    ; CHECK: [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0
60    ; CHECK: [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0
61    ; CHECK: [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
62    ; CHECK: [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
63    ; CHECK: [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
64    ; CHECK: .1:
65    ; CHECK: successors: %bb.2(0x40000000), %bb.1(0x40000000)
66    ; CHECK: [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF]], %bb.0, %9, %bb.1
67    ; CHECK: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
68    ; CHECK: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
69    ; CHECK: [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
70    ; CHECK: [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
71    ; CHECK: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
72    ; CHECK: [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
73    ; CHECK: [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
74    ; CHECK: [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
75    ; CHECK: [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
76    ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
77    ; CHECK: [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR]](<4 x s32>), [[C1]](s32), [[COPY2]], [[C]], 0, 0, 0 :: (dereferenceable invariant load (s128), align 4)
78    ; CHECK: [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
79    ; CHECK: $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
80    ; CHECK: S_CBRANCH_EXECNZ %bb.1, implicit $exec
81    ; CHECK: .2:
82    ; CHECK: successors: %bb.3(0x80000000)
83    ; CHECK: $exec = S_MOV_B64_term [[S_MOV_B64_term]]
84    ; CHECK: .3:
85    %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
86    %1:_(s32) = COPY $sgpr0
87    %2:_(<4 x s32>) = G_AMDGPU_S_BUFFER_LOAD %0, %1, 0
88
89...
90
91---
92name: buffer_load_vv
93legalized: true
94tracksRegLiveness: true
95body: |
96  bb.0:
97    liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4
98
99    ; CHECK-LABEL: name: buffer_load_vv
100    ; CHECK: successors: %bb.1(0x80000000)
101    ; CHECK: liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4
102    ; CHECK: [[COPY:%[0-9]+]]:vgpr(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
103    ; CHECK: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr4
104    ; CHECK: [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0
105    ; CHECK: [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0
106    ; CHECK: [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
107    ; CHECK: [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
108    ; CHECK: [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
109    ; CHECK: .1:
110    ; CHECK: successors: %bb.2(0x40000000), %bb.1(0x40000000)
111    ; CHECK: [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF]], %bb.0, %8, %bb.1
112    ; CHECK: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
113    ; CHECK: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
114    ; CHECK: [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
115    ; CHECK: [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
116    ; CHECK: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
117    ; CHECK: [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
118    ; CHECK: [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
119    ; CHECK: [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
120    ; CHECK: [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
121    ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
122    ; CHECK: [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR]](<4 x s32>), [[C1]](s32), [[COPY1]], [[C]], 0, 0, 0 :: (dereferenceable invariant load (s128), align 4)
123    ; CHECK: [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
124    ; CHECK: $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
125    ; CHECK: S_CBRANCH_EXECNZ %bb.1, implicit $exec
126    ; CHECK: .2:
127    ; CHECK: successors: %bb.3(0x80000000)
128    ; CHECK: $exec = S_MOV_B64_term [[S_MOV_B64_term]]
129    ; CHECK: .3:
130    %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
131    %1:_(s32) = COPY $vgpr4
132    %2:_(<4 x s32>) = G_AMDGPU_S_BUFFER_LOAD %0, %1, 0
133
134...
135