1 ; RUN: llc -show-mc-encoding -mattr=-code-object-v3,+promote-alloca -disable-promote-alloca-to-vector -amdgpu-load-store-vectorizer=0 -enable-amdgpu-aa=0 -verify-machineinstrs -march=amdgcn < %s | FileCheck -enable-var-scope -check-prefix=SI-PROMOTE -check-prefix=SI -check-prefix=FUNC %s
2 ; RUN: llc -show-mc-encoding -mattr=-code-object-v3,+promote-alloca -disable-promote-alloca-to-vector -amdgpu-load-store-vectorizer=0 -enable-amdgpu-aa=0 -verify-machineinstrs -mtriple=amdgcn--amdhsa -mcpu=kaveri -mattr=-code-object-v3,-unaligned-buffer-access < %s | FileCheck -enable-var-scope -check-prefix=SI-PROMOTE -check-prefix=SI -check-prefix=FUNC -check-prefix=HSA-PROMOTE %s
3 ; RUN: llc -show-mc-encoding -mattr=-code-object-v3,-promote-alloca -amdgpu-load-store-vectorizer=0 -enable-amdgpu-aa=0 -verify-machineinstrs -march=amdgcn < %s | FileCheck %s -check-prefix=SI-ALLOCA -check-prefix=SI -check-prefix=FUNC
4 ; RUN: llc -show-mc-encoding -mattr=-code-object-v3,-promote-alloca -amdgpu-load-store-vectorizer=0 -enable-amdgpu-aa=0 -verify-machineinstrs -mtriple=amdgcn-amdhsa -mcpu=kaveri -mattr=-code-object-v3,-unaligned-buffer-access < %s | FileCheck -enable-var-scope -check-prefix=SI-ALLOCA -check-prefix=SI -check-prefix=FUNC -check-prefix=HSA-ALLOCA %s
5 ; RUN: llc -show-mc-encoding -mattr=-code-object-v3,+promote-alloca -disable-promote-alloca-to-vector -amdgpu-load-store-vectorizer=0 -enable-amdgpu-aa=0 -verify-machineinstrs -mtriple=amdgcn-amdhsa -march=amdgcn -mcpu=tonga -mattr=-code-object-v3,-unaligned-buffer-access < %s | FileCheck -enable-var-scope -check-prefix=SI-PROMOTE -check-prefix=SI -check-prefix=FUNC %s
6 ; RUN: llc -show-mc-encoding -mattr=-code-object-v3,+promote-alloca -amdgpu-load-store-vectorizer=0 -enable-amdgpu-aa=0 -verify-machineinstrs -mtriple=amdgcn-amdhsa -march=amdgcn -mcpu=tonga -mattr=-code-object-v3,-unaligned-buffer-access < %s | FileCheck -enable-var-scope -check-prefix=SI-PROMOTE-VECT -check-prefix=SI -check-prefix=FUNC %s
7 ; RUN: llc -show-mc-encoding -mattr=-code-object-v3,-promote-alloca -amdgpu-load-store-vectorizer=0 -enable-amdgpu-aa=0 -verify-machineinstrs -mtriple=amdgcn-amdhsa -march=amdgcn -mcpu=tonga -mattr=-code-object-v3,-unaligned-buffer-access < %s | FileCheck -enable-var-scope -check-prefix=SI-ALLOCA -check-prefix=SI -check-prefix=FUNC %s
9 ; RUN: opt -S -mtriple=amdgcn-unknown-amdhsa -data-layout=A5 -mcpu=kaveri -amdgpu-promote-alloca -disable-promote-alloca-to-vector < %s | FileCheck -enable-var-scope -check-prefix=HSAOPT -check-prefix=OPT %s
10 ; RUN: opt -S -mtriple=amdgcn-unknown-unknown -data-layout=A5 -mcpu=kaveri -amdgpu-promote-alloca -disable-promote-alloca-to-vector < %s | FileCheck -enable-var-scope -check-prefix=NOHSAOPT -check-prefix=OPT %s
12 ; RUN: llc -march=r600 -mcpu=cypress -disable-promote-alloca-to-vector < %s | FileCheck %s -check-prefix=R600 -check-prefix=FUNC
13 ; RUN: llc -march=r600 -mcpu=cypress < %s | FileCheck %s -check-prefix=R600-VECT -check-prefix=FUNC
15 ; HSAOPT: @mova_same_clause.stack = internal unnamed_addr addrspace(3) global [256 x [5 x i32]] undef, align 4
16 ; HSAOPT: @high_alignment.stack = internal unnamed_addr addrspace(3) global [256 x [8 x i32]] undef, align 16
19 ; FUNC-LABEL: {{^}}mova_same_clause:
20 ; OPT-LABEL: @mova_same_clause(
27 ; HSA-PROMOTE: .amd_kernel_code_t
28 ; HSA-PROMOTE: workgroup_group_segment_byte_size = 5120
29 ; HSA-PROMOTE: .end_amd_kernel_code_t
31 ; HSA-PROMOTE: s_load_dword s{{[0-9]+}}, s[4:5], 0x2
33 ; SI-PROMOTE: ds_write_b32
34 ; SI-PROMOTE: ds_write_b32
35 ; SI-PROMOTE: ds_read_b32
36 ; SI-PROMOTE: ds_read_b32
38 ; HSA-ALLOCA: .amd_kernel_code_t
39 ; FIXME: Creating the emergency stack slots causes us to over-estimate scratch
41 ; HSA-ALLOCA: workitem_private_segment_byte_size = 24
42 ; HSA-ALLOCA: .end_amd_kernel_code_t
44 ; HSA-ALLOCA: s_mov_b32 flat_scratch_lo, s7
45 ; HSA-ALLOCA: s_add_u32 s6, s6, s9
46 ; HSA-ALLOCA: s_lshr_b32 flat_scratch_hi, s6, 8
48 ; SI-ALLOCA: buffer_store_dword v{{[0-9]+}}, v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} offen ; encoding: [0x00,0x10,0x70,0xe0
49 ; SI-ALLOCA: buffer_store_dword v{{[0-9]+}}, v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} offen ; encoding: [0x00,0x10,0x70,0xe0
52 ; HSAOPT: [[DISPATCH_PTR:%[0-9]+]] = call noalias nonnull dereferenceable(64) i8 addrspace(4)* @llvm.amdgcn.dispatch.ptr()
53 ; HSAOPT: [[CAST_DISPATCH_PTR:%[0-9]+]] = bitcast i8 addrspace(4)* [[DISPATCH_PTR]] to i32 addrspace(4)*
54 ; HSAOPT: [[GEP0:%[0-9]+]] = getelementptr inbounds i32, i32 addrspace(4)* [[CAST_DISPATCH_PTR]], i64 1
55 ; HSAOPT: [[LDXY:%[0-9]+]] = load i32, i32 addrspace(4)* [[GEP0]], align 4, !invariant.load !0
56 ; HSAOPT: [[GEP1:%[0-9]+]] = getelementptr inbounds i32, i32 addrspace(4)* [[CAST_DISPATCH_PTR]], i64 2
57 ; HSAOPT: [[LDZU:%[0-9]+]] = load i32, i32 addrspace(4)* [[GEP1]], align 4, !range !1, !invariant.load !0
58 ; HSAOPT: [[EXTRACTY:%[0-9]+]] = lshr i32 [[LDXY]], 16
60 ; HSAOPT: [[WORKITEM_ID_X:%[0-9]+]] = call i32 @llvm.amdgcn.workitem.id.x(), !range !2
61 ; HSAOPT: [[WORKITEM_ID_Y:%[0-9]+]] = call i32 @llvm.amdgcn.workitem.id.y(), !range !2
62 ; HSAOPT: [[WORKITEM_ID_Z:%[0-9]+]] = call i32 @llvm.amdgcn.workitem.id.z(), !range !2
64 ; HSAOPT: [[Y_SIZE_X_Z_SIZE:%[0-9]+]] = mul nuw nsw i32 [[EXTRACTY]], [[LDZU]]
65 ; HSAOPT: [[YZ_X_XID:%[0-9]+]] = mul i32 [[Y_SIZE_X_Z_SIZE]], [[WORKITEM_ID_X]]
66 ; HSAOPT: [[Y_X_Z_SIZE:%[0-9]+]] = mul nuw nsw i32 [[WORKITEM_ID_Y]], [[LDZU]]
67 ; HSAOPT: [[ADD_YZ_X_X_YZ_SIZE:%[0-9]+]] = add i32 [[YZ_X_XID]], [[Y_X_Z_SIZE]]
68 ; HSAOPT: [[ADD_ZID:%[0-9]+]] = add i32 [[ADD_YZ_X_X_YZ_SIZE]], [[WORKITEM_ID_Z]]
70 ; HSAOPT: [[LOCAL_GEP:%[0-9]+]] = getelementptr inbounds [256 x [5 x i32]], [256 x [5 x i32]] addrspace(3)* @mova_same_clause.stack, i32 0, i32 [[ADD_ZID]]
71 ; HSAOPT: %arrayidx1 = getelementptr inbounds [5 x i32], [5 x i32] addrspace(3)* [[LOCAL_GEP]], i32 0, i32 {{%[0-9]+}}
72 ; HSAOPT: %arrayidx3 = getelementptr inbounds [5 x i32], [5 x i32] addrspace(3)* [[LOCAL_GEP]], i32 0, i32 {{%[0-9]+}}
73 ; HSAOPT: %arrayidx10 = getelementptr inbounds [5 x i32], [5 x i32] addrspace(3)* [[LOCAL_GEP]], i32 0, i32 0
74 ; HSAOPT: %arrayidx12 = getelementptr inbounds [5 x i32], [5 x i32] addrspace(3)* [[LOCAL_GEP]], i32 0, i32 1
77 ; NOHSAOPT: call i32 @llvm.r600.read.local.size.y(), !range !0
78 ; NOHSAOPT: call i32 @llvm.r600.read.local.size.z(), !range !0
79 ; NOHSAOPT: call i32 @llvm.amdgcn.workitem.id.x(), !range !1
80 ; NOHSAOPT: call i32 @llvm.amdgcn.workitem.id.y(), !range !1
81 ; NOHSAOPT: call i32 @llvm.amdgcn.workitem.id.z(), !range !1
82 define amdgpu_kernel void @mova_same_clause(i32 addrspace(1)* nocapture %out, i32 addrspace(1)* nocapture %in) #0 {
84 %stack = alloca [5 x i32], align 4, addrspace(5)
85 %0 = load i32, i32 addrspace(1)* %in, align 4
86 %arrayidx1 = getelementptr inbounds [5 x i32], [5 x i32] addrspace(5)* %stack, i32 0, i32 %0
87 store i32 4, i32 addrspace(5)* %arrayidx1, align 4
88 %arrayidx2 = getelementptr inbounds i32, i32 addrspace(1)* %in, i32 1
89 %1 = load i32, i32 addrspace(1)* %arrayidx2, align 4
90 %arrayidx3 = getelementptr inbounds [5 x i32], [5 x i32] addrspace(5)* %stack, i32 0, i32 %1
91 store i32 5, i32 addrspace(5)* %arrayidx3, align 4
92 %arrayidx10 = getelementptr inbounds [5 x i32], [5 x i32] addrspace(5)* %stack, i32 0, i32 0
93 %2 = load i32, i32 addrspace(5)* %arrayidx10, align 4
94 store i32 %2, i32 addrspace(1)* %out, align 4
95 %arrayidx12 = getelementptr inbounds [5 x i32], [5 x i32] addrspace(5)* %stack, i32 0, i32 1
96 %3 = load i32, i32 addrspace(5)* %arrayidx12
97 %arrayidx13 = getelementptr inbounds i32, i32 addrspace(1)* %out, i32 1
98 store i32 %3, i32 addrspace(1)* %arrayidx13
102 ; OPT-LABEL: @high_alignment(
103 ; OPT: getelementptr inbounds [256 x [8 x i32]], [256 x [8 x i32]] addrspace(3)* @high_alignment.stack, i32 0, i32 %{{[0-9]+}}
104 define amdgpu_kernel void @high_alignment(i32 addrspace(1)* nocapture %out, i32 addrspace(1)* nocapture %in) #0 {
106 %stack = alloca [8 x i32], align 16, addrspace(5)
107 %0 = load i32, i32 addrspace(1)* %in, align 4
108 %arrayidx1 = getelementptr inbounds [8 x i32], [8 x i32] addrspace(5)* %stack, i32 0, i32 %0
109 store i32 4, i32 addrspace(5)* %arrayidx1, align 4
110 %arrayidx2 = getelementptr inbounds i32, i32 addrspace(1)* %in, i32 1
111 %1 = load i32, i32 addrspace(1)* %arrayidx2, align 4
112 %arrayidx3 = getelementptr inbounds [8 x i32], [8 x i32] addrspace(5)* %stack, i32 0, i32 %1
113 store i32 5, i32 addrspace(5)* %arrayidx3, align 4
114 %arrayidx10 = getelementptr inbounds [8 x i32], [8 x i32] addrspace(5)* %stack, i32 0, i32 0
115 %2 = load i32, i32 addrspace(5)* %arrayidx10, align 4
116 store i32 %2, i32 addrspace(1)* %out, align 4
117 %arrayidx12 = getelementptr inbounds [8 x i32], [8 x i32] addrspace(5)* %stack, i32 0, i32 1
118 %3 = load i32, i32 addrspace(5)* %arrayidx12
119 %arrayidx13 = getelementptr inbounds i32, i32 addrspace(1)* %out, i32 1
120 store i32 %3, i32 addrspace(1)* %arrayidx13
124 ; FUNC-LABEL: {{^}}no_replace_inbounds_gep:
125 ; OPT-LABEL: @no_replace_inbounds_gep(
126 ; OPT: alloca [5 x i32]
129 define amdgpu_kernel void @no_replace_inbounds_gep(i32 addrspace(1)* nocapture %out, i32 addrspace(1)* nocapture %in) #0 {
131 %stack = alloca [5 x i32], align 4, addrspace(5)
132 %0 = load i32, i32 addrspace(1)* %in, align 4
133 %arrayidx1 = getelementptr [5 x i32], [5 x i32] addrspace(5)* %stack, i32 0, i32 %0
134 store i32 4, i32 addrspace(5)* %arrayidx1, align 4
135 %arrayidx2 = getelementptr inbounds i32, i32 addrspace(1)* %in, i32 1
136 %1 = load i32, i32 addrspace(1)* %arrayidx2, align 4
137 %arrayidx3 = getelementptr inbounds [5 x i32], [5 x i32] addrspace(5)* %stack, i32 0, i32 %1
138 store i32 5, i32 addrspace(5)* %arrayidx3, align 4
139 %arrayidx10 = getelementptr inbounds [5 x i32], [5 x i32] addrspace(5)* %stack, i32 0, i32 0
140 %2 = load i32, i32 addrspace(5)* %arrayidx10, align 4
141 store i32 %2, i32 addrspace(1)* %out, align 4
142 %arrayidx12 = getelementptr inbounds [5 x i32], [5 x i32] addrspace(5)* %stack, i32 0, i32 1
143 %3 = load i32, i32 addrspace(5)* %arrayidx12
144 %arrayidx13 = getelementptr inbounds i32, i32 addrspace(1)* %out, i32 1
145 store i32 %3, i32 addrspace(1)* %arrayidx13
149 ; This test checks that the stack offset is calculated correctly for structs.
150 ; All register loads/stores should be optimized away, so there shouldn't be
151 ; any MOVA instructions.
153 ; XXX: This generated code has unnecessary MOVs, we should be able to optimize
156 ; FUNC-LABEL: {{^}}multiple_structs:
157 ; OPT-LABEL: @multiple_structs(
162 %struct.point = type { i32, i32 }
164 define amdgpu_kernel void @multiple_structs(i32 addrspace(1)* %out) #0 {
166 %a = alloca %struct.point, addrspace(5)
167 %b = alloca %struct.point, addrspace(5)
168 %a.x.ptr = getelementptr %struct.point, %struct.point addrspace(5)* %a, i32 0, i32 0
169 %a.y.ptr = getelementptr %struct.point, %struct.point addrspace(5)* %a, i32 0, i32 1
170 %b.x.ptr = getelementptr %struct.point, %struct.point addrspace(5)* %b, i32 0, i32 0
171 %b.y.ptr = getelementptr %struct.point, %struct.point addrspace(5)* %b, i32 0, i32 1
172 store i32 0, i32 addrspace(5)* %a.x.ptr
173 store i32 1, i32 addrspace(5)* %a.y.ptr
174 store i32 2, i32 addrspace(5)* %b.x.ptr
175 store i32 3, i32 addrspace(5)* %b.y.ptr
176 %a.indirect.ptr = getelementptr %struct.point, %struct.point addrspace(5)* %a, i32 0, i32 0
177 %b.indirect.ptr = getelementptr %struct.point, %struct.point addrspace(5)* %b, i32 0, i32 0
178 %a.indirect = load i32, i32 addrspace(5)* %a.indirect.ptr
179 %b.indirect = load i32, i32 addrspace(5)* %b.indirect.ptr
180 %0 = add i32 %a.indirect, %b.indirect
181 store i32 %0, i32 addrspace(1)* %out
185 ; Test direct access of a private array inside a loop. The private array
186 ; loads and stores should be lowered to copies, so there shouldn't be any
189 ; FUNC-LABEL: {{^}}direct_loop:
193 define amdgpu_kernel void @direct_loop(i32 addrspace(1)* %out, i32 addrspace(1)* %in) #0 {
195 %prv_array_const = alloca [2 x i32], addrspace(5)
196 %prv_array = alloca [2 x i32], addrspace(5)
197 %a = load i32, i32 addrspace(1)* %in
198 %b_src_ptr = getelementptr inbounds i32, i32 addrspace(1)* %in, i32 1
199 %b = load i32, i32 addrspace(1)* %b_src_ptr
200 %a_dst_ptr = getelementptr inbounds [2 x i32], [2 x i32] addrspace(5)* %prv_array_const, i32 0, i32 0
201 store i32 %a, i32 addrspace(5)* %a_dst_ptr
202 %b_dst_ptr = getelementptr inbounds [2 x i32], [2 x i32] addrspace(5)* %prv_array_const, i32 0, i32 1
203 store i32 %b, i32 addrspace(5)* %b_dst_ptr
207 %inc = phi i32 [0, %entry], [%count, %for.body]
208 %x_ptr = getelementptr inbounds [2 x i32], [2 x i32] addrspace(5)* %prv_array_const, i32 0, i32 0
209 %x = load i32, i32 addrspace(5)* %x_ptr
210 %y_ptr = getelementptr inbounds [2 x i32], [2 x i32] addrspace(5)* %prv_array, i32 0, i32 0
211 %y = load i32, i32 addrspace(5)* %y_ptr
213 store i32 %xy, i32 addrspace(5)* %y_ptr
214 %count = add i32 %inc, 1
215 %done = icmp eq i32 %count, 4095
216 br i1 %done, label %for.end, label %for.body
219 %value_ptr = getelementptr inbounds [2 x i32], [2 x i32] addrspace(5)* %prv_array, i32 0, i32 0
220 %value = load i32, i32 addrspace(5)* %value_ptr
221 store i32 %value, i32 addrspace(1)* %out
225 ; FUNC-LABEL: {{^}}short_array:
227 ; R600-VECT: MOVA_INT
229 ; SI-ALLOCA-DAG: buffer_store_short v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} offset:6 ; encoding: [0x06,0x00,0x68,0xe0
230 ; SI-ALLOCA-DAG: buffer_store_short v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} offset:4 ; encoding: [0x04,0x00,0x68,0xe0
231 ; Loaded value is 0 or 1, so sext will become zext, so we get buffer_load_ushort instead of buffer_load_sshort.
232 ; SI-ALLOCA: buffer_load_sshort v{{[0-9]+}}, v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}}
234 ; SI-PROMOTE-VECT: s_load_dword [[IDX:s[0-9]+]]
235 ; SI-PROMOTE-VECT: s_mov_b32 [[SREG:s[0-9]+]], 0x10000
236 ; SI-PROMOTE-VECT: s_lshl_b32 [[SCALED_IDX:s[0-9]+]], [[IDX]], 4
237 ; SI-PROMOTE-VECT: v_mov_b32_e32 [[VREG:v[0-9]+]], [[SCALED_IDX]]
238 ; SI-PROMOTE-VECT: v_bfe_u32 v{{[0-9]+}}, [[SREG]], [[VREG]], 16
239 define amdgpu_kernel void @short_array(i32 addrspace(1)* %out, i32 %index) #0 {
241 %0 = alloca [2 x i16], addrspace(5)
242 %1 = getelementptr inbounds [2 x i16], [2 x i16] addrspace(5)* %0, i32 0, i32 0
243 %2 = getelementptr inbounds [2 x i16], [2 x i16] addrspace(5)* %0, i32 0, i32 1
244 store i16 0, i16 addrspace(5)* %1
245 store i16 1, i16 addrspace(5)* %2
246 %3 = getelementptr inbounds [2 x i16], [2 x i16] addrspace(5)* %0, i32 0, i32 %index
247 %4 = load i16, i16 addrspace(5)* %3
248 %5 = sext i16 %4 to i32
249 store i32 %5, i32 addrspace(1)* %out
253 ; FUNC-LABEL: {{^}}char_array:
255 ; R600-VECT: MOVA_INT
257 ; SI-PROMOTE-VECT-DAG: s_lshl_b32
258 ; SI-PROMOTE-VECT-DAG: v_lshrrev
260 ; SI-ALLOCA-DAG: buffer_store_byte v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} offset:4 ; encoding: [0x04,0x00,0x60,0xe0
261 ; SI-ALLOCA-DAG: buffer_store_byte v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} offset:5 ; encoding: [0x05,0x00,0x60,0xe0
262 define amdgpu_kernel void @char_array(i32 addrspace(1)* %out, i32 %index) #0 {
264 %0 = alloca [2 x i8], addrspace(5)
265 %1 = getelementptr inbounds [2 x i8], [2 x i8] addrspace(5)* %0, i32 0, i32 0
266 %2 = getelementptr inbounds [2 x i8], [2 x i8] addrspace(5)* %0, i32 0, i32 1
267 store i8 0, i8 addrspace(5)* %1
268 store i8 1, i8 addrspace(5)* %2
269 %3 = getelementptr inbounds [2 x i8], [2 x i8] addrspace(5)* %0, i32 0, i32 %index
270 %4 = load i8, i8 addrspace(5)* %3
271 %5 = sext i8 %4 to i32
272 store i32 %5, i32 addrspace(1)* %out
276 ; Test that two stack objects are not stored in the same register
277 ; The second stack object should be in T3.X
278 ; FUNC-LABEL: {{^}}no_overlap:
280 ; R600-CHECK: [[CHAN:[XYZW]]]+
281 ; R600-NOT: [[CHAN]]+
283 ; A total of 5 bytes should be allocated and used.
284 ; SI: buffer_store_byte v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} offset:4 ;
285 define amdgpu_kernel void @no_overlap(i32 addrspace(1)* %out, i32 %in) #0 {
287 %0 = alloca [3 x i8], align 1, addrspace(5)
288 %1 = alloca [2 x i8], align 1, addrspace(5)
289 %2 = getelementptr [3 x i8], [3 x i8] addrspace(5)* %0, i32 0, i32 0
290 %3 = getelementptr [3 x i8], [3 x i8] addrspace(5)* %0, i32 0, i32 1
291 %4 = getelementptr [3 x i8], [3 x i8] addrspace(5)* %0, i32 0, i32 2
292 %5 = getelementptr [2 x i8], [2 x i8] addrspace(5)* %1, i32 0, i32 0
293 %6 = getelementptr [2 x i8], [2 x i8] addrspace(5)* %1, i32 0, i32 1
294 store i8 0, i8 addrspace(5)* %2
295 store i8 1, i8 addrspace(5)* %3
296 store i8 2, i8 addrspace(5)* %4
297 store i8 1, i8 addrspace(5)* %5
298 store i8 0, i8 addrspace(5)* %6
299 %7 = getelementptr [3 x i8], [3 x i8] addrspace(5)* %0, i32 0, i32 %in
300 %8 = getelementptr [2 x i8], [2 x i8] addrspace(5)* %1, i32 0, i32 %in
301 %9 = load i8, i8 addrspace(5)* %7
302 %10 = load i8, i8 addrspace(5)* %8
304 %12 = sext i8 %11 to i32
305 store i32 %12, i32 addrspace(1)* %out
309 define amdgpu_kernel void @char_array_array(i32 addrspace(1)* %out, i32 %index) #0 {
311 %alloca = alloca [2 x [2 x i8]], addrspace(5)
312 %gep0 = getelementptr [2 x [2 x i8]], [2 x [2 x i8]] addrspace(5)* %alloca, i32 0, i32 0, i32 0
313 %gep1 = getelementptr [2 x [2 x i8]], [2 x [2 x i8]] addrspace(5)* %alloca, i32 0, i32 0, i32 1
314 store i8 0, i8 addrspace(5)* %gep0
315 store i8 1, i8 addrspace(5)* %gep1
316 %gep2 = getelementptr [2 x [2 x i8]], [2 x [2 x i8]] addrspace(5)* %alloca, i32 0, i32 0, i32 %index
317 %load = load i8, i8 addrspace(5)* %gep2
318 %sext = sext i8 %load to i32
319 store i32 %sext, i32 addrspace(1)* %out
323 define amdgpu_kernel void @i32_array_array(i32 addrspace(1)* %out, i32 %index) #0 {
325 %alloca = alloca [2 x [2 x i32]], addrspace(5)
326 %gep0 = getelementptr [2 x [2 x i32]], [2 x [2 x i32]] addrspace(5)* %alloca, i32 0, i32 0, i32 0
327 %gep1 = getelementptr [2 x [2 x i32]], [2 x [2 x i32]] addrspace(5)* %alloca, i32 0, i32 0, i32 1
328 store i32 0, i32 addrspace(5)* %gep0
329 store i32 1, i32 addrspace(5)* %gep1
330 %gep2 = getelementptr [2 x [2 x i32]], [2 x [2 x i32]] addrspace(5)* %alloca, i32 0, i32 0, i32 %index
331 %load = load i32, i32 addrspace(5)* %gep2
332 store i32 %load, i32 addrspace(1)* %out
336 define amdgpu_kernel void @i64_array_array(i64 addrspace(1)* %out, i32 %index) #0 {
338 %alloca = alloca [2 x [2 x i64]], addrspace(5)
339 %gep0 = getelementptr [2 x [2 x i64]], [2 x [2 x i64]] addrspace(5)* %alloca, i32 0, i32 0, i32 0
340 %gep1 = getelementptr [2 x [2 x i64]], [2 x [2 x i64]] addrspace(5)* %alloca, i32 0, i32 0, i32 1
341 store i64 0, i64 addrspace(5)* %gep0
342 store i64 1, i64 addrspace(5)* %gep1
343 %gep2 = getelementptr [2 x [2 x i64]], [2 x [2 x i64]] addrspace(5)* %alloca, i32 0, i32 0, i32 %index
344 %load = load i64, i64 addrspace(5)* %gep2
345 store i64 %load, i64 addrspace(1)* %out
349 %struct.pair32 = type { i32, i32 }
351 define amdgpu_kernel void @struct_array_array(i32 addrspace(1)* %out, i32 %index) #0 {
353 %alloca = alloca [2 x [2 x %struct.pair32]], addrspace(5)
354 %gep0 = getelementptr [2 x [2 x %struct.pair32]], [2 x [2 x %struct.pair32]] addrspace(5)* %alloca, i32 0, i32 0, i32 0, i32 1
355 %gep1 = getelementptr [2 x [2 x %struct.pair32]], [2 x [2 x %struct.pair32]] addrspace(5)* %alloca, i32 0, i32 0, i32 1, i32 1
356 store i32 0, i32 addrspace(5)* %gep0
357 store i32 1, i32 addrspace(5)* %gep1
358 %gep2 = getelementptr [2 x [2 x %struct.pair32]], [2 x [2 x %struct.pair32]] addrspace(5)* %alloca, i32 0, i32 0, i32 %index, i32 0
359 %load = load i32, i32 addrspace(5)* %gep2
360 store i32 %load, i32 addrspace(1)* %out
364 define amdgpu_kernel void @struct_pair32_array(i32 addrspace(1)* %out, i32 %index) #0 {
366 %alloca = alloca [2 x %struct.pair32], addrspace(5)
367 %gep0 = getelementptr [2 x %struct.pair32], [2 x %struct.pair32] addrspace(5)* %alloca, i32 0, i32 0, i32 1
368 %gep1 = getelementptr [2 x %struct.pair32], [2 x %struct.pair32] addrspace(5)* %alloca, i32 0, i32 1, i32 0
369 store i32 0, i32 addrspace(5)* %gep0
370 store i32 1, i32 addrspace(5)* %gep1
371 %gep2 = getelementptr [2 x %struct.pair32], [2 x %struct.pair32] addrspace(5)* %alloca, i32 0, i32 %index, i32 0
372 %load = load i32, i32 addrspace(5)* %gep2
373 store i32 %load, i32 addrspace(1)* %out
377 define amdgpu_kernel void @select_private(i32 addrspace(1)* %out, i32 %in) nounwind {
379 %tmp = alloca [2 x i32], addrspace(5)
380 %tmp1 = getelementptr [2 x i32], [2 x i32] addrspace(5)* %tmp, i32 0, i32 0
381 %tmp2 = getelementptr [2 x i32], [2 x i32] addrspace(5)* %tmp, i32 0, i32 1
382 store i32 0, i32 addrspace(5)* %tmp1
383 store i32 1, i32 addrspace(5)* %tmp2
384 %cmp = icmp eq i32 %in, 0
385 %sel = select i1 %cmp, i32 addrspace(5)* %tmp1, i32 addrspace(5)* %tmp2
386 %load = load i32, i32 addrspace(5)* %sel
387 store i32 %load, i32 addrspace(1)* %out
391 ; AMDGPUPromoteAlloca does not know how to handle ptrtoint. When it
392 ; finds one, it should stop trying to promote.
394 ; FUNC-LABEL: ptrtoint:
396 ; SI: buffer_store_dword v{{[0-9]+}}, v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} offen
397 ; SI: v_add_{{[iu]}}32_e32 [[ADD_OFFSET:v[0-9]+]], vcc, 5,
398 ; SI: buffer_load_dword v{{[0-9]+}}, [[ADD_OFFSET:v[0-9]+]], s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} offen ;
399 define amdgpu_kernel void @ptrtoint(i32 addrspace(1)* %out, i32 %a, i32 %b) #0 {
400 %alloca = alloca [16 x i32], addrspace(5)
401 %tmp0 = getelementptr [16 x i32], [16 x i32] addrspace(5)* %alloca, i32 0, i32 %a
402 store i32 5, i32 addrspace(5)* %tmp0
403 %tmp1 = ptrtoint [16 x i32] addrspace(5)* %alloca to i32
404 %tmp2 = add i32 %tmp1, 5
405 %tmp3 = inttoptr i32 %tmp2 to i32 addrspace(5)*
406 %tmp4 = getelementptr i32, i32 addrspace(5)* %tmp3, i32 %b
407 %tmp5 = load i32, i32 addrspace(5)* %tmp4
408 store i32 %tmp5, i32 addrspace(1)* %out
412 ; OPT-LABEL: @pointer_typed_alloca(
413 ; OPT: getelementptr inbounds [256 x i32 addrspace(1)*], [256 x i32 addrspace(1)*] addrspace(3)* @pointer_typed_alloca.A.addr, i32 0, i32 %{{[0-9]+}}
414 ; OPT: load i32 addrspace(1)*, i32 addrspace(1)* addrspace(3)* %{{[0-9]+}}, align 4
415 define amdgpu_kernel void @pointer_typed_alloca(i32 addrspace(1)* %A) {
417 %A.addr = alloca i32 addrspace(1)*, align 4, addrspace(5)
418 store i32 addrspace(1)* %A, i32 addrspace(1)* addrspace(5)* %A.addr, align 4
419 %ld0 = load i32 addrspace(1)*, i32 addrspace(1)* addrspace(5)* %A.addr, align 4
420 %arrayidx = getelementptr inbounds i32, i32 addrspace(1)* %ld0, i32 0
421 store i32 1, i32 addrspace(1)* %arrayidx, align 4
422 %ld1 = load i32 addrspace(1)*, i32 addrspace(1)* addrspace(5)* %A.addr, align 4
423 %arrayidx1 = getelementptr inbounds i32, i32 addrspace(1)* %ld1, i32 1
424 store i32 2, i32 addrspace(1)* %arrayidx1, align 4
425 %ld2 = load i32 addrspace(1)*, i32 addrspace(1)* addrspace(5)* %A.addr, align 4
426 %arrayidx2 = getelementptr inbounds i32, i32 addrspace(1)* %ld2, i32 2
427 store i32 3, i32 addrspace(1)* %arrayidx2, align 4
431 ; FUNC-LABEL: v16i32_stack:
450 ; SI: buffer_load_dword
451 ; SI: buffer_load_dword
452 ; SI: buffer_load_dword
453 ; SI: buffer_load_dword
454 ; SI: buffer_load_dword
455 ; SI: buffer_load_dword
456 ; SI: buffer_load_dword
457 ; SI: buffer_load_dword
458 ; SI: buffer_load_dword
459 ; SI: buffer_load_dword
460 ; SI: buffer_load_dword
461 ; SI: buffer_load_dword
462 ; SI: buffer_load_dword
463 ; SI: buffer_load_dword
464 ; SI: buffer_load_dword
465 ; SI: buffer_load_dword
467 define amdgpu_kernel void @v16i32_stack(<16 x i32> addrspace(1)* %out, i32 %a) {
468 %alloca = alloca [2 x <16 x i32>], addrspace(5)
469 %tmp0 = getelementptr [2 x <16 x i32>], [2 x <16 x i32>] addrspace(5)* %alloca, i32 0, i32 %a
470 %tmp5 = load <16 x i32>, <16 x i32> addrspace(5)* %tmp0
471 store <16 x i32> %tmp5, <16 x i32> addrspace(1)* %out
475 ; FUNC-LABEL: v16float_stack:
494 ; SI: buffer_load_dword
495 ; SI: buffer_load_dword
496 ; SI: buffer_load_dword
497 ; SI: buffer_load_dword
498 ; SI: buffer_load_dword
499 ; SI: buffer_load_dword
500 ; SI: buffer_load_dword
501 ; SI: buffer_load_dword
502 ; SI: buffer_load_dword
503 ; SI: buffer_load_dword
504 ; SI: buffer_load_dword
505 ; SI: buffer_load_dword
506 ; SI: buffer_load_dword
507 ; SI: buffer_load_dword
508 ; SI: buffer_load_dword
509 ; SI: buffer_load_dword
511 define amdgpu_kernel void @v16float_stack(<16 x float> addrspace(1)* %out, i32 %a) {
512 %alloca = alloca [2 x <16 x float>], addrspace(5)
513 %tmp0 = getelementptr [2 x <16 x float>], [2 x <16 x float>] addrspace(5)* %alloca, i32 0, i32 %a
514 %tmp5 = load <16 x float>, <16 x float> addrspace(5)* %tmp0
515 store <16 x float> %tmp5, <16 x float> addrspace(1)* %out
519 ; FUNC-LABEL: v2float_stack:
524 ; SI: buffer_load_dword
525 ; SI: buffer_load_dword
527 define amdgpu_kernel void @v2float_stack(<2 x float> addrspace(1)* %out, i32 %a) {
528 %alloca = alloca [16 x <2 x float>], addrspace(5)
529 %tmp0 = getelementptr [16 x <2 x float>], [16 x <2 x float>] addrspace(5)* %alloca, i32 0, i32 %a
530 %tmp5 = load <2 x float>, <2 x float> addrspace(5)* %tmp0
531 store <2 x float> %tmp5, <2 x float> addrspace(1)* %out
535 ; OPT-LABEL: @direct_alloca_read_0xi32(
536 ; OPT: store [0 x i32] undef, [0 x i32] addrspace(3)*
537 ; OPT: load [0 x i32], [0 x i32] addrspace(3)*
538 define amdgpu_kernel void @direct_alloca_read_0xi32([0 x i32] addrspace(1)* %out, i32 %index) {
540 %tmp = alloca [0 x i32], addrspace(5)
541 store [0 x i32] [], [0 x i32] addrspace(5)* %tmp
542 %load = load [0 x i32], [0 x i32] addrspace(5)* %tmp
543 store [0 x i32] %load, [0 x i32] addrspace(1)* %out
547 ; OPT-LABEL: @direct_alloca_read_1xi32(
548 ; OPT: store [1 x i32] zeroinitializer, [1 x i32] addrspace(3)*
549 ; OPT: load [1 x i32], [1 x i32] addrspace(3)*
550 define amdgpu_kernel void @direct_alloca_read_1xi32([1 x i32] addrspace(1)* %out, i32 %index) {
552 %tmp = alloca [1 x i32], addrspace(5)
553 store [1 x i32] [i32 0], [1 x i32] addrspace(5)* %tmp
554 %load = load [1 x i32], [1 x i32] addrspace(5)* %tmp
555 store [1 x i32] %load, [1 x i32] addrspace(1)* %out
559 attributes #0 = { nounwind "amdgpu-waves-per-eu"="1,2" }
562 ; HSAOPT: !1 = !{i32 0, i32 257}
563 ; HSAOPT: !2 = !{i32 0, i32 256}
565 ; NOHSAOPT: !0 = !{i32 0, i32 257}
566 ; NOHSAOPT: !1 = !{i32 0, i32 256}