1; RUN: llc -verify-machineinstrs -mtriple=amdgcn-amd-amdhsa -mcpu=kaveri -mattr=-flat-for-global,+max-private-element-size-16 < %s | FileCheck -enable-var-scope -check-prefixes=GCN,SI,GCN-NO-TONGA %s
2; RUN: llc -verify-machineinstrs -mtriple=amdgcn-amd-amdhsa -mcpu=tonga -mattr=-flat-for-global -mattr=+max-private-element-size-16 < %s | FileCheck -enable-var-scope -check-prefixes=GCN,VI,GCN-TONGA %s
3
4; FIXME: Broken on evergreen
5; FIXME: For some reason the 8 and 16 vectors are being stored as
6; individual elements instead of 128-bit stores.
7
8
9; FIXME: Why is the constant moved into the intermediate register and
10; not just directly into the vector component?
11
12; GCN-LABEL: {{^}}insertelement_v4f32_0:
13; GCN: s_load_dwordx4
14; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}}
15; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}}
16; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}}
17; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}}
18; GCN-DAG: s_mov_b32 [[CONSTREG:s[0-9]+]], 0x40a00000
19; GCN-DAG: v_mov_b32_e32 v[[LOW_REG:[0-9]+]], [[CONSTREG]]
20; GCN: buffer_store_dwordx4 v{{\[}}[[LOW_REG]]:
21define amdgpu_kernel void @insertelement_v4f32_0(<4 x float> addrspace(1)* %out, <4 x float> %a) nounwind {
22  %vecins = insertelement <4 x float> %a, float 5.000000e+00, i32 0
23  store <4 x float> %vecins, <4 x float> addrspace(1)* %out, align 16
24  ret void
25}
26
27; GCN-LABEL: {{^}}insertelement_v4f32_1:
28define amdgpu_kernel void @insertelement_v4f32_1(<4 x float> addrspace(1)* %out, <4 x float> %a) nounwind {
29  %vecins = insertelement <4 x float> %a, float 5.000000e+00, i32 1
30  store <4 x float> %vecins, <4 x float> addrspace(1)* %out, align 16
31  ret void
32}
33
34; GCN-LABEL: {{^}}insertelement_v4f32_2:
35define amdgpu_kernel void @insertelement_v4f32_2(<4 x float> addrspace(1)* %out, <4 x float> %a) nounwind {
36  %vecins = insertelement <4 x float> %a, float 5.000000e+00, i32 2
37  store <4 x float> %vecins, <4 x float> addrspace(1)* %out, align 16
38  ret void
39}
40
41; GCN-LABEL: {{^}}insertelement_v4f32_3:
42define amdgpu_kernel void @insertelement_v4f32_3(<4 x float> addrspace(1)* %out, <4 x float> %a) nounwind {
43  %vecins = insertelement <4 x float> %a, float 5.000000e+00, i32 3
44  store <4 x float> %vecins, <4 x float> addrspace(1)* %out, align 16
45  ret void
46}
47
48; GCN-LABEL: {{^}}insertelement_v4i32_0:
49define amdgpu_kernel void @insertelement_v4i32_0(<4 x i32> addrspace(1)* %out, <4 x i32> %a) nounwind {
50  %vecins = insertelement <4 x i32> %a, i32 999, i32 0
51  store <4 x i32> %vecins, <4 x i32> addrspace(1)* %out, align 16
52  ret void
53}
54
55; GCN-LABEL: {{^}}insertelement_v3f32_1:
56define amdgpu_kernel void @insertelement_v3f32_1(<3 x float> addrspace(1)* %out, <3 x float> %a) nounwind {
57  %vecins = insertelement <3 x float> %a, float 5.000000e+00, i32 1
58  store <3 x float> %vecins, <3 x float> addrspace(1)* %out, align 16
59  ret void
60}
61
62; GCN-LABEL: {{^}}insertelement_v3f32_2:
63define amdgpu_kernel void @insertelement_v3f32_2(<3 x float> addrspace(1)* %out, <3 x float> %a) nounwind {
64  %vecins = insertelement <3 x float> %a, float 5.000000e+00, i32 2
65  store <3 x float> %vecins, <3 x float> addrspace(1)* %out, align 16
66  ret void
67}
68
69; GCN-LABEL: {{^}}insertelement_v3f32_3:
70define amdgpu_kernel void @insertelement_v3f32_3(<3 x float> addrspace(1)* %out, <3 x float> %a) nounwind {
71  %vecins = insertelement <3 x float> %a, float 5.000000e+00, i32 3
72  store <3 x float> %vecins, <3 x float> addrspace(1)* %out, align 16
73  ret void
74}
75
76; GCN-LABEL: {{^}}insertelement_to_sgpr:
77; GCN-NOT: v_readfirstlane
78define <4 x float> @insertelement_to_sgpr() nounwind {
79  %tmp = load <4 x i32>, <4 x i32> addrspace(4)* undef
80  %tmp1 = insertelement <4 x i32> %tmp, i32 0, i32 0
81  %tmp2 = call <4 x float> @llvm.amdgcn.image.gather4.lz.2d.v4f32.f32(i32 1, float undef, float undef, <8 x i32> undef, <4 x i32> %tmp1, i1 0, i32 0, i32 0)
82  ret <4 x float> %tmp2
83}
84
85; GCN-LABEL: {{^}}dynamic_insertelement_v2f32:
86; GCN: v_mov_b32_e32 [[CONST:v[0-9]+]], 0x40a00000
87; GCN: v_movreld_b32_e32 v[[LOW_RESULT_REG:[0-9]+]], [[CONST]]
88; GCN: buffer_store_dwordx2 {{v\[}}[[LOW_RESULT_REG]]:
89define amdgpu_kernel void @dynamic_insertelement_v2f32(<2 x float> addrspace(1)* %out, <2 x float> %a, i32 %b) nounwind {
90  %vecins = insertelement <2 x float> %a, float 5.000000e+00, i32 %b
91  store <2 x float> %vecins, <2 x float> addrspace(1)* %out, align 8
92  ret void
93}
94
95; GCN-LABEL: {{^}}dynamic_insertelement_v3f32:
96; GCN: v_mov_b32_e32 [[CONST:v[0-9]+]], 0x40a00000
97; GCN: v_movreld_b32_e32 v[[LOW_RESULT_REG:[0-9]+]], [[CONST]]
98; GCN-DAG: buffer_store_dwordx2 {{v\[}}[[LOW_RESULT_REG]]:
99; GCN-DAG: buffer_store_dword v
100define amdgpu_kernel void @dynamic_insertelement_v3f32(<3 x float> addrspace(1)* %out, <3 x float> %a, i32 %b) nounwind {
101  %vecins = insertelement <3 x float> %a, float 5.000000e+00, i32 %b
102  store <3 x float> %vecins, <3 x float> addrspace(1)* %out, align 16
103  ret void
104}
105
106; GCN-LABEL: {{^}}dynamic_insertelement_v4f32:
107; GCN: v_mov_b32_e32 [[CONST:v[0-9]+]], 0x40a00000
108; GCN: v_movreld_b32_e32 v[[LOW_RESULT_REG:[0-9]+]], [[CONST]]
109; GCN: buffer_store_dwordx4 {{v\[}}[[LOW_RESULT_REG]]:
110define amdgpu_kernel void @dynamic_insertelement_v4f32(<4 x float> addrspace(1)* %out, <4 x float> %a, i32 %b) nounwind {
111  %vecins = insertelement <4 x float> %a, float 5.000000e+00, i32 %b
112  store <4 x float> %vecins, <4 x float> addrspace(1)* %out, align 16
113  ret void
114}
115
116; GCN-LABEL: {{^}}dynamic_insertelement_v8f32:
117; GCN: v_movreld_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}
118; GCN: buffer_store_dwordx4
119; GCN: buffer_store_dwordx4
120define amdgpu_kernel void @dynamic_insertelement_v8f32(<8 x float> addrspace(1)* %out, <8 x float> %a, i32 %b) nounwind {
121  %vecins = insertelement <8 x float> %a, float 5.000000e+00, i32 %b
122  store <8 x float> %vecins, <8 x float> addrspace(1)* %out, align 32
123  ret void
124}
125
126; GCN-LABEL: {{^}}dynamic_insertelement_v16f32:
127; GCN: v_movreld_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}
128; GCN: buffer_store_dwordx4
129; GCN: buffer_store_dwordx4
130; GCN: buffer_store_dwordx4
131; GCN: buffer_store_dwordx4
132define amdgpu_kernel void @dynamic_insertelement_v16f32(<16 x float> addrspace(1)* %out, <16 x float> %a, i32 %b) nounwind {
133  %vecins = insertelement <16 x float> %a, float 5.000000e+00, i32 %b
134  store <16 x float> %vecins, <16 x float> addrspace(1)* %out, align 64
135  ret void
136}
137
138; GCN-LABEL: {{^}}dynamic_insertelement_v2i32:
139; GCN: v_movreld_b32
140; GCN: buffer_store_dwordx2
141define amdgpu_kernel void @dynamic_insertelement_v2i32(<2 x i32> addrspace(1)* %out, <2 x i32> %a, i32 %b) nounwind {
142  %vecins = insertelement <2 x i32> %a, i32 5, i32 %b
143  store <2 x i32> %vecins, <2 x i32> addrspace(1)* %out, align 8
144  ret void
145}
146
147; GCN-LABEL: {{^}}dynamic_insertelement_v3i32:
148; GCN: v_movreld_b32_e32 v[[LOW_RESULT_REG:[0-9]+]], 5
149; GCN-DAG: buffer_store_dwordx2 {{v\[}}[[LOW_RESULT_REG]]:
150; GCN-DAG: buffer_store_dword v
151define amdgpu_kernel void @dynamic_insertelement_v3i32(<3 x i32> addrspace(1)* %out, <3 x i32> %a, i32 %b) nounwind {
152  %vecins = insertelement <3 x i32> %a, i32 5, i32 %b
153  store <3 x i32> %vecins, <3 x i32> addrspace(1)* %out, align 16
154  ret void
155}
156
157; GCN-LABEL: {{^}}dynamic_insertelement_v4i32:
158; GCN: s_load_dword [[SVAL:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, {{0x11|0x44}}
159; GCN: v_mov_b32_e32 [[VVAL:v[0-9]+]], [[SVAL]]
160; GCN: v_movreld_b32_e32 v{{[0-9]+}}, [[VVAL]]
161; GCN: buffer_store_dwordx4
162define amdgpu_kernel void @dynamic_insertelement_v4i32(<4 x i32> addrspace(1)* %out, <4 x i32> %a, i32 %b, [8 x i32], i32 %val) nounwind {
163  %vecins = insertelement <4 x i32> %a, i32 %val, i32 %b
164  store <4 x i32> %vecins, <4 x i32> addrspace(1)* %out, align 16
165  ret void
166}
167
168; GCN-LABEL: {{^}}dynamic_insertelement_v8i32:
169; GCN: v_movreld_b32
170; GCN: buffer_store_dwordx4
171; GCN: buffer_store_dwordx4
172define amdgpu_kernel void @dynamic_insertelement_v8i32(<8 x i32> addrspace(1)* %out, <8 x i32> %a, i32 %b) nounwind {
173  %vecins = insertelement <8 x i32> %a, i32 5, i32 %b
174  store <8 x i32> %vecins, <8 x i32> addrspace(1)* %out, align 32
175  ret void
176}
177
178; GCN-LABEL: {{^}}dynamic_insertelement_v16i32:
179; GCN: v_movreld_b32
180; GCN: buffer_store_dwordx4
181; GCN: buffer_store_dwordx4
182; GCN: buffer_store_dwordx4
183; GCN: buffer_store_dwordx4
184define amdgpu_kernel void @dynamic_insertelement_v16i32(<16 x i32> addrspace(1)* %out, <16 x i32> %a, i32 %b) nounwind {
185  %vecins = insertelement <16 x i32> %a, i32 5, i32 %b
186  store <16 x i32> %vecins, <16 x i32> addrspace(1)* %out, align 64
187  ret void
188}
189
190; GCN-LABEL: {{^}}dynamic_insertelement_v2i16:
191define amdgpu_kernel void @dynamic_insertelement_v2i16(<2 x i16> addrspace(1)* %out, <2 x i16> %a, i32 %b) nounwind {
192  %vecins = insertelement <2 x i16> %a, i16 5, i32 %b
193  store <2 x i16> %vecins, <2 x i16> addrspace(1)* %out, align 8
194  ret void
195}
196
197; GCN-LABEL: {{^}}dynamic_insertelement_v3i16:
198define amdgpu_kernel void @dynamic_insertelement_v3i16(<3 x i16> addrspace(1)* %out, <3 x i16> %a, i32 %b) nounwind {
199  %vecins = insertelement <3 x i16> %a, i16 5, i32 %b
200  store <3 x i16> %vecins, <3 x i16> addrspace(1)* %out, align 8
201  ret void
202}
203
204; GCN-LABEL: {{^}}dynamic_insertelement_v2i8:
205; VI: s_load_dword [[LOAD:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0x28
206; VI-NEXT: s_load_dword [[IDX:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0x4c
207; VI-NOT: _load
208; VI: s_lshl_b32 [[SCALED_IDX:s[0-9]+]], [[IDX]], 3
209; VI: v_lshlrev_b16_e64 [[MASK:v[0-9]+]], [[SCALED_IDX]], -1
210; VI: v_and_b32_e32 [[INSERT:v[0-9]+]], 5, [[MASK]]
211; VI: v_xor_b32_e32 [[NOT_MASK:v[0-9]+]], -1, [[MASK]]
212; VI: v_and_b32_e32 [[AND_NOT_MASK:v[0-9]+]], [[LOAD]], [[NOT_MASK]]
213; VI: v_or_b32_e32 [[OR:v[0-9]+]], [[INSERT]], [[AND_NOT_MASK]]
214; VI: buffer_store_short [[OR]]
215define amdgpu_kernel void @dynamic_insertelement_v2i8(<2 x i8> addrspace(1)* %out, [8 x i32], <2 x i8> %a, [8 x i32], i32 %b) nounwind {
216  %vecins = insertelement <2 x i8> %a, i8 5, i32 %b
217  store <2 x i8> %vecins, <2 x i8> addrspace(1)* %out, align 8
218  ret void
219}
220
221; FIXME: post legalize i16 and i32 shifts aren't merged because of
222; isTypeDesirableForOp in SimplifyDemandedBits
223
224; GCN-LABEL: {{^}}dynamic_insertelement_v3i8:
225; VI: s_load_dword [[LOAD:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0x28
226; VI-NEXT: s_load_dword [[IDX:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0x4c
227; VI-NOT: _load
228
229; VI: v_mov_b32_e32 [[V_LOAD:v[0-9]+]], [[LOAD]]
230; VI: s_lshl_b32 [[SCALED_IDX:s[0-9]+]], [[IDX]], 3
231; VI: s_lshl_b32 [[SHIFTED_MASK:s[0-9]+]], 0xffff, [[SCALED_IDX]]
232; VI: s_not_b32 [[NOT_MASK:s[0-9]+]], [[SHIFTED_MASK]]
233; VI: s_and_b32 [[AND_NOT_MASK:s[0-9]+]], [[NOT_MASK]], [[LOAD]]
234; VI: v_bfi_b32 [[BFI:v[0-9]+]], [[SHIFTED_MASK]], 5, [[V_LOAD]]
235; VI: s_lshr_b32 [[HI2:s[0-9]+]], [[AND_NOT_MASK]], 16
236
237; VI-DAG: buffer_store_short [[BFI]]
238; VI-DAG: v_mov_b32_e32 [[V_HI2:v[0-9]+]], [[HI2]]
239; VI: buffer_store_byte [[V_HI2]]
240define amdgpu_kernel void @dynamic_insertelement_v3i8(<3 x i8> addrspace(1)* %out, [8 x i32], <3 x i8> %a, [8 x i32], i32 %b) nounwind {
241  %vecins = insertelement <3 x i8> %a, i8 5, i32 %b
242  store <3 x i8> %vecins, <3 x i8> addrspace(1)* %out, align 4
243  ret void
244}
245
246; GCN-LABEL: {{^}}dynamic_insertelement_v4i8:
247; VI: s_load_dword [[LOAD:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0x28
248; VI-NEXT: s_load_dword [[IDX:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0x4c
249; VI-NOT: _load
250
251; VI: v_mov_b32_e32 [[V_LOAD:v[0-9]+]], [[LOAD]]
252; VI-DAG: s_lshl_b32 [[SCALED_IDX:s[0-9]+]], [[IDX]], 3
253; VI: s_lshl_b32 [[SHIFTED_MASK:s[0-9]+]], 0xffff, [[SCALED_IDX]]
254; VI: v_bfi_b32 [[BFI:v[0-9]+]], [[SHIFTED_MASK]], 5, [[V_LOAD]]
255; VI: buffer_store_dword [[BFI]]
256define amdgpu_kernel void @dynamic_insertelement_v4i8(<4 x i8> addrspace(1)* %out, [8 x i32], <4 x i8> %a, [8 x i32], i32 %b) nounwind {
257  %vecins = insertelement <4 x i8> %a, i8 5, i32 %b
258  store <4 x i8> %vecins, <4 x i8> addrspace(1)* %out, align 4
259  ret void
260}
261
262; GCN-LABEL: {{^}}s_dynamic_insertelement_v8i8:
263; VI-NOT: {{buffer|flat|global}}_load
264; VI-DAG: s_load_dwordx4 s{{\[[0-9]+:[0-9]+\]}}, s[4:5], 0x0
265; VI-DAG: s_load_dword [[IDX:s[0-9]]], s[4:5], 0x10
266; VI-DAG: s_mov_b32 s[[MASK_HI:[0-9]+]], 0{{$}}
267; VI-DAG: s_load_dwordx2 [[VEC:s\[[0-9]+:[0-9]+\]]], s{{\[[0-9]+:[0-9]+\]}}, 0x0
268
269; VI-DAG: s_lshl_b32 [[SCALED_IDX:s[0-9]+]], [[IDX]], 3
270; VI-DAG: s_mov_b32 s[[MASK_LO:[0-9]+]], 0xffff
271; VI: s_lshl_b64 s{{\[}}[[MASK_SHIFT_LO:[0-9]+]]:[[MASK_SHIFT_HI:[0-9]+]]{{\]}}, s{{\[}}[[MASK_LO]]:[[MASK_HI]]{{\]}}, [[SCALED_IDX]]
272; VI: s_not_b64 [[NOT_MASK:s\[[0-9]+:[0-9]+\]]], s{{\[}}[[MASK_SHIFT_LO]]:[[MASK_SHIFT_HI]]{{\]}}
273; VI: s_and_b64 [[AND:s\[[0-9]+:[0-9]+\]]], [[NOT_MASK]], [[VEC]]
274; VI: s_and_b32 s[[INS:[0-9]+]], s[[MASK_SHIFT_LO]], 5
275; VI: s_or_b64 s{{\[}}[[RESULT0:[0-9]+]]:[[RESULT1:[0-9]+]]{{\]}}, s{{\[}}[[INS]]:[[MASK_HI]]{{\]}}, [[AND]]
276; VI: v_mov_b32_e32 v[[V_RESULT0:[0-9]+]], s[[RESULT0]]
277; VI: v_mov_b32_e32 v[[V_RESULT1:[0-9]+]], s[[RESULT1]]
278; VI: buffer_store_dwordx2 v{{\[}}[[V_RESULT0]]:[[V_RESULT1]]{{\]}}
279define amdgpu_kernel void @s_dynamic_insertelement_v8i8(<8 x i8> addrspace(1)* %out, <8 x i8> addrspace(4)* %a.ptr, i32 %b) nounwind {
280  %a = load <8 x i8>, <8 x i8> addrspace(4)* %a.ptr, align 4
281  %vecins = insertelement <8 x i8> %a, i8 5, i32 %b
282  store <8 x i8> %vecins, <8 x i8> addrspace(1)* %out, align 8
283  ret void
284}
285
286; GCN-LABEL: {{^}}dynamic_insertelement_v16i8:
287; GCN: s_load_dwordx2
288; GCN: s_load_dwordx4
289; GCN: s_load_dword s
290
291; GCN: buffer_store_byte
292; GCN: buffer_store_byte
293; GCN: buffer_store_byte
294; GCN: buffer_store_byte
295; GCN: buffer_store_byte
296; GCN: buffer_store_byte
297; GCN: buffer_store_byte
298; GCN: buffer_store_byte
299; GCN: buffer_store_byte
300; GCN: buffer_store_byte
301; GCN: buffer_store_byte
302; GCN: buffer_store_byte
303; GCN: buffer_store_byte
304; GCN: buffer_store_byte
305; GCN: buffer_store_byte
306; GCN: buffer_store_byte
307
308; GCN: buffer_store_byte
309; GCN: buffer_store_dwordx4
310define amdgpu_kernel void @dynamic_insertelement_v16i8(<16 x i8> addrspace(1)* %out, <16 x i8> %a, i32 %b) nounwind {
311  %vecins = insertelement <16 x i8> %a, i8 5, i32 %b
312  store <16 x i8> %vecins, <16 x i8> addrspace(1)* %out, align 16
313  ret void
314}
315
316; This test requires handling INSERT_SUBREG in SIFixSGPRCopies.  Check that
317; the compiler doesn't crash.
318; GCN-LABEL: {{^}}insert_split_bb:
319define amdgpu_kernel void @insert_split_bb(<2 x i32> addrspace(1)* %out, i32 addrspace(1)* %in, i32 %a, i32 %b) {
320entry:
321  %0 = insertelement <2 x i32> undef, i32 %a, i32 0
322  %1 = icmp eq i32 %a, 0
323  br i1 %1, label %if, label %else
324
325if:
326  %2 = load i32, i32 addrspace(1)* %in
327  %3 = insertelement <2 x i32> %0, i32 %2, i32 1
328  br label %endif
329
330else:
331  %4 = getelementptr i32, i32 addrspace(1)* %in, i32 1
332  %5 = load i32, i32 addrspace(1)* %4
333  %6 = insertelement <2 x i32> %0, i32 %5, i32 1
334  br label %endif
335
336endif:
337  %7 = phi <2 x i32> [%3, %if], [%6, %else]
338  store <2 x i32> %7, <2 x i32> addrspace(1)* %out
339  ret void
340}
341
342; GCN-LABEL: {{^}}dynamic_insertelement_v2f64:
343; GCN-DAG: s_load_dwordx4 s{{\[}}[[A_ELT0:[0-9]+]]:[[A_ELT3:[0-9]+]]{{\]}}
344; GCN-DAG: s_load_dword [[IDX:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, {{0x18|0x60}}{{$}}
345
346; GCN-DAG: s_lshl_b32 [[SCALEDIDX:s[0-9]+]], [[IDX]], 1{{$}}
347
348; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}}
349; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}}
350; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}}
351; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}}
352; GCN-DAG: v_mov_b32_e32 [[ELT1:v[0-9]+]], 0x40200000
353
354; GCN-DAG: s_mov_b32 m0, [[SCALEDIDX]]
355; GCN-DAG: v_movreld_b32_e32 v{{[0-9]+}}, 0
356
357; Increment to next element folded into base register, but FileCheck
358; can't do math expressions
359
360; FIXME: Should be able to manipulate m0 directly instead of s_lshl_b32 + copy to m0
361
362; GCN: v_movreld_b32_e32 v{{[0-9]+}}, [[ELT1]]
363
364; GCN: buffer_store_dwordx4
365; GCN: s_endpgm
366define amdgpu_kernel void @dynamic_insertelement_v2f64(<2 x double> addrspace(1)* %out, [8 x i32], <2 x double> %a, [8 x i32], i32 %b) nounwind {
367  %vecins = insertelement <2 x double> %a, double 8.0, i32 %b
368  store <2 x double> %vecins, <2 x double> addrspace(1)* %out, align 16
369  ret void
370}
371
372; GCN-LABEL: {{^}}dynamic_insertelement_v2i64:
373
374; GCN-DAG: v_movreld_b32_e32 v{{[0-9]+}}, 5
375; GCN-DAG: v_movreld_b32_e32 v{{[0-9]+}}, 0
376
377; GCN: buffer_store_dwordx4
378; GCN: s_endpgm
379define amdgpu_kernel void @dynamic_insertelement_v2i64(<2 x i64> addrspace(1)* %out, <2 x i64> %a, i32 %b) nounwind {
380  %vecins = insertelement <2 x i64> %a, i64 5, i32 %b
381  store <2 x i64> %vecins, <2 x i64> addrspace(1)* %out, align 8
382  ret void
383}
384
385; GCN-LABEL: {{^}}dynamic_insertelement_v3i64:
386define amdgpu_kernel void @dynamic_insertelement_v3i64(<3 x i64> addrspace(1)* %out, <3 x i64> %a, i32 %b) nounwind {
387  %vecins = insertelement <3 x i64> %a, i64 5, i32 %b
388  store <3 x i64> %vecins, <3 x i64> addrspace(1)* %out, align 32
389  ret void
390}
391
392; FIXME: Should be able to do without stack access. The used stack
393; space is also 2x what should be required.
394
395; GCN-LABEL: {{^}}dynamic_insertelement_v4f64:
396
397; Stack store
398
399; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s[0:3], {{s[0-9]+}} offset:32{{$}}
400; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s[0:3], {{s[0-9]+}} offset:48{{$}}
401
402; Write element
403; GCN: buffer_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s[0:3], {{s[0-9]+}} offen{{$}}
404
405; Stack reload
406; GCN-DAG: buffer_load_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s[0:3], {{s[0-9]+}} offset:32{{$}}
407; GCN-DAG: buffer_load_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s[0:3], {{s[0-9]+}} offset:48{{$}}
408
409; Store result
410; GCN: buffer_store_dwordx4
411; GCN: buffer_store_dwordx4
412; GCN: s_endpgm
413; GCN: ScratchSize: 64
414
415define amdgpu_kernel void @dynamic_insertelement_v4f64(<4 x double> addrspace(1)* %out, <4 x double> %a, i32 %b) nounwind {
416  %vecins = insertelement <4 x double> %a, double 8.0, i32 %b
417  store <4 x double> %vecins, <4 x double> addrspace(1)* %out, align 16
418  ret void
419}
420
421; GCN-LABEL: {{^}}dynamic_insertelement_v8f64:
422; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s[0:3], {{s[0-9]+}} offset:64{{$}}
423; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s[0:3], {{s[0-9]+}} offset:80{{$}}
424; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s[0:3], {{s[0-9]+}} offset:96{{$}}
425; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s[0:3], {{s[0-9]+}} offset:112{{$}}
426
427; GCN: buffer_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s[0:3], {{s[0-9]+}} offen{{$}}
428
429; GCN-DAG: buffer_load_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s[0:3], {{s[0-9]+}} offset:64{{$}}
430; GCN-DAG: buffer_load_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s[0:3], {{s[0-9]+}} offset:80{{$}}
431; GCN-DAG: buffer_load_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s[0:3], {{s[0-9]+}} offset:96{{$}}
432; GCN-DAG: buffer_load_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s[0:3], {{s[0-9]+}} offset:112{{$}}
433
434; GCN: buffer_store_dwordx4
435; GCN: buffer_store_dwordx4
436; GCN: buffer_store_dwordx4
437; GCN: buffer_store_dwordx4
438; GCN: s_endpgm
439; GCN: ScratchSize: 128
440define amdgpu_kernel void @dynamic_insertelement_v8f64(<8 x double> addrspace(1)* %out, <8 x double> %a, i32 %b) #0 {
441  %vecins = insertelement <8 x double> %a, double 8.0, i32 %b
442  store <8 x double> %vecins, <8 x double> addrspace(1)* %out, align 16
443  ret void
444}
445
446declare <4 x float> @llvm.amdgcn.image.gather4.lz.2d.v4f32.f32(i32, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
447
448attributes #0 = { nounwind }
449attributes #1 = { nounwind readnone }
450