1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
2 ; RUN: llc -mtriple=r600 -mcpu=redwood < %s | FileCheck --check-prefixes=R600 %s
3 ; RUN: llc -mtriple=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefixes=SI %s
4 ; RUN: llc -mtriple=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX8 %s
5 ; RUN: llc -mtriple=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX10 %s
6 ; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX11 %s
8 define amdgpu_kernel void @rotr_i32(ptr addrspace(1) %in, i32 %x, i32 %y) {
9 ; R600-LABEL: rotr_i32:
10 ; R600: ; %bb.0: ; %entry
11 ; R600-NEXT: ALU 2, @4, KC0[CB0:0-32], KC1[]
12 ; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
15 ; R600-NEXT: ALU clause starting at 4:
16 ; R600-NEXT: LSHR * T0.X, KC0[2].Y, literal.x,
17 ; R600-NEXT: 2(2.802597e-45), 0(0.000000e+00)
18 ; R600-NEXT: BIT_ALIGN_INT * T1.X, KC0[2].Z, KC0[2].Z, KC0[2].W,
21 ; SI: ; %bb.0: ; %entry
22 ; SI-NEXT: s_load_dwordx4 s[0:3], s[2:3], 0x9
23 ; SI-NEXT: s_mov_b32 s7, 0xf000
24 ; SI-NEXT: s_mov_b32 s6, -1
25 ; SI-NEXT: s_waitcnt lgkmcnt(0)
26 ; SI-NEXT: s_mov_b32 s4, s0
27 ; SI-NEXT: s_mov_b32 s5, s1
28 ; SI-NEXT: v_mov_b32_e32 v0, s3
29 ; SI-NEXT: v_alignbit_b32 v0, s2, s2, v0
30 ; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0
33 ; GFX8-LABEL: rotr_i32:
34 ; GFX8: ; %bb.0: ; %entry
35 ; GFX8-NEXT: s_load_dwordx4 s[0:3], s[2:3], 0x24
36 ; GFX8-NEXT: s_waitcnt lgkmcnt(0)
37 ; GFX8-NEXT: v_mov_b32_e32 v0, s3
38 ; GFX8-NEXT: v_alignbit_b32 v2, s2, s2, v0
39 ; GFX8-NEXT: v_mov_b32_e32 v0, s0
40 ; GFX8-NEXT: v_mov_b32_e32 v1, s1
41 ; GFX8-NEXT: flat_store_dword v[0:1], v2
44 ; GFX10-LABEL: rotr_i32:
45 ; GFX10: ; %bb.0: ; %entry
46 ; GFX10-NEXT: s_load_dwordx4 s[4:7], s[2:3], 0x24
47 ; GFX10-NEXT: v_mov_b32_e32 v0, 0
48 ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
49 ; GFX10-NEXT: v_alignbit_b32 v1, s6, s6, s7
50 ; GFX10-NEXT: global_store_dword v0, v1, s[4:5]
51 ; GFX10-NEXT: s_endpgm
53 ; GFX11-LABEL: rotr_i32:
54 ; GFX11: ; %bb.0: ; %entry
55 ; GFX11-NEXT: s_load_b128 s[0:3], s[2:3], 0x24
56 ; GFX11-NEXT: v_mov_b32_e32 v0, 0
57 ; GFX11-NEXT: s_waitcnt lgkmcnt(0)
58 ; GFX11-NEXT: v_alignbit_b32 v1, s2, s2, s3
59 ; GFX11-NEXT: global_store_b32 v0, v1, s[0:1]
61 ; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
62 ; GFX11-NEXT: s_endpgm
64 %tmp0 = sub i32 32, %y
65 %tmp1 = shl i32 %x, %tmp0
66 %tmp2 = lshr i32 %x, %y
67 %tmp3 = or i32 %tmp1, %tmp2
68 store i32 %tmp3, ptr addrspace(1) %in
72 define amdgpu_kernel void @rotr_v2i32(ptr addrspace(1) %in, <2 x i32> %x, <2 x i32> %y) {
73 ; R600-LABEL: rotr_v2i32:
74 ; R600: ; %bb.0: ; %entry
75 ; R600-NEXT: ALU 3, @4, KC0[CB0:0-32], KC1[]
76 ; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
79 ; R600-NEXT: ALU clause starting at 4:
80 ; R600-NEXT: BIT_ALIGN_INT * T0.Y, KC0[3].X, KC0[3].X, KC0[3].Z,
81 ; R600-NEXT: BIT_ALIGN_INT * T0.X, KC0[2].W, KC0[2].W, KC0[3].Y,
82 ; R600-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
83 ; R600-NEXT: 2(2.802597e-45), 0(0.000000e+00)
85 ; SI-LABEL: rotr_v2i32:
86 ; SI: ; %bb.0: ; %entry
87 ; SI-NEXT: s_load_dwordx4 s[4:7], s[2:3], 0xb
88 ; SI-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x9
89 ; SI-NEXT: s_mov_b32 s3, 0xf000
90 ; SI-NEXT: s_mov_b32 s2, -1
91 ; SI-NEXT: s_waitcnt lgkmcnt(0)
92 ; SI-NEXT: v_mov_b32_e32 v0, s7
93 ; SI-NEXT: v_alignbit_b32 v1, s5, s5, v0
94 ; SI-NEXT: v_mov_b32_e32 v0, s6
95 ; SI-NEXT: v_alignbit_b32 v0, s4, s4, v0
96 ; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
99 ; GFX8-LABEL: rotr_v2i32:
100 ; GFX8: ; %bb.0: ; %entry
101 ; GFX8-NEXT: s_load_dwordx4 s[4:7], s[2:3], 0x2c
102 ; GFX8-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x24
103 ; GFX8-NEXT: s_waitcnt lgkmcnt(0)
104 ; GFX8-NEXT: v_mov_b32_e32 v0, s7
105 ; GFX8-NEXT: v_mov_b32_e32 v2, s6
106 ; GFX8-NEXT: v_alignbit_b32 v1, s5, s5, v0
107 ; GFX8-NEXT: v_alignbit_b32 v0, s4, s4, v2
108 ; GFX8-NEXT: v_mov_b32_e32 v3, s1
109 ; GFX8-NEXT: v_mov_b32_e32 v2, s0
110 ; GFX8-NEXT: flat_store_dwordx2 v[2:3], v[0:1]
111 ; GFX8-NEXT: s_endpgm
113 ; GFX10-LABEL: rotr_v2i32:
114 ; GFX10: ; %bb.0: ; %entry
115 ; GFX10-NEXT: s_clause 0x1
116 ; GFX10-NEXT: s_load_dwordx4 s[4:7], s[2:3], 0x2c
117 ; GFX10-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x24
118 ; GFX10-NEXT: v_mov_b32_e32 v2, 0
119 ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
120 ; GFX10-NEXT: v_alignbit_b32 v1, s5, s5, s7
121 ; GFX10-NEXT: v_alignbit_b32 v0, s4, s4, s6
122 ; GFX10-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1]
123 ; GFX10-NEXT: s_endpgm
125 ; GFX11-LABEL: rotr_v2i32:
126 ; GFX11: ; %bb.0: ; %entry
127 ; GFX11-NEXT: s_clause 0x1
128 ; GFX11-NEXT: s_load_b128 s[4:7], s[2:3], 0x2c
129 ; GFX11-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
130 ; GFX11-NEXT: v_mov_b32_e32 v2, 0
131 ; GFX11-NEXT: s_waitcnt lgkmcnt(0)
132 ; GFX11-NEXT: v_alignbit_b32 v1, s5, s5, s7
133 ; GFX11-NEXT: v_alignbit_b32 v0, s4, s4, s6
134 ; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1]
135 ; GFX11-NEXT: s_nop 0
136 ; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
137 ; GFX11-NEXT: s_endpgm
139 %tmp0 = sub <2 x i32> <i32 32, i32 32>, %y
140 %tmp1 = shl <2 x i32> %x, %tmp0
141 %tmp2 = lshr <2 x i32> %x, %y
142 %tmp3 = or <2 x i32> %tmp1, %tmp2
143 store <2 x i32> %tmp3, ptr addrspace(1) %in
147 define amdgpu_kernel void @rotr_v4i32(ptr addrspace(1) %in, <4 x i32> %x, <4 x i32> %y) {
148 ; R600-LABEL: rotr_v4i32:
149 ; R600: ; %bb.0: ; %entry
150 ; R600-NEXT: ALU 5, @4, KC0[CB0:0-32], KC1[]
151 ; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XYZW, T1.X, 1
154 ; R600-NEXT: ALU clause starting at 4:
155 ; R600-NEXT: BIT_ALIGN_INT * T0.W, KC0[4].X, KC0[4].X, KC0[5].X,
156 ; R600-NEXT: BIT_ALIGN_INT * T0.Z, KC0[3].W, KC0[3].W, KC0[4].W,
157 ; R600-NEXT: BIT_ALIGN_INT * T0.Y, KC0[3].Z, KC0[3].Z, KC0[4].Z,
158 ; R600-NEXT: BIT_ALIGN_INT * T0.X, KC0[3].Y, KC0[3].Y, KC0[4].Y,
159 ; R600-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
160 ; R600-NEXT: 2(2.802597e-45), 0(0.000000e+00)
162 ; SI-LABEL: rotr_v4i32:
163 ; SI: ; %bb.0: ; %entry
164 ; SI-NEXT: s_load_dwordx8 s[4:11], s[2:3], 0xd
165 ; SI-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x9
166 ; SI-NEXT: s_mov_b32 s3, 0xf000
167 ; SI-NEXT: s_mov_b32 s2, -1
168 ; SI-NEXT: s_waitcnt lgkmcnt(0)
169 ; SI-NEXT: v_mov_b32_e32 v0, s11
170 ; SI-NEXT: v_alignbit_b32 v3, s7, s7, v0
171 ; SI-NEXT: v_mov_b32_e32 v0, s10
172 ; SI-NEXT: v_alignbit_b32 v2, s6, s6, v0
173 ; SI-NEXT: v_mov_b32_e32 v0, s9
174 ; SI-NEXT: v_alignbit_b32 v1, s5, s5, v0
175 ; SI-NEXT: v_mov_b32_e32 v0, s8
176 ; SI-NEXT: v_alignbit_b32 v0, s4, s4, v0
177 ; SI-NEXT: buffer_store_dwordx4 v[0:3], off, s[0:3], 0
180 ; GFX8-LABEL: rotr_v4i32:
181 ; GFX8: ; %bb.0: ; %entry
182 ; GFX8-NEXT: s_load_dwordx8 s[4:11], s[2:3], 0x34
183 ; GFX8-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x24
184 ; GFX8-NEXT: s_waitcnt lgkmcnt(0)
185 ; GFX8-NEXT: v_mov_b32_e32 v0, s11
186 ; GFX8-NEXT: v_mov_b32_e32 v1, s10
187 ; GFX8-NEXT: v_mov_b32_e32 v4, s9
188 ; GFX8-NEXT: v_alignbit_b32 v3, s7, s7, v0
189 ; GFX8-NEXT: v_alignbit_b32 v2, s6, s6, v1
190 ; GFX8-NEXT: v_alignbit_b32 v1, s5, s5, v4
191 ; GFX8-NEXT: v_mov_b32_e32 v0, s8
192 ; GFX8-NEXT: v_mov_b32_e32 v5, s1
193 ; GFX8-NEXT: v_alignbit_b32 v0, s4, s4, v0
194 ; GFX8-NEXT: v_mov_b32_e32 v4, s0
195 ; GFX8-NEXT: flat_store_dwordx4 v[4:5], v[0:3]
196 ; GFX8-NEXT: s_endpgm
198 ; GFX10-LABEL: rotr_v4i32:
199 ; GFX10: ; %bb.0: ; %entry
200 ; GFX10-NEXT: s_clause 0x1
201 ; GFX10-NEXT: s_load_dwordx8 s[4:11], s[2:3], 0x34
202 ; GFX10-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x24
203 ; GFX10-NEXT: v_mov_b32_e32 v4, 0
204 ; GFX10-NEXT: s_waitcnt lgkmcnt(0)
205 ; GFX10-NEXT: v_alignbit_b32 v3, s7, s7, s11
206 ; GFX10-NEXT: v_alignbit_b32 v2, s6, s6, s10
207 ; GFX10-NEXT: v_alignbit_b32 v1, s5, s5, s9
208 ; GFX10-NEXT: v_alignbit_b32 v0, s4, s4, s8
209 ; GFX10-NEXT: global_store_dwordx4 v4, v[0:3], s[0:1]
210 ; GFX10-NEXT: s_endpgm
212 ; GFX11-LABEL: rotr_v4i32:
213 ; GFX11: ; %bb.0: ; %entry
214 ; GFX11-NEXT: s_clause 0x1
215 ; GFX11-NEXT: s_load_b256 s[4:11], s[2:3], 0x34
216 ; GFX11-NEXT: s_load_b64 s[0:1], s[2:3], 0x24
217 ; GFX11-NEXT: v_mov_b32_e32 v4, 0
218 ; GFX11-NEXT: s_waitcnt lgkmcnt(0)
219 ; GFX11-NEXT: v_alignbit_b32 v3, s7, s7, s11
220 ; GFX11-NEXT: v_alignbit_b32 v2, s6, s6, s10
221 ; GFX11-NEXT: v_alignbit_b32 v1, s5, s5, s9
222 ; GFX11-NEXT: v_alignbit_b32 v0, s4, s4, s8
223 ; GFX11-NEXT: global_store_b128 v4, v[0:3], s[0:1]
224 ; GFX11-NEXT: s_nop 0
225 ; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
226 ; GFX11-NEXT: s_endpgm
228 %tmp0 = sub <4 x i32> <i32 32, i32 32, i32 32, i32 32>, %y
229 %tmp1 = shl <4 x i32> %x, %tmp0
230 %tmp2 = lshr <4 x i32> %x, %y
231 %tmp3 = or <4 x i32> %tmp1, %tmp2
232 store <4 x i32> %tmp3, ptr addrspace(1) %in
236 declare i16 @llvm.fshr.i16(i16, i16, i16)
238 define void @test_rotr_i16(ptr addrspace(1) nocapture readonly %sourceA, ptr addrspace(1) nocapture readonly %sourceB, ptr addrspace(1) nocapture %destValues) {
239 ; R600-LABEL: test_rotr_i16:
240 ; R600: ; %bb.0: ; %entry
241 ; R600-NEXT: ALU 0, @12, KC0[CB0:0-32], KC1[]
242 ; R600-NEXT: TEX 0 @8
243 ; R600-NEXT: ALU 0, @13, KC0[CB0:0-32], KC1[]
244 ; R600-NEXT: TEX 0 @10
245 ; R600-NEXT: ALU 21, @14, KC0[CB0:0-32], KC1[]
246 ; R600-NEXT: MEM_RAT MSKOR T0.XW, T1.X
249 ; R600-NEXT: Fetch clause starting at 8:
250 ; R600-NEXT: VTX_READ_16 T0.X, T0.X, 48, #1
251 ; R600-NEXT: Fetch clause starting at 10:
252 ; R600-NEXT: VTX_READ_16 T1.X, T1.X, 32, #1
253 ; R600-NEXT: ALU clause starting at 12:
254 ; R600-NEXT: MOV * T0.X, KC0[2].Z,
255 ; R600-NEXT: ALU clause starting at 13:
256 ; R600-NEXT: MOV * T1.X, KC0[2].Y,
257 ; R600-NEXT: ALU clause starting at 14:
258 ; R600-NEXT: SUB_INT T0.W, 0.0, T0.X,
259 ; R600-NEXT: AND_INT * T1.W, T0.X, literal.x,
260 ; R600-NEXT: 15(2.101948e-44), 0(0.000000e+00)
261 ; R600-NEXT: AND_INT * T0.W, PV.W, literal.x,
262 ; R600-NEXT: 15(2.101948e-44), 0(0.000000e+00)
263 ; R600-NEXT: LSHL T0.Z, T1.X, PV.W,
264 ; R600-NEXT: LSHR T0.W, T1.X, T1.W,
265 ; R600-NEXT: ADD_INT * T1.W, KC0[2].W, literal.x,
266 ; R600-NEXT: 8(1.121039e-44), 0(0.000000e+00)
267 ; R600-NEXT: AND_INT T2.W, PS, literal.x,
268 ; R600-NEXT: OR_INT * T0.W, PV.W, PV.Z,
269 ; R600-NEXT: 3(4.203895e-45), 0(0.000000e+00)
270 ; R600-NEXT: AND_INT T0.W, PS, literal.x,
271 ; R600-NEXT: LSHL * T2.W, PV.W, literal.y,
272 ; R600-NEXT: 65535(9.183409e-41), 3(4.203895e-45)
273 ; R600-NEXT: LSHL T0.X, PV.W, PS,
274 ; R600-NEXT: LSHL * T0.W, literal.x, PS,
275 ; R600-NEXT: 65535(9.183409e-41), 0(0.000000e+00)
276 ; R600-NEXT: MOV T0.Y, 0.0,
277 ; R600-NEXT: MOV * T0.Z, 0.0,
278 ; R600-NEXT: LSHR * T1.X, T1.W, literal.x,
279 ; R600-NEXT: 2(2.802597e-45), 0(0.000000e+00)
281 ; SI-LABEL: test_rotr_i16:
282 ; SI: ; %bb.0: ; %entry
283 ; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
284 ; SI-NEXT: s_mov_b32 s6, 0
285 ; SI-NEXT: s_mov_b32 s7, 0xf000
286 ; SI-NEXT: s_mov_b32 s4, s6
287 ; SI-NEXT: s_mov_b32 s5, s6
288 ; SI-NEXT: buffer_load_ushort v2, v[2:3], s[4:7], 0 addr64 offset:48
289 ; SI-NEXT: buffer_load_ushort v0, v[0:1], s[4:7], 0 addr64 offset:32
290 ; SI-NEXT: s_waitcnt vmcnt(1)
291 ; SI-NEXT: v_and_b32_e32 v1, 15, v2
292 ; SI-NEXT: v_sub_i32_e32 v2, vcc, 0, v2
293 ; SI-NEXT: s_waitcnt vmcnt(0)
294 ; SI-NEXT: v_lshrrev_b32_e32 v1, v1, v0
295 ; SI-NEXT: v_and_b32_e32 v2, 15, v2
296 ; SI-NEXT: v_lshlrev_b32_e32 v0, v2, v0
297 ; SI-NEXT: v_or_b32_e32 v0, v1, v0
298 ; SI-NEXT: buffer_store_short v0, v[4:5], s[4:7], 0 addr64 offset:8
299 ; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
300 ; SI-NEXT: s_setpc_b64 s[30:31]
302 ; GFX8-LABEL: test_rotr_i16:
303 ; GFX8: ; %bb.0: ; %entry
304 ; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
305 ; GFX8-NEXT: v_add_u32_e32 v0, vcc, 32, v0
306 ; GFX8-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
307 ; GFX8-NEXT: v_add_u32_e32 v2, vcc, 48, v2
308 ; GFX8-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc
309 ; GFX8-NEXT: flat_load_ushort v2, v[2:3]
310 ; GFX8-NEXT: flat_load_ushort v0, v[0:1]
311 ; GFX8-NEXT: s_waitcnt vmcnt(0)
312 ; GFX8-NEXT: v_lshrrev_b16_e32 v1, v2, v0
313 ; GFX8-NEXT: v_sub_u16_e32 v2, 0, v2
314 ; GFX8-NEXT: v_lshlrev_b16_e32 v0, v2, v0
315 ; GFX8-NEXT: v_or_b32_e32 v2, v1, v0
316 ; GFX8-NEXT: v_add_u32_e32 v0, vcc, 8, v4
317 ; GFX8-NEXT: v_addc_u32_e32 v1, vcc, 0, v5, vcc
318 ; GFX8-NEXT: flat_store_short v[0:1], v2
319 ; GFX8-NEXT: s_waitcnt vmcnt(0)
320 ; GFX8-NEXT: s_setpc_b64 s[30:31]
322 ; GFX10-LABEL: test_rotr_i16:
323 ; GFX10: ; %bb.0: ; %entry
324 ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
325 ; GFX10-NEXT: global_load_ushort v6, v[2:3], off offset:48
326 ; GFX10-NEXT: global_load_ushort v7, v[0:1], off offset:32
327 ; GFX10-NEXT: s_waitcnt vmcnt(1)
328 ; GFX10-NEXT: v_sub_nc_u16 v0, 0, v6
329 ; GFX10-NEXT: s_waitcnt vmcnt(0)
330 ; GFX10-NEXT: v_lshrrev_b16 v1, v6, v7
331 ; GFX10-NEXT: v_lshlrev_b16 v0, v0, v7
332 ; GFX10-NEXT: v_or_b32_e32 v0, v1, v0
333 ; GFX10-NEXT: global_store_short v[4:5], v0, off offset:8
334 ; GFX10-NEXT: s_setpc_b64 s[30:31]
336 ; GFX11-LABEL: test_rotr_i16:
337 ; GFX11: ; %bb.0: ; %entry
338 ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
339 ; GFX11-NEXT: global_load_u16 v2, v[2:3], off offset:48
340 ; GFX11-NEXT: global_load_u16 v0, v[0:1], off offset:32
341 ; GFX11-NEXT: s_waitcnt vmcnt(1)
342 ; GFX11-NEXT: v_sub_nc_u16 v1, 0, v2
343 ; GFX11-NEXT: s_waitcnt vmcnt(0)
344 ; GFX11-NEXT: v_lshrrev_b16 v2, v2, v0
345 ; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
346 ; GFX11-NEXT: v_lshlrev_b16 v0, v1, v0
347 ; GFX11-NEXT: v_or_b32_e32 v0, v2, v0
348 ; GFX11-NEXT: global_store_b16 v[4:5], v0, off offset:8
349 ; GFX11-NEXT: s_setpc_b64 s[30:31]
351 %arrayidx = getelementptr inbounds i16, ptr addrspace(1) %sourceA, i64 16
352 %a = load i16, ptr addrspace(1) %arrayidx
353 %arrayidx2 = getelementptr inbounds i16, ptr addrspace(1) %sourceB, i64 24
354 %b = load i16, ptr addrspace(1) %arrayidx2
355 %c = tail call i16 @llvm.fshr.i16(i16 %a, i16 %a, i16 %b)
356 %arrayidx5 = getelementptr inbounds i16, ptr addrspace(1) %destValues, i64 4
357 store i16 %c, ptr addrspace(1) %arrayidx5