xref: /llvm-project/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-int-splat.ll (revision 319adf5de72a46345d0328122ebcd93c74d29592) 
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc -mtriple=riscv32 -mattr=+v -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=8 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,LMULMAX8,LMULMAX8-RV32
3; RUN: llc -mtriple=riscv32 -mattr=+v -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=2 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,LMULMAX2,LMULMAX2-RV32
4; RUN: llc -mtriple=riscv32 -mattr=+v -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=1 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,LMULMAX1,LMULMAX1-RV32
5; RUN: llc -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=8 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,LMULMAX8,LMULMAX8-RV64
6; RUN: llc -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=2 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,LMULMAX2,LMULMAX2-RV64
7; RUN: llc -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-min=128 -riscv-v-fixed-length-vector-lmul-max=1 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,LMULMAX1,LMULMAX1-RV64
8
9define void @splat_v16i8(ptr %x, i8 %y) {
10; CHECK-LABEL: splat_v16i8:
11; CHECK:       # %bb.0:
12; CHECK-NEXT:    vsetivli zero, 16, e8, m1, ta, ma
13; CHECK-NEXT:    vmv.v.x v8, a1
14; CHECK-NEXT:    vse8.v v8, (a0)
15; CHECK-NEXT:    ret
16  %a = insertelement <16 x i8> poison, i8 %y, i32 0
17  %b = shufflevector <16 x i8> %a, <16 x i8> poison, <16 x i32> zeroinitializer
18  store <16 x i8> %b, ptr %x
19  ret void
20}
21
22define void @splat_v8i16(ptr %x, i16 %y) {
23; CHECK-LABEL: splat_v8i16:
24; CHECK:       # %bb.0:
25; CHECK-NEXT:    vsetivli zero, 8, e16, m1, ta, ma
26; CHECK-NEXT:    vmv.v.x v8, a1
27; CHECK-NEXT:    vse16.v v8, (a0)
28; CHECK-NEXT:    ret
29  %a = insertelement <8 x i16> poison, i16 %y, i32 0
30  %b = shufflevector <8 x i16> %a, <8 x i16> poison, <8 x i32> zeroinitializer
31  store <8 x i16> %b, ptr %x
32  ret void
33}
34
35define void @splat_v4i32(ptr %x, i32 %y) {
36; CHECK-LABEL: splat_v4i32:
37; CHECK:       # %bb.0:
38; CHECK-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
39; CHECK-NEXT:    vmv.v.x v8, a1
40; CHECK-NEXT:    vse32.v v8, (a0)
41; CHECK-NEXT:    ret
42  %a = insertelement <4 x i32> poison, i32 %y, i32 0
43  %b = shufflevector <4 x i32> %a, <4 x i32> poison, <4 x i32> zeroinitializer
44  store <4 x i32> %b, ptr %x
45  ret void
46}
47
48define void @splat_v2i64(ptr %x, i64 %y) {
49; LMULMAX8-RV32-LABEL: splat_v2i64:
50; LMULMAX8-RV32:       # %bb.0:
51; LMULMAX8-RV32-NEXT:    addi sp, sp, -16
52; LMULMAX8-RV32-NEXT:    .cfi_def_cfa_offset 16
53; LMULMAX8-RV32-NEXT:    sw a2, 12(sp)
54; LMULMAX8-RV32-NEXT:    sw a1, 8(sp)
55; LMULMAX8-RV32-NEXT:    addi a1, sp, 8
56; LMULMAX8-RV32-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
57; LMULMAX8-RV32-NEXT:    vlse64.v v8, (a1), zero
58; LMULMAX8-RV32-NEXT:    vse64.v v8, (a0)
59; LMULMAX8-RV32-NEXT:    addi sp, sp, 16
60; LMULMAX8-RV32-NEXT:    ret
61;
62; LMULMAX2-RV32-LABEL: splat_v2i64:
63; LMULMAX2-RV32:       # %bb.0:
64; LMULMAX2-RV32-NEXT:    addi sp, sp, -16
65; LMULMAX2-RV32-NEXT:    .cfi_def_cfa_offset 16
66; LMULMAX2-RV32-NEXT:    sw a2, 12(sp)
67; LMULMAX2-RV32-NEXT:    sw a1, 8(sp)
68; LMULMAX2-RV32-NEXT:    addi a1, sp, 8
69; LMULMAX2-RV32-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
70; LMULMAX2-RV32-NEXT:    vlse64.v v8, (a1), zero
71; LMULMAX2-RV32-NEXT:    vse64.v v8, (a0)
72; LMULMAX2-RV32-NEXT:    addi sp, sp, 16
73; LMULMAX2-RV32-NEXT:    ret
74;
75; LMULMAX1-RV32-LABEL: splat_v2i64:
76; LMULMAX1-RV32:       # %bb.0:
77; LMULMAX1-RV32-NEXT:    addi sp, sp, -16
78; LMULMAX1-RV32-NEXT:    .cfi_def_cfa_offset 16
79; LMULMAX1-RV32-NEXT:    sw a2, 12(sp)
80; LMULMAX1-RV32-NEXT:    sw a1, 8(sp)
81; LMULMAX1-RV32-NEXT:    addi a1, sp, 8
82; LMULMAX1-RV32-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
83; LMULMAX1-RV32-NEXT:    vlse64.v v8, (a1), zero
84; LMULMAX1-RV32-NEXT:    vse64.v v8, (a0)
85; LMULMAX1-RV32-NEXT:    addi sp, sp, 16
86; LMULMAX1-RV32-NEXT:    ret
87;
88; LMULMAX8-RV64-LABEL: splat_v2i64:
89; LMULMAX8-RV64:       # %bb.0:
90; LMULMAX8-RV64-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
91; LMULMAX8-RV64-NEXT:    vmv.v.x v8, a1
92; LMULMAX8-RV64-NEXT:    vse64.v v8, (a0)
93; LMULMAX8-RV64-NEXT:    ret
94;
95; LMULMAX2-RV64-LABEL: splat_v2i64:
96; LMULMAX2-RV64:       # %bb.0:
97; LMULMAX2-RV64-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
98; LMULMAX2-RV64-NEXT:    vmv.v.x v8, a1
99; LMULMAX2-RV64-NEXT:    vse64.v v8, (a0)
100; LMULMAX2-RV64-NEXT:    ret
101;
102; LMULMAX1-RV64-LABEL: splat_v2i64:
103; LMULMAX1-RV64:       # %bb.0:
104; LMULMAX1-RV64-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
105; LMULMAX1-RV64-NEXT:    vmv.v.x v8, a1
106; LMULMAX1-RV64-NEXT:    vse64.v v8, (a0)
107; LMULMAX1-RV64-NEXT:    ret
108  %a = insertelement <2 x i64> poison, i64 %y, i32 0
109  %b = shufflevector <2 x i64> %a, <2 x i64> poison, <2 x i32> zeroinitializer
110  store <2 x i64> %b, ptr %x
111  ret void
112}
113
114define void @splat_v32i8(ptr %x, i8 %y) {
115; LMULMAX8-LABEL: splat_v32i8:
116; LMULMAX8:       # %bb.0:
117; LMULMAX8-NEXT:    li a2, 32
118; LMULMAX8-NEXT:    vsetvli zero, a2, e8, m2, ta, ma
119; LMULMAX8-NEXT:    vmv.v.x v8, a1
120; LMULMAX8-NEXT:    vse8.v v8, (a0)
121; LMULMAX8-NEXT:    ret
122;
123; LMULMAX2-LABEL: splat_v32i8:
124; LMULMAX2:       # %bb.0:
125; LMULMAX2-NEXT:    li a2, 32
126; LMULMAX2-NEXT:    vsetvli zero, a2, e8, m2, ta, ma
127; LMULMAX2-NEXT:    vmv.v.x v8, a1
128; LMULMAX2-NEXT:    vse8.v v8, (a0)
129; LMULMAX2-NEXT:    ret
130;
131; LMULMAX1-LABEL: splat_v32i8:
132; LMULMAX1:       # %bb.0:
133; LMULMAX1-NEXT:    vsetivli zero, 16, e8, m1, ta, ma
134; LMULMAX1-NEXT:    vmv.v.x v8, a1
135; LMULMAX1-NEXT:    addi a1, a0, 16
136; LMULMAX1-NEXT:    vse8.v v8, (a1)
137; LMULMAX1-NEXT:    vse8.v v8, (a0)
138; LMULMAX1-NEXT:    ret
139  %a = insertelement <32 x i8> poison, i8 %y, i32 0
140  %b = shufflevector <32 x i8> %a, <32 x i8> poison, <32 x i32> zeroinitializer
141  store <32 x i8> %b, ptr %x
142  ret void
143}
144
145define void @splat_v16i16(ptr %x, i16 %y) {
146; LMULMAX8-LABEL: splat_v16i16:
147; LMULMAX8:       # %bb.0:
148; LMULMAX8-NEXT:    vsetivli zero, 16, e16, m2, ta, ma
149; LMULMAX8-NEXT:    vmv.v.x v8, a1
150; LMULMAX8-NEXT:    vse16.v v8, (a0)
151; LMULMAX8-NEXT:    ret
152;
153; LMULMAX2-LABEL: splat_v16i16:
154; LMULMAX2:       # %bb.0:
155; LMULMAX2-NEXT:    vsetivli zero, 16, e16, m2, ta, ma
156; LMULMAX2-NEXT:    vmv.v.x v8, a1
157; LMULMAX2-NEXT:    vse16.v v8, (a0)
158; LMULMAX2-NEXT:    ret
159;
160; LMULMAX1-LABEL: splat_v16i16:
161; LMULMAX1:       # %bb.0:
162; LMULMAX1-NEXT:    vsetivli zero, 8, e16, m1, ta, ma
163; LMULMAX1-NEXT:    vmv.v.x v8, a1
164; LMULMAX1-NEXT:    addi a1, a0, 16
165; LMULMAX1-NEXT:    vse16.v v8, (a1)
166; LMULMAX1-NEXT:    vse16.v v8, (a0)
167; LMULMAX1-NEXT:    ret
168  %a = insertelement <16 x i16> poison, i16 %y, i32 0
169  %b = shufflevector <16 x i16> %a, <16 x i16> poison, <16 x i32> zeroinitializer
170  store <16 x i16> %b, ptr %x
171  ret void
172}
173
174define void @splat_v8i32(ptr %x, i32 %y) {
175; LMULMAX8-LABEL: splat_v8i32:
176; LMULMAX8:       # %bb.0:
177; LMULMAX8-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
178; LMULMAX8-NEXT:    vmv.v.x v8, a1
179; LMULMAX8-NEXT:    vse32.v v8, (a0)
180; LMULMAX8-NEXT:    ret
181;
182; LMULMAX2-LABEL: splat_v8i32:
183; LMULMAX2:       # %bb.0:
184; LMULMAX2-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
185; LMULMAX2-NEXT:    vmv.v.x v8, a1
186; LMULMAX2-NEXT:    vse32.v v8, (a0)
187; LMULMAX2-NEXT:    ret
188;
189; LMULMAX1-LABEL: splat_v8i32:
190; LMULMAX1:       # %bb.0:
191; LMULMAX1-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
192; LMULMAX1-NEXT:    vmv.v.x v8, a1
193; LMULMAX1-NEXT:    addi a1, a0, 16
194; LMULMAX1-NEXT:    vse32.v v8, (a1)
195; LMULMAX1-NEXT:    vse32.v v8, (a0)
196; LMULMAX1-NEXT:    ret
197  %a = insertelement <8 x i32> poison, i32 %y, i32 0
198  %b = shufflevector <8 x i32> %a, <8 x i32> poison, <8 x i32> zeroinitializer
199  store <8 x i32> %b, ptr %x
200  ret void
201}
202
203define void @splat_v4i64(ptr %x, i64 %y) {
204; LMULMAX8-RV32-LABEL: splat_v4i64:
205; LMULMAX8-RV32:       # %bb.0:
206; LMULMAX8-RV32-NEXT:    addi sp, sp, -16
207; LMULMAX8-RV32-NEXT:    .cfi_def_cfa_offset 16
208; LMULMAX8-RV32-NEXT:    sw a2, 12(sp)
209; LMULMAX8-RV32-NEXT:    sw a1, 8(sp)
210; LMULMAX8-RV32-NEXT:    addi a1, sp, 8
211; LMULMAX8-RV32-NEXT:    vsetivli zero, 4, e64, m2, ta, ma
212; LMULMAX8-RV32-NEXT:    vlse64.v v8, (a1), zero
213; LMULMAX8-RV32-NEXT:    vse64.v v8, (a0)
214; LMULMAX8-RV32-NEXT:    addi sp, sp, 16
215; LMULMAX8-RV32-NEXT:    ret
216;
217; LMULMAX2-RV32-LABEL: splat_v4i64:
218; LMULMAX2-RV32:       # %bb.0:
219; LMULMAX2-RV32-NEXT:    addi sp, sp, -16
220; LMULMAX2-RV32-NEXT:    .cfi_def_cfa_offset 16
221; LMULMAX2-RV32-NEXT:    sw a2, 12(sp)
222; LMULMAX2-RV32-NEXT:    sw a1, 8(sp)
223; LMULMAX2-RV32-NEXT:    addi a1, sp, 8
224; LMULMAX2-RV32-NEXT:    vsetivli zero, 4, e64, m2, ta, ma
225; LMULMAX2-RV32-NEXT:    vlse64.v v8, (a1), zero
226; LMULMAX2-RV32-NEXT:    vse64.v v8, (a0)
227; LMULMAX2-RV32-NEXT:    addi sp, sp, 16
228; LMULMAX2-RV32-NEXT:    ret
229;
230; LMULMAX1-RV32-LABEL: splat_v4i64:
231; LMULMAX1-RV32:       # %bb.0:
232; LMULMAX1-RV32-NEXT:    li a3, 5
233; LMULMAX1-RV32-NEXT:    vsetivli zero, 1, e8, mf8, ta, ma
234; LMULMAX1-RV32-NEXT:    vmv.s.x v0, a3
235; LMULMAX1-RV32-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
236; LMULMAX1-RV32-NEXT:    vmv.v.x v8, a2
237; LMULMAX1-RV32-NEXT:    vmerge.vxm v8, v8, a1, v0
238; LMULMAX1-RV32-NEXT:    addi a1, a0, 16
239; LMULMAX1-RV32-NEXT:    vse32.v v8, (a1)
240; LMULMAX1-RV32-NEXT:    vse32.v v8, (a0)
241; LMULMAX1-RV32-NEXT:    ret
242;
243; LMULMAX8-RV64-LABEL: splat_v4i64:
244; LMULMAX8-RV64:       # %bb.0:
245; LMULMAX8-RV64-NEXT:    vsetivli zero, 4, e64, m2, ta, ma
246; LMULMAX8-RV64-NEXT:    vmv.v.x v8, a1
247; LMULMAX8-RV64-NEXT:    vse64.v v8, (a0)
248; LMULMAX8-RV64-NEXT:    ret
249;
250; LMULMAX2-RV64-LABEL: splat_v4i64:
251; LMULMAX2-RV64:       # %bb.0:
252; LMULMAX2-RV64-NEXT:    vsetivli zero, 4, e64, m2, ta, ma
253; LMULMAX2-RV64-NEXT:    vmv.v.x v8, a1
254; LMULMAX2-RV64-NEXT:    vse64.v v8, (a0)
255; LMULMAX2-RV64-NEXT:    ret
256;
257; LMULMAX1-RV64-LABEL: splat_v4i64:
258; LMULMAX1-RV64:       # %bb.0:
259; LMULMAX1-RV64-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
260; LMULMAX1-RV64-NEXT:    vmv.v.x v8, a1
261; LMULMAX1-RV64-NEXT:    addi a1, a0, 16
262; LMULMAX1-RV64-NEXT:    vse64.v v8, (a1)
263; LMULMAX1-RV64-NEXT:    vse64.v v8, (a0)
264; LMULMAX1-RV64-NEXT:    ret
265  %a = insertelement <4 x i64> poison, i64 %y, i32 0
266  %b = shufflevector <4 x i64> %a, <4 x i64> poison, <4 x i32> zeroinitializer
267  store <4 x i64> %b, ptr %x
268  ret void
269}
270
271define void @splat_zero_v16i8(ptr %x) {
272; CHECK-LABEL: splat_zero_v16i8:
273; CHECK:       # %bb.0:
274; CHECK-NEXT:    vsetivli zero, 16, e8, m1, ta, ma
275; CHECK-NEXT:    vmv.v.i v8, 0
276; CHECK-NEXT:    vse8.v v8, (a0)
277; CHECK-NEXT:    ret
278  %a = insertelement <16 x i8> poison, i8 0, i32 0
279  %b = shufflevector <16 x i8> %a, <16 x i8> poison, <16 x i32> zeroinitializer
280  store <16 x i8> %b, ptr %x
281  ret void
282}
283
284define void @splat_zero_v8i16(ptr %x) {
285; CHECK-LABEL: splat_zero_v8i16:
286; CHECK:       # %bb.0:
287; CHECK-NEXT:    vsetivli zero, 8, e16, m1, ta, ma
288; CHECK-NEXT:    vmv.v.i v8, 0
289; CHECK-NEXT:    vse16.v v8, (a0)
290; CHECK-NEXT:    ret
291  %a = insertelement <8 x i16> poison, i16 0, i32 0
292  %b = shufflevector <8 x i16> %a, <8 x i16> poison, <8 x i32> zeroinitializer
293  store <8 x i16> %b, ptr %x
294  ret void
295}
296
297define void @splat_zero_v4i32(ptr %x) {
298; CHECK-LABEL: splat_zero_v4i32:
299; CHECK:       # %bb.0:
300; CHECK-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
301; CHECK-NEXT:    vmv.v.i v8, 0
302; CHECK-NEXT:    vse32.v v8, (a0)
303; CHECK-NEXT:    ret
304  %a = insertelement <4 x i32> poison, i32 0, i32 0
305  %b = shufflevector <4 x i32> %a, <4 x i32> poison, <4 x i32> zeroinitializer
306  store <4 x i32> %b, ptr %x
307  ret void
308}
309
310define void @splat_zero_v2i64(ptr %x) {
311; CHECK-LABEL: splat_zero_v2i64:
312; CHECK:       # %bb.0:
313; CHECK-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
314; CHECK-NEXT:    vmv.v.i v8, 0
315; CHECK-NEXT:    vse64.v v8, (a0)
316; CHECK-NEXT:    ret
317  %a = insertelement <2 x i64> poison, i64 0, i32 0
318  %b = shufflevector <2 x i64> %a, <2 x i64> poison, <2 x i32> zeroinitializer
319  store <2 x i64> %b, ptr %x
320  ret void
321}
322
323define void @splat_zero_v32i8(ptr %x) {
324; LMULMAX8-LABEL: splat_zero_v32i8:
325; LMULMAX8:       # %bb.0:
326; LMULMAX8-NEXT:    li a1, 32
327; LMULMAX8-NEXT:    vsetvli zero, a1, e8, m2, ta, ma
328; LMULMAX8-NEXT:    vmv.v.i v8, 0
329; LMULMAX8-NEXT:    vse8.v v8, (a0)
330; LMULMAX8-NEXT:    ret
331;
332; LMULMAX2-LABEL: splat_zero_v32i8:
333; LMULMAX2:       # %bb.0:
334; LMULMAX2-NEXT:    li a1, 32
335; LMULMAX2-NEXT:    vsetvli zero, a1, e8, m2, ta, ma
336; LMULMAX2-NEXT:    vmv.v.i v8, 0
337; LMULMAX2-NEXT:    vse8.v v8, (a0)
338; LMULMAX2-NEXT:    ret
339;
340; LMULMAX1-LABEL: splat_zero_v32i8:
341; LMULMAX1:       # %bb.0:
342; LMULMAX1-NEXT:    vsetivli zero, 16, e8, m1, ta, ma
343; LMULMAX1-NEXT:    vmv.v.i v8, 0
344; LMULMAX1-NEXT:    vse8.v v8, (a0)
345; LMULMAX1-NEXT:    addi a0, a0, 16
346; LMULMAX1-NEXT:    vse8.v v8, (a0)
347; LMULMAX1-NEXT:    ret
348  %a = insertelement <32 x i8> poison, i8 0, i32 0
349  %b = shufflevector <32 x i8> %a, <32 x i8> poison, <32 x i32> zeroinitializer
350  store <32 x i8> %b, ptr %x
351  ret void
352}
353
354define void @splat_zero_v16i16(ptr %x) {
355; LMULMAX8-LABEL: splat_zero_v16i16:
356; LMULMAX8:       # %bb.0:
357; LMULMAX8-NEXT:    vsetivli zero, 16, e16, m2, ta, ma
358; LMULMAX8-NEXT:    vmv.v.i v8, 0
359; LMULMAX8-NEXT:    vse16.v v8, (a0)
360; LMULMAX8-NEXT:    ret
361;
362; LMULMAX2-LABEL: splat_zero_v16i16:
363; LMULMAX2:       # %bb.0:
364; LMULMAX2-NEXT:    vsetivli zero, 16, e16, m2, ta, ma
365; LMULMAX2-NEXT:    vmv.v.i v8, 0
366; LMULMAX2-NEXT:    vse16.v v8, (a0)
367; LMULMAX2-NEXT:    ret
368;
369; LMULMAX1-LABEL: splat_zero_v16i16:
370; LMULMAX1:       # %bb.0:
371; LMULMAX1-NEXT:    vsetivli zero, 8, e16, m1, ta, ma
372; LMULMAX1-NEXT:    vmv.v.i v8, 0
373; LMULMAX1-NEXT:    vse16.v v8, (a0)
374; LMULMAX1-NEXT:    addi a0, a0, 16
375; LMULMAX1-NEXT:    vse16.v v8, (a0)
376; LMULMAX1-NEXT:    ret
377  %a = insertelement <16 x i16> poison, i16 0, i32 0
378  %b = shufflevector <16 x i16> %a, <16 x i16> poison, <16 x i32> zeroinitializer
379  store <16 x i16> %b, ptr %x
380  ret void
381}
382
383define void @splat_zero_v8i32(ptr %x) {
384; LMULMAX8-LABEL: splat_zero_v8i32:
385; LMULMAX8:       # %bb.0:
386; LMULMAX8-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
387; LMULMAX8-NEXT:    vmv.v.i v8, 0
388; LMULMAX8-NEXT:    vse32.v v8, (a0)
389; LMULMAX8-NEXT:    ret
390;
391; LMULMAX2-LABEL: splat_zero_v8i32:
392; LMULMAX2:       # %bb.0:
393; LMULMAX2-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
394; LMULMAX2-NEXT:    vmv.v.i v8, 0
395; LMULMAX2-NEXT:    vse32.v v8, (a0)
396; LMULMAX2-NEXT:    ret
397;
398; LMULMAX1-LABEL: splat_zero_v8i32:
399; LMULMAX1:       # %bb.0:
400; LMULMAX1-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
401; LMULMAX1-NEXT:    vmv.v.i v8, 0
402; LMULMAX1-NEXT:    vse32.v v8, (a0)
403; LMULMAX1-NEXT:    addi a0, a0, 16
404; LMULMAX1-NEXT:    vse32.v v8, (a0)
405; LMULMAX1-NEXT:    ret
406  %a = insertelement <8 x i32> poison, i32 0, i32 0
407  %b = shufflevector <8 x i32> %a, <8 x i32> poison, <8 x i32> zeroinitializer
408  store <8 x i32> %b, ptr %x
409  ret void
410}
411
412define void @splat_zero_v4i64(ptr %x) {
413; LMULMAX8-LABEL: splat_zero_v4i64:
414; LMULMAX8:       # %bb.0:
415; LMULMAX8-NEXT:    vsetivli zero, 4, e64, m2, ta, ma
416; LMULMAX8-NEXT:    vmv.v.i v8, 0
417; LMULMAX8-NEXT:    vse64.v v8, (a0)
418; LMULMAX8-NEXT:    ret
419;
420; LMULMAX2-LABEL: splat_zero_v4i64:
421; LMULMAX2:       # %bb.0:
422; LMULMAX2-NEXT:    vsetivli zero, 4, e64, m2, ta, ma
423; LMULMAX2-NEXT:    vmv.v.i v8, 0
424; LMULMAX2-NEXT:    vse64.v v8, (a0)
425; LMULMAX2-NEXT:    ret
426;
427; LMULMAX1-RV32-LABEL: splat_zero_v4i64:
428; LMULMAX1-RV32:       # %bb.0:
429; LMULMAX1-RV32-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
430; LMULMAX1-RV32-NEXT:    vmv.v.i v8, 0
431; LMULMAX1-RV32-NEXT:    vse32.v v8, (a0)
432; LMULMAX1-RV32-NEXT:    addi a0, a0, 16
433; LMULMAX1-RV32-NEXT:    vse32.v v8, (a0)
434; LMULMAX1-RV32-NEXT:    ret
435;
436; LMULMAX1-RV64-LABEL: splat_zero_v4i64:
437; LMULMAX1-RV64:       # %bb.0:
438; LMULMAX1-RV64-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
439; LMULMAX1-RV64-NEXT:    vmv.v.i v8, 0
440; LMULMAX1-RV64-NEXT:    vse64.v v8, (a0)
441; LMULMAX1-RV64-NEXT:    addi a0, a0, 16
442; LMULMAX1-RV64-NEXT:    vse64.v v8, (a0)
443; LMULMAX1-RV64-NEXT:    ret
444  %a = insertelement <4 x i64> poison, i64 0, i32 0
445  %b = shufflevector <4 x i64> %a, <4 x i64> poison, <4 x i32> zeroinitializer
446  store <4 x i64> %b, ptr %x
447  ret void
448}
449
450define void @splat_zero_v2i16(ptr %p) {
451; CHECK-LABEL: splat_zero_v2i16:
452; CHECK:       # %bb.0:
453; CHECK-NEXT:    sw zero, 0(a0)
454; CHECK-NEXT:    ret
455  store <2 x i16> zeroinitializer, ptr %p
456  ret void
457}
458
459define void @splat_zero_v2i16_unaligned(ptr %p) {
460; CHECK-LABEL: splat_zero_v2i16_unaligned:
461; CHECK:       # %bb.0:
462; CHECK-NEXT:    vsetivli zero, 2, e16, mf4, ta, ma
463; CHECK-NEXT:    vmv.v.i v8, 0
464; CHECK-NEXT:    vsetivli zero, 4, e8, mf4, ta, ma
465; CHECK-NEXT:    vse8.v v8, (a0)
466; CHECK-NEXT:    ret
467  store <2 x i16> zeroinitializer, ptr %p, align 1
468  ret void
469}
470
471define void @splat_zero_v4i16(ptr %p) {
472; LMULMAX8-RV32-LABEL: splat_zero_v4i16:
473; LMULMAX8-RV32:       # %bb.0:
474; LMULMAX8-RV32-NEXT:    vsetivli zero, 4, e16, mf2, ta, ma
475; LMULMAX8-RV32-NEXT:    vmv.v.i v8, 0
476; LMULMAX8-RV32-NEXT:    vse16.v v8, (a0)
477; LMULMAX8-RV32-NEXT:    ret
478;
479; LMULMAX2-RV32-LABEL: splat_zero_v4i16:
480; LMULMAX2-RV32:       # %bb.0:
481; LMULMAX2-RV32-NEXT:    vsetivli zero, 4, e16, mf2, ta, ma
482; LMULMAX2-RV32-NEXT:    vmv.v.i v8, 0
483; LMULMAX2-RV32-NEXT:    vse16.v v8, (a0)
484; LMULMAX2-RV32-NEXT:    ret
485;
486; LMULMAX1-RV32-LABEL: splat_zero_v4i16:
487; LMULMAX1-RV32:       # %bb.0:
488; LMULMAX1-RV32-NEXT:    vsetivli zero, 4, e16, mf2, ta, ma
489; LMULMAX1-RV32-NEXT:    vmv.v.i v8, 0
490; LMULMAX1-RV32-NEXT:    vse16.v v8, (a0)
491; LMULMAX1-RV32-NEXT:    ret
492;
493; LMULMAX8-RV64-LABEL: splat_zero_v4i16:
494; LMULMAX8-RV64:       # %bb.0:
495; LMULMAX8-RV64-NEXT:    sd zero, 0(a0)
496; LMULMAX8-RV64-NEXT:    ret
497;
498; LMULMAX2-RV64-LABEL: splat_zero_v4i16:
499; LMULMAX2-RV64:       # %bb.0:
500; LMULMAX2-RV64-NEXT:    sd zero, 0(a0)
501; LMULMAX2-RV64-NEXT:    ret
502;
503; LMULMAX1-RV64-LABEL: splat_zero_v4i16:
504; LMULMAX1-RV64:       # %bb.0:
505; LMULMAX1-RV64-NEXT:    sd zero, 0(a0)
506; LMULMAX1-RV64-NEXT:    ret
507  store <4 x i16> zeroinitializer, ptr %p
508  ret void
509}
510
511define void @splat_zero_v2i32(ptr %p) {
512; LMULMAX8-RV32-LABEL: splat_zero_v2i32:
513; LMULMAX8-RV32:       # %bb.0:
514; LMULMAX8-RV32-NEXT:    vsetivli zero, 2, e32, mf2, ta, ma
515; LMULMAX8-RV32-NEXT:    vmv.v.i v8, 0
516; LMULMAX8-RV32-NEXT:    vse32.v v8, (a0)
517; LMULMAX8-RV32-NEXT:    ret
518;
519; LMULMAX2-RV32-LABEL: splat_zero_v2i32:
520; LMULMAX2-RV32:       # %bb.0:
521; LMULMAX2-RV32-NEXT:    vsetivli zero, 2, e32, mf2, ta, ma
522; LMULMAX2-RV32-NEXT:    vmv.v.i v8, 0
523; LMULMAX2-RV32-NEXT:    vse32.v v8, (a0)
524; LMULMAX2-RV32-NEXT:    ret
525;
526; LMULMAX1-RV32-LABEL: splat_zero_v2i32:
527; LMULMAX1-RV32:       # %bb.0:
528; LMULMAX1-RV32-NEXT:    vsetivli zero, 2, e32, mf2, ta, ma
529; LMULMAX1-RV32-NEXT:    vmv.v.i v8, 0
530; LMULMAX1-RV32-NEXT:    vse32.v v8, (a0)
531; LMULMAX1-RV32-NEXT:    ret
532;
533; LMULMAX8-RV64-LABEL: splat_zero_v2i32:
534; LMULMAX8-RV64:       # %bb.0:
535; LMULMAX8-RV64-NEXT:    sd zero, 0(a0)
536; LMULMAX8-RV64-NEXT:    ret
537;
538; LMULMAX2-RV64-LABEL: splat_zero_v2i32:
539; LMULMAX2-RV64:       # %bb.0:
540; LMULMAX2-RV64-NEXT:    sd zero, 0(a0)
541; LMULMAX2-RV64-NEXT:    ret
542;
543; LMULMAX1-RV64-LABEL: splat_zero_v2i32:
544; LMULMAX1-RV64:       # %bb.0:
545; LMULMAX1-RV64-NEXT:    sd zero, 0(a0)
546; LMULMAX1-RV64-NEXT:    ret
547  store <2 x i32> zeroinitializer, ptr %p
548  ret void
549}
550
551; Not a power of two and requires more than two scalar stores.
552define void @splat_zero_v7i16(ptr %p) {
553; LMULMAX8-RV32-LABEL: splat_zero_v7i16:
554; LMULMAX8-RV32:       # %bb.0:
555; LMULMAX8-RV32-NEXT:    sh zero, 12(a0)
556; LMULMAX8-RV32-NEXT:    vsetivli zero, 4, e16, mf2, ta, ma
557; LMULMAX8-RV32-NEXT:    vmv.v.i v8, 0
558; LMULMAX8-RV32-NEXT:    vse16.v v8, (a0)
559; LMULMAX8-RV32-NEXT:    vsetivli zero, 8, e16, m1, ta, ma
560; LMULMAX8-RV32-NEXT:    vmv.v.i v8, 0
561; LMULMAX8-RV32-NEXT:    vsetivli zero, 1, e32, m1, ta, ma
562; LMULMAX8-RV32-NEXT:    vslidedown.vi v8, v8, 2
563; LMULMAX8-RV32-NEXT:    addi a0, a0, 8
564; LMULMAX8-RV32-NEXT:    vse32.v v8, (a0)
565; LMULMAX8-RV32-NEXT:    ret
566;
567; LMULMAX2-RV32-LABEL: splat_zero_v7i16:
568; LMULMAX2-RV32:       # %bb.0:
569; LMULMAX2-RV32-NEXT:    sh zero, 12(a0)
570; LMULMAX2-RV32-NEXT:    vsetivli zero, 4, e16, mf2, ta, ma
571; LMULMAX2-RV32-NEXT:    vmv.v.i v8, 0
572; LMULMAX2-RV32-NEXT:    vse16.v v8, (a0)
573; LMULMAX2-RV32-NEXT:    vsetivli zero, 8, e16, m1, ta, ma
574; LMULMAX2-RV32-NEXT:    vmv.v.i v8, 0
575; LMULMAX2-RV32-NEXT:    vsetivli zero, 1, e32, m1, ta, ma
576; LMULMAX2-RV32-NEXT:    vslidedown.vi v8, v8, 2
577; LMULMAX2-RV32-NEXT:    addi a0, a0, 8
578; LMULMAX2-RV32-NEXT:    vse32.v v8, (a0)
579; LMULMAX2-RV32-NEXT:    ret
580;
581; LMULMAX1-RV32-LABEL: splat_zero_v7i16:
582; LMULMAX1-RV32:       # %bb.0:
583; LMULMAX1-RV32-NEXT:    sh zero, 12(a0)
584; LMULMAX1-RV32-NEXT:    vsetivli zero, 4, e16, mf2, ta, ma
585; LMULMAX1-RV32-NEXT:    vmv.v.i v8, 0
586; LMULMAX1-RV32-NEXT:    vse16.v v8, (a0)
587; LMULMAX1-RV32-NEXT:    vsetivli zero, 8, e16, m1, ta, ma
588; LMULMAX1-RV32-NEXT:    vmv.v.i v8, 0
589; LMULMAX1-RV32-NEXT:    vsetivli zero, 1, e32, m1, ta, ma
590; LMULMAX1-RV32-NEXT:    vslidedown.vi v8, v8, 2
591; LMULMAX1-RV32-NEXT:    addi a0, a0, 8
592; LMULMAX1-RV32-NEXT:    vse32.v v8, (a0)
593; LMULMAX1-RV32-NEXT:    ret
594;
595; LMULMAX8-RV64-LABEL: splat_zero_v7i16:
596; LMULMAX8-RV64:       # %bb.0:
597; LMULMAX8-RV64-NEXT:    sh zero, 12(a0)
598; LMULMAX8-RV64-NEXT:    sw zero, 8(a0)
599; LMULMAX8-RV64-NEXT:    sd zero, 0(a0)
600; LMULMAX8-RV64-NEXT:    ret
601;
602; LMULMAX2-RV64-LABEL: splat_zero_v7i16:
603; LMULMAX2-RV64:       # %bb.0:
604; LMULMAX2-RV64-NEXT:    sh zero, 12(a0)
605; LMULMAX2-RV64-NEXT:    sw zero, 8(a0)
606; LMULMAX2-RV64-NEXT:    sd zero, 0(a0)
607; LMULMAX2-RV64-NEXT:    ret
608;
609; LMULMAX1-RV64-LABEL: splat_zero_v7i16:
610; LMULMAX1-RV64:       # %bb.0:
611; LMULMAX1-RV64-NEXT:    sh zero, 12(a0)
612; LMULMAX1-RV64-NEXT:    sw zero, 8(a0)
613; LMULMAX1-RV64-NEXT:    sd zero, 0(a0)
614; LMULMAX1-RV64-NEXT:    ret
615  store <7 x i16> zeroinitializer, ptr %p
616  ret void
617}
618
619define void @splat_allones_v16i8(ptr %x) {
620; CHECK-LABEL: splat_allones_v16i8:
621; CHECK:       # %bb.0:
622; CHECK-NEXT:    vsetivli zero, 16, e8, m1, ta, ma
623; CHECK-NEXT:    vmv.v.i v8, -1
624; CHECK-NEXT:    vse8.v v8, (a0)
625; CHECK-NEXT:    ret
626  %a = insertelement <16 x i8> poison, i8 -1, i32 0
627  %b = shufflevector <16 x i8> %a, <16 x i8> poison, <16 x i32> zeroinitializer
628  store <16 x i8> %b, ptr %x
629  ret void
630}
631
632define void @splat_allones_v8i16(ptr %x) {
633; CHECK-LABEL: splat_allones_v8i16:
634; CHECK:       # %bb.0:
635; CHECK-NEXT:    vsetivli zero, 8, e16, m1, ta, ma
636; CHECK-NEXT:    vmv.v.i v8, -1
637; CHECK-NEXT:    vse16.v v8, (a0)
638; CHECK-NEXT:    ret
639  %a = insertelement <8 x i16> poison, i16 -1, i32 0
640  %b = shufflevector <8 x i16> %a, <8 x i16> poison, <8 x i32> zeroinitializer
641  store <8 x i16> %b, ptr %x
642  ret void
643}
644
645define void @splat_allones_v4i32(ptr %x) {
646; CHECK-LABEL: splat_allones_v4i32:
647; CHECK:       # %bb.0:
648; CHECK-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
649; CHECK-NEXT:    vmv.v.i v8, -1
650; CHECK-NEXT:    vse32.v v8, (a0)
651; CHECK-NEXT:    ret
652  %a = insertelement <4 x i32> poison, i32 -1, i32 0
653  %b = shufflevector <4 x i32> %a, <4 x i32> poison, <4 x i32> zeroinitializer
654  store <4 x i32> %b, ptr %x
655  ret void
656}
657
658define void @splat_allones_v2i64(ptr %x) {
659; CHECK-LABEL: splat_allones_v2i64:
660; CHECK:       # %bb.0:
661; CHECK-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
662; CHECK-NEXT:    vmv.v.i v8, -1
663; CHECK-NEXT:    vse64.v v8, (a0)
664; CHECK-NEXT:    ret
665  %a = insertelement <2 x i64> poison, i64 -1, i32 0
666  %b = shufflevector <2 x i64> %a, <2 x i64> poison, <2 x i32> zeroinitializer
667  store <2 x i64> %b, ptr %x
668  ret void
669}
670
671define void @splat_allones_v32i8(ptr %x) {
672; LMULMAX8-LABEL: splat_allones_v32i8:
673; LMULMAX8:       # %bb.0:
674; LMULMAX8-NEXT:    li a1, 32
675; LMULMAX8-NEXT:    vsetvli zero, a1, e8, m2, ta, ma
676; LMULMAX8-NEXT:    vmv.v.i v8, -1
677; LMULMAX8-NEXT:    vse8.v v8, (a0)
678; LMULMAX8-NEXT:    ret
679;
680; LMULMAX2-LABEL: splat_allones_v32i8:
681; LMULMAX2:       # %bb.0:
682; LMULMAX2-NEXT:    li a1, 32
683; LMULMAX2-NEXT:    vsetvli zero, a1, e8, m2, ta, ma
684; LMULMAX2-NEXT:    vmv.v.i v8, -1
685; LMULMAX2-NEXT:    vse8.v v8, (a0)
686; LMULMAX2-NEXT:    ret
687;
688; LMULMAX1-LABEL: splat_allones_v32i8:
689; LMULMAX1:       # %bb.0:
690; LMULMAX1-NEXT:    vsetivli zero, 16, e8, m1, ta, ma
691; LMULMAX1-NEXT:    vmv.v.i v8, -1
692; LMULMAX1-NEXT:    vse8.v v8, (a0)
693; LMULMAX1-NEXT:    addi a0, a0, 16
694; LMULMAX1-NEXT:    vse8.v v8, (a0)
695; LMULMAX1-NEXT:    ret
696  %a = insertelement <32 x i8> poison, i8 -1, i32 0
697  %b = shufflevector <32 x i8> %a, <32 x i8> poison, <32 x i32> zeroinitializer
698  store <32 x i8> %b, ptr %x
699  ret void
700}
701
702define void @splat_allones_v16i16(ptr %x) {
703; LMULMAX8-LABEL: splat_allones_v16i16:
704; LMULMAX8:       # %bb.0:
705; LMULMAX8-NEXT:    vsetivli zero, 16, e16, m2, ta, ma
706; LMULMAX8-NEXT:    vmv.v.i v8, -1
707; LMULMAX8-NEXT:    vse16.v v8, (a0)
708; LMULMAX8-NEXT:    ret
709;
710; LMULMAX2-LABEL: splat_allones_v16i16:
711; LMULMAX2:       # %bb.0:
712; LMULMAX2-NEXT:    vsetivli zero, 16, e16, m2, ta, ma
713; LMULMAX2-NEXT:    vmv.v.i v8, -1
714; LMULMAX2-NEXT:    vse16.v v8, (a0)
715; LMULMAX2-NEXT:    ret
716;
717; LMULMAX1-LABEL: splat_allones_v16i16:
718; LMULMAX1:       # %bb.0:
719; LMULMAX1-NEXT:    vsetivli zero, 8, e16, m1, ta, ma
720; LMULMAX1-NEXT:    vmv.v.i v8, -1
721; LMULMAX1-NEXT:    vse16.v v8, (a0)
722; LMULMAX1-NEXT:    addi a0, a0, 16
723; LMULMAX1-NEXT:    vse16.v v8, (a0)
724; LMULMAX1-NEXT:    ret
725  %a = insertelement <16 x i16> poison, i16 -1, i32 0
726  %b = shufflevector <16 x i16> %a, <16 x i16> poison, <16 x i32> zeroinitializer
727  store <16 x i16> %b, ptr %x
728  ret void
729}
730
731define void @splat_allones_v8i32(ptr %x) {
732; LMULMAX8-LABEL: splat_allones_v8i32:
733; LMULMAX8:       # %bb.0:
734; LMULMAX8-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
735; LMULMAX8-NEXT:    vmv.v.i v8, -1
736; LMULMAX8-NEXT:    vse32.v v8, (a0)
737; LMULMAX8-NEXT:    ret
738;
739; LMULMAX2-LABEL: splat_allones_v8i32:
740; LMULMAX2:       # %bb.0:
741; LMULMAX2-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
742; LMULMAX2-NEXT:    vmv.v.i v8, -1
743; LMULMAX2-NEXT:    vse32.v v8, (a0)
744; LMULMAX2-NEXT:    ret
745;
746; LMULMAX1-LABEL: splat_allones_v8i32:
747; LMULMAX1:       # %bb.0:
748; LMULMAX1-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
749; LMULMAX1-NEXT:    vmv.v.i v8, -1
750; LMULMAX1-NEXT:    vse32.v v8, (a0)
751; LMULMAX1-NEXT:    addi a0, a0, 16
752; LMULMAX1-NEXT:    vse32.v v8, (a0)
753; LMULMAX1-NEXT:    ret
754  %a = insertelement <8 x i32> poison, i32 -1, i32 0
755  %b = shufflevector <8 x i32> %a, <8 x i32> poison, <8 x i32> zeroinitializer
756  store <8 x i32> %b, ptr %x
757  ret void
758}
759
760define void @splat_allones_v4i64(ptr %x) {
761; LMULMAX8-LABEL: splat_allones_v4i64:
762; LMULMAX8:       # %bb.0:
763; LMULMAX8-NEXT:    vsetivli zero, 4, e64, m2, ta, ma
764; LMULMAX8-NEXT:    vmv.v.i v8, -1
765; LMULMAX8-NEXT:    vse64.v v8, (a0)
766; LMULMAX8-NEXT:    ret
767;
768; LMULMAX2-LABEL: splat_allones_v4i64:
769; LMULMAX2:       # %bb.0:
770; LMULMAX2-NEXT:    vsetivli zero, 4, e64, m2, ta, ma
771; LMULMAX2-NEXT:    vmv.v.i v8, -1
772; LMULMAX2-NEXT:    vse64.v v8, (a0)
773; LMULMAX2-NEXT:    ret
774;
775; LMULMAX1-RV32-LABEL: splat_allones_v4i64:
776; LMULMAX1-RV32:       # %bb.0:
777; LMULMAX1-RV32-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
778; LMULMAX1-RV32-NEXT:    vmv.v.i v8, -1
779; LMULMAX1-RV32-NEXT:    vse32.v v8, (a0)
780; LMULMAX1-RV32-NEXT:    addi a0, a0, 16
781; LMULMAX1-RV32-NEXT:    vse32.v v8, (a0)
782; LMULMAX1-RV32-NEXT:    ret
783;
784; LMULMAX1-RV64-LABEL: splat_allones_v4i64:
785; LMULMAX1-RV64:       # %bb.0:
786; LMULMAX1-RV64-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
787; LMULMAX1-RV64-NEXT:    vmv.v.i v8, -1
788; LMULMAX1-RV64-NEXT:    vse64.v v8, (a0)
789; LMULMAX1-RV64-NEXT:    addi a0, a0, 16
790; LMULMAX1-RV64-NEXT:    vse64.v v8, (a0)
791; LMULMAX1-RV64-NEXT:    ret
792  %a = insertelement <4 x i64> poison, i64 -1, i32 0
793  %b = shufflevector <4 x i64> %a, <4 x i64> poison, <4 x i32> zeroinitializer
794  store <4 x i64> %b, ptr %x
795  ret void
796}
797
798; This requires a bitcast on RV32 due to type legalization rewriting the
799; build_vector to v8i32.
800; FIXME: We should prevent this and use the implicit sign extension of vmv.v.x
801; with SEW=64 on RV32.
802define void @splat_allones_with_use_v4i64(ptr %x) {
803; LMULMAX8-LABEL: splat_allones_with_use_v4i64:
804; LMULMAX8:       # %bb.0:
805; LMULMAX8-NEXT:    vsetivli zero, 4, e64, m2, ta, ma
806; LMULMAX8-NEXT:    vle64.v v8, (a0)
807; LMULMAX8-NEXT:    vadd.vi v8, v8, -1
808; LMULMAX8-NEXT:    vse64.v v8, (a0)
809; LMULMAX8-NEXT:    ret
810;
811; LMULMAX2-LABEL: splat_allones_with_use_v4i64:
812; LMULMAX2:       # %bb.0:
813; LMULMAX2-NEXT:    vsetivli zero, 4, e64, m2, ta, ma
814; LMULMAX2-NEXT:    vle64.v v8, (a0)
815; LMULMAX2-NEXT:    vadd.vi v8, v8, -1
816; LMULMAX2-NEXT:    vse64.v v8, (a0)
817; LMULMAX2-NEXT:    ret
818;
819; LMULMAX1-RV32-LABEL: splat_allones_with_use_v4i64:
820; LMULMAX1-RV32:       # %bb.0:
821; LMULMAX1-RV32-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
822; LMULMAX1-RV32-NEXT:    vle64.v v8, (a0)
823; LMULMAX1-RV32-NEXT:    addi a1, a0, 16
824; LMULMAX1-RV32-NEXT:    vle64.v v9, (a1)
825; LMULMAX1-RV32-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
826; LMULMAX1-RV32-NEXT:    vmv.v.i v10, -1
827; LMULMAX1-RV32-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
828; LMULMAX1-RV32-NEXT:    vadd.vv v9, v9, v10
829; LMULMAX1-RV32-NEXT:    vadd.vv v8, v8, v10
830; LMULMAX1-RV32-NEXT:    vse64.v v8, (a0)
831; LMULMAX1-RV32-NEXT:    vse64.v v9, (a1)
832; LMULMAX1-RV32-NEXT:    ret
833;
834; LMULMAX1-RV64-LABEL: splat_allones_with_use_v4i64:
835; LMULMAX1-RV64:       # %bb.0:
836; LMULMAX1-RV64-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
837; LMULMAX1-RV64-NEXT:    addi a1, a0, 16
838; LMULMAX1-RV64-NEXT:    vle64.v v8, (a1)
839; LMULMAX1-RV64-NEXT:    vle64.v v9, (a0)
840; LMULMAX1-RV64-NEXT:    vadd.vi v8, v8, -1
841; LMULMAX1-RV64-NEXT:    vadd.vi v9, v9, -1
842; LMULMAX1-RV64-NEXT:    vse64.v v9, (a0)
843; LMULMAX1-RV64-NEXT:    vse64.v v8, (a1)
844; LMULMAX1-RV64-NEXT:    ret
845  %a = load <4 x i64>, ptr %x
846  %b = add <4 x i64> %a, <i64 -1, i64 -1, i64 -1, i64 -1>
847  store <4 x i64> %b, ptr %x
848  ret void
849}
850
851; This test used to crash at LMUL=8 when inserting a v16i64 subvector into
852; nxv8i64 at index 0: the v16i64 type was used to get the LMUL, the size of
853; which exceeded maximum-expected size of 512. The scalable container type of
854; nxv8i64 should have been used instead.
855define void @vadd_vx_v16i64(ptr %a, i64 %b, ptr %c) {
856; LMULMAX8-RV32-LABEL: vadd_vx_v16i64:
857; LMULMAX8-RV32:       # %bb.0:
858; LMULMAX8-RV32-NEXT:    addi sp, sp, -16
859; LMULMAX8-RV32-NEXT:    .cfi_def_cfa_offset 16
860; LMULMAX8-RV32-NEXT:    vsetivli zero, 16, e64, m8, ta, ma
861; LMULMAX8-RV32-NEXT:    vle64.v v8, (a0)
862; LMULMAX8-RV32-NEXT:    sw a2, 12(sp)
863; LMULMAX8-RV32-NEXT:    sw a1, 8(sp)
864; LMULMAX8-RV32-NEXT:    addi a0, sp, 8
865; LMULMAX8-RV32-NEXT:    vlse64.v v16, (a0), zero
866; LMULMAX8-RV32-NEXT:    vadd.vv v8, v8, v16
867; LMULMAX8-RV32-NEXT:    vse64.v v8, (a3)
868; LMULMAX8-RV32-NEXT:    addi sp, sp, 16
869; LMULMAX8-RV32-NEXT:    ret
870;
871; LMULMAX2-RV32-LABEL: vadd_vx_v16i64:
872; LMULMAX2-RV32:       # %bb.0:
873; LMULMAX2-RV32-NEXT:    addi a4, a0, 64
874; LMULMAX2-RV32-NEXT:    vsetivli zero, 4, e64, m2, ta, ma
875; LMULMAX2-RV32-NEXT:    vle64.v v8, (a4)
876; LMULMAX2-RV32-NEXT:    addi a4, a0, 96
877; LMULMAX2-RV32-NEXT:    vle64.v v10, (a4)
878; LMULMAX2-RV32-NEXT:    vle64.v v12, (a0)
879; LMULMAX2-RV32-NEXT:    addi a0, a0, 32
880; LMULMAX2-RV32-NEXT:    vle64.v v14, (a0)
881; LMULMAX2-RV32-NEXT:    li a0, 85
882; LMULMAX2-RV32-NEXT:    vmv.s.x v0, a0
883; LMULMAX2-RV32-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
884; LMULMAX2-RV32-NEXT:    vmv.v.x v16, a2
885; LMULMAX2-RV32-NEXT:    vmerge.vxm v16, v16, a1, v0
886; LMULMAX2-RV32-NEXT:    vsetivli zero, 4, e64, m2, ta, ma
887; LMULMAX2-RV32-NEXT:    vadd.vv v14, v14, v16
888; LMULMAX2-RV32-NEXT:    vadd.vv v12, v12, v16
889; LMULMAX2-RV32-NEXT:    vadd.vv v10, v10, v16
890; LMULMAX2-RV32-NEXT:    vadd.vv v8, v8, v16
891; LMULMAX2-RV32-NEXT:    addi a0, a3, 64
892; LMULMAX2-RV32-NEXT:    vse64.v v8, (a0)
893; LMULMAX2-RV32-NEXT:    addi a0, a3, 96
894; LMULMAX2-RV32-NEXT:    vse64.v v10, (a0)
895; LMULMAX2-RV32-NEXT:    vse64.v v12, (a3)
896; LMULMAX2-RV32-NEXT:    addi a0, a3, 32
897; LMULMAX2-RV32-NEXT:    vse64.v v14, (a0)
898; LMULMAX2-RV32-NEXT:    ret
899;
900; LMULMAX1-RV32-LABEL: vadd_vx_v16i64:
901; LMULMAX1-RV32:       # %bb.0:
902; LMULMAX1-RV32-NEXT:    addi a4, a0, 96
903; LMULMAX1-RV32-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
904; LMULMAX1-RV32-NEXT:    vle64.v v8, (a4)
905; LMULMAX1-RV32-NEXT:    addi a4, a0, 112
906; LMULMAX1-RV32-NEXT:    vle64.v v9, (a4)
907; LMULMAX1-RV32-NEXT:    addi a4, a0, 64
908; LMULMAX1-RV32-NEXT:    vle64.v v10, (a4)
909; LMULMAX1-RV32-NEXT:    addi a4, a0, 80
910; LMULMAX1-RV32-NEXT:    vle64.v v11, (a4)
911; LMULMAX1-RV32-NEXT:    addi a4, a0, 32
912; LMULMAX1-RV32-NEXT:    vle64.v v12, (a4)
913; LMULMAX1-RV32-NEXT:    addi a4, a0, 48
914; LMULMAX1-RV32-NEXT:    vle64.v v13, (a4)
915; LMULMAX1-RV32-NEXT:    vle64.v v14, (a0)
916; LMULMAX1-RV32-NEXT:    addi a0, a0, 16
917; LMULMAX1-RV32-NEXT:    vle64.v v15, (a0)
918; LMULMAX1-RV32-NEXT:    li a0, 5
919; LMULMAX1-RV32-NEXT:    vmv.s.x v0, a0
920; LMULMAX1-RV32-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
921; LMULMAX1-RV32-NEXT:    vmv.v.x v16, a2
922; LMULMAX1-RV32-NEXT:    vmerge.vxm v16, v16, a1, v0
923; LMULMAX1-RV32-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
924; LMULMAX1-RV32-NEXT:    vadd.vv v15, v15, v16
925; LMULMAX1-RV32-NEXT:    vadd.vv v14, v14, v16
926; LMULMAX1-RV32-NEXT:    vadd.vv v13, v13, v16
927; LMULMAX1-RV32-NEXT:    vadd.vv v12, v12, v16
928; LMULMAX1-RV32-NEXT:    vadd.vv v11, v11, v16
929; LMULMAX1-RV32-NEXT:    vadd.vv v10, v10, v16
930; LMULMAX1-RV32-NEXT:    vadd.vv v9, v9, v16
931; LMULMAX1-RV32-NEXT:    vadd.vv v8, v8, v16
932; LMULMAX1-RV32-NEXT:    addi a0, a3, 96
933; LMULMAX1-RV32-NEXT:    vse64.v v8, (a0)
934; LMULMAX1-RV32-NEXT:    addi a0, a3, 112
935; LMULMAX1-RV32-NEXT:    vse64.v v9, (a0)
936; LMULMAX1-RV32-NEXT:    addi a0, a3, 64
937; LMULMAX1-RV32-NEXT:    vse64.v v10, (a0)
938; LMULMAX1-RV32-NEXT:    addi a0, a3, 80
939; LMULMAX1-RV32-NEXT:    vse64.v v11, (a0)
940; LMULMAX1-RV32-NEXT:    addi a0, a3, 32
941; LMULMAX1-RV32-NEXT:    vse64.v v12, (a0)
942; LMULMAX1-RV32-NEXT:    addi a0, a3, 48
943; LMULMAX1-RV32-NEXT:    vse64.v v13, (a0)
944; LMULMAX1-RV32-NEXT:    vse64.v v14, (a3)
945; LMULMAX1-RV32-NEXT:    addi a3, a3, 16
946; LMULMAX1-RV32-NEXT:    vse64.v v15, (a3)
947; LMULMAX1-RV32-NEXT:    ret
948;
949; LMULMAX8-RV64-LABEL: vadd_vx_v16i64:
950; LMULMAX8-RV64:       # %bb.0:
951; LMULMAX8-RV64-NEXT:    vsetivli zero, 16, e64, m8, ta, ma
952; LMULMAX8-RV64-NEXT:    vle64.v v8, (a0)
953; LMULMAX8-RV64-NEXT:    vadd.vx v8, v8, a1
954; LMULMAX8-RV64-NEXT:    vse64.v v8, (a2)
955; LMULMAX8-RV64-NEXT:    ret
956;
957; LMULMAX2-RV64-LABEL: vadd_vx_v16i64:
958; LMULMAX2-RV64:       # %bb.0:
959; LMULMAX2-RV64-NEXT:    vsetivli zero, 4, e64, m2, ta, ma
960; LMULMAX2-RV64-NEXT:    addi a3, a0, 96
961; LMULMAX2-RV64-NEXT:    vle64.v v8, (a3)
962; LMULMAX2-RV64-NEXT:    addi a3, a0, 32
963; LMULMAX2-RV64-NEXT:    vle64.v v10, (a3)
964; LMULMAX2-RV64-NEXT:    addi a3, a0, 64
965; LMULMAX2-RV64-NEXT:    vle64.v v12, (a3)
966; LMULMAX2-RV64-NEXT:    vle64.v v14, (a0)
967; LMULMAX2-RV64-NEXT:    vadd.vx v10, v10, a1
968; LMULMAX2-RV64-NEXT:    vadd.vx v8, v8, a1
969; LMULMAX2-RV64-NEXT:    vadd.vx v12, v12, a1
970; LMULMAX2-RV64-NEXT:    vadd.vx v14, v14, a1
971; LMULMAX2-RV64-NEXT:    vse64.v v14, (a2)
972; LMULMAX2-RV64-NEXT:    addi a0, a2, 64
973; LMULMAX2-RV64-NEXT:    vse64.v v12, (a0)
974; LMULMAX2-RV64-NEXT:    addi a0, a2, 96
975; LMULMAX2-RV64-NEXT:    vse64.v v8, (a0)
976; LMULMAX2-RV64-NEXT:    addi a0, a2, 32
977; LMULMAX2-RV64-NEXT:    vse64.v v10, (a0)
978; LMULMAX2-RV64-NEXT:    ret
979;
980; LMULMAX1-RV64-LABEL: vadd_vx_v16i64:
981; LMULMAX1-RV64:       # %bb.0:
982; LMULMAX1-RV64-NEXT:    vsetivli zero, 2, e64, m1, ta, ma
983; LMULMAX1-RV64-NEXT:    vle64.v v8, (a0)
984; LMULMAX1-RV64-NEXT:    addi a3, a0, 96
985; LMULMAX1-RV64-NEXT:    vle64.v v9, (a3)
986; LMULMAX1-RV64-NEXT:    addi a3, a0, 112
987; LMULMAX1-RV64-NEXT:    vle64.v v10, (a3)
988; LMULMAX1-RV64-NEXT:    addi a3, a0, 64
989; LMULMAX1-RV64-NEXT:    vle64.v v11, (a3)
990; LMULMAX1-RV64-NEXT:    addi a3, a0, 48
991; LMULMAX1-RV64-NEXT:    vle64.v v12, (a3)
992; LMULMAX1-RV64-NEXT:    addi a3, a0, 16
993; LMULMAX1-RV64-NEXT:    vle64.v v13, (a3)
994; LMULMAX1-RV64-NEXT:    addi a3, a0, 80
995; LMULMAX1-RV64-NEXT:    addi a0, a0, 32
996; LMULMAX1-RV64-NEXT:    vle64.v v14, (a0)
997; LMULMAX1-RV64-NEXT:    vle64.v v15, (a3)
998; LMULMAX1-RV64-NEXT:    vadd.vx v13, v13, a1
999; LMULMAX1-RV64-NEXT:    vadd.vx v12, v12, a1
1000; LMULMAX1-RV64-NEXT:    vadd.vx v14, v14, a1
1001; LMULMAX1-RV64-NEXT:    vadd.vx v15, v15, a1
1002; LMULMAX1-RV64-NEXT:    vadd.vx v11, v11, a1
1003; LMULMAX1-RV64-NEXT:    vadd.vx v10, v10, a1
1004; LMULMAX1-RV64-NEXT:    vadd.vx v9, v9, a1
1005; LMULMAX1-RV64-NEXT:    vadd.vx v8, v8, a1
1006; LMULMAX1-RV64-NEXT:    vse64.v v8, (a2)
1007; LMULMAX1-RV64-NEXT:    addi a0, a2, 96
1008; LMULMAX1-RV64-NEXT:    vse64.v v9, (a0)
1009; LMULMAX1-RV64-NEXT:    addi a0, a2, 112
1010; LMULMAX1-RV64-NEXT:    vse64.v v10, (a0)
1011; LMULMAX1-RV64-NEXT:    addi a0, a2, 64
1012; LMULMAX1-RV64-NEXT:    vse64.v v11, (a0)
1013; LMULMAX1-RV64-NEXT:    addi a0, a2, 80
1014; LMULMAX1-RV64-NEXT:    vse64.v v15, (a0)
1015; LMULMAX1-RV64-NEXT:    addi a0, a2, 32
1016; LMULMAX1-RV64-NEXT:    vse64.v v14, (a0)
1017; LMULMAX1-RV64-NEXT:    addi a0, a2, 48
1018; LMULMAX1-RV64-NEXT:    vse64.v v12, (a0)
1019; LMULMAX1-RV64-NEXT:    addi a2, a2, 16
1020; LMULMAX1-RV64-NEXT:    vse64.v v13, (a2)
1021; LMULMAX1-RV64-NEXT:    ret
1022  %va = load <16 x i64>, ptr %a
1023  %head = insertelement <16 x i64> poison, i64 %b, i32 0
1024  %splat = shufflevector <16 x i64> %head, <16 x i64> poison, <16 x i32> zeroinitializer
1025  %vc = add <16 x i64> %va, %splat
1026  store <16 x i64> %vc, ptr %c
1027  ret void
1028}
1029