xref: /llvm-project/llvm/test/CodeGen/RISCV/rvv/fnearbyint-sdnode.ll (revision 9122c5235ec85ce0c0ad337e862b006e7b349d84) 
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc -mtriple=riscv32 -mattr=+d,+zvfh,+zfbfmin,+zvfbfmin,+v \
3; RUN:     -target-abi=ilp32d -verify-machineinstrs < %s | FileCheck %s \
4; RUN:     --check-prefixes=CHECK,ZVFH
5; RUN: llc -mtriple=riscv64 -mattr=+d,+zvfh,+zfbfmin,+zvfbfmin,+v \
6; RUN:     -target-abi=lp64d -verify-machineinstrs < %s | FileCheck %s \
7; RUN:     --check-prefixes=CHECK,ZVFH
8; RUN: llc -mtriple=riscv32 -mattr=+d,+zfhmin,+zvfhmin,+zfbfmin,+zvfbfmin,+v \
9; RUN:     -target-abi=ilp32d -verify-machineinstrs < %s | FileCheck %s \
10; RUN:     --check-prefixes=CHECK,ZVFHMIN
11; RUN: llc -mtriple=riscv64 -mattr=+d,+zfhmin,+zvfhmin,+zfbfmin,+zvfbfmin,+v \
12; RUN:     -target-abi=lp64d -verify-machineinstrs < %s | FileCheck %s \
13; RUN:     --check-prefixes=CHECK,ZVFHMIN
14
15define <vscale x 1 x bfloat> @nearbyint_nxv1bf16(<vscale x 1 x bfloat> %x) {
16; CHECK-LABEL: nearbyint_nxv1bf16:
17; CHECK:       # %bb.0:
18; CHECK-NEXT:    vsetvli a0, zero, e16, mf4, ta, ma
19; CHECK-NEXT:    vfwcvtbf16.f.f.v v9, v8
20; CHECK-NEXT:    lui a0, 307200
21; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
22; CHECK-NEXT:    vfabs.v v8, v9
23; CHECK-NEXT:    fmv.w.x fa5, a0
24; CHECK-NEXT:    vmflt.vf v0, v8, fa5
25; CHECK-NEXT:    frflags a0
26; CHECK-NEXT:    vfcvt.x.f.v v8, v9, v0.t
27; CHECK-NEXT:    vfcvt.f.x.v v8, v8, v0.t
28; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, mu
29; CHECK-NEXT:    vfsgnj.vv v9, v8, v9, v0.t
30; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
31; CHECK-NEXT:    vfncvtbf16.f.f.w v8, v9
32; CHECK-NEXT:    fsflags a0
33; CHECK-NEXT:    ret
34  %a = call <vscale x 1 x bfloat> @llvm.nearbyint.nxv1bf16(<vscale x 1 x bfloat> %x)
35  ret <vscale x 1 x bfloat> %a
36}
37
38define <vscale x 2 x bfloat> @nearbyint_nxv2bf16(<vscale x 2 x bfloat> %x) {
39; CHECK-LABEL: nearbyint_nxv2bf16:
40; CHECK:       # %bb.0:
41; CHECK-NEXT:    vsetvli a0, zero, e16, mf2, ta, ma
42; CHECK-NEXT:    vfwcvtbf16.f.f.v v9, v8
43; CHECK-NEXT:    lui a0, 307200
44; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, ma
45; CHECK-NEXT:    vfabs.v v8, v9
46; CHECK-NEXT:    fmv.w.x fa5, a0
47; CHECK-NEXT:    vmflt.vf v0, v8, fa5
48; CHECK-NEXT:    frflags a0
49; CHECK-NEXT:    vfcvt.x.f.v v8, v9, v0.t
50; CHECK-NEXT:    vfcvt.f.x.v v8, v8, v0.t
51; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, mu
52; CHECK-NEXT:    vfsgnj.vv v9, v8, v9, v0.t
53; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, ma
54; CHECK-NEXT:    vfncvtbf16.f.f.w v8, v9
55; CHECK-NEXT:    fsflags a0
56; CHECK-NEXT:    ret
57  %a = call <vscale x 2 x bfloat> @llvm.nearbyint.nxv2bf16(<vscale x 2 x bfloat> %x)
58  ret <vscale x 2 x bfloat> %a
59}
60
61define <vscale x 4 x bfloat> @nearbyint_nxv4bf16(<vscale x 4 x bfloat> %x) {
62; CHECK-LABEL: nearbyint_nxv4bf16:
63; CHECK:       # %bb.0:
64; CHECK-NEXT:    vsetvli a0, zero, e16, m1, ta, ma
65; CHECK-NEXT:    vfwcvtbf16.f.f.v v10, v8
66; CHECK-NEXT:    lui a0, 307200
67; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
68; CHECK-NEXT:    vfabs.v v8, v10
69; CHECK-NEXT:    fmv.w.x fa5, a0
70; CHECK-NEXT:    vmflt.vf v0, v8, fa5
71; CHECK-NEXT:    frflags a0
72; CHECK-NEXT:    vfcvt.x.f.v v8, v10, v0.t
73; CHECK-NEXT:    vfcvt.f.x.v v8, v8, v0.t
74; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, mu
75; CHECK-NEXT:    vfsgnj.vv v10, v8, v10, v0.t
76; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, ma
77; CHECK-NEXT:    vfncvtbf16.f.f.w v8, v10
78; CHECK-NEXT:    fsflags a0
79; CHECK-NEXT:    ret
80  %a = call <vscale x 4 x bfloat> @llvm.nearbyint.nxv4bf16(<vscale x 4 x bfloat> %x)
81  ret <vscale x 4 x bfloat> %a
82}
83
84define <vscale x 8 x bfloat> @nearbyint_nxv8bf16(<vscale x 8 x bfloat> %x) {
85; CHECK-LABEL: nearbyint_nxv8bf16:
86; CHECK:       # %bb.0:
87; CHECK-NEXT:    vsetvli a0, zero, e16, m2, ta, ma
88; CHECK-NEXT:    vfwcvtbf16.f.f.v v12, v8
89; CHECK-NEXT:    lui a0, 307200
90; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
91; CHECK-NEXT:    vfabs.v v8, v12
92; CHECK-NEXT:    fmv.w.x fa5, a0
93; CHECK-NEXT:    vmflt.vf v0, v8, fa5
94; CHECK-NEXT:    frflags a0
95; CHECK-NEXT:    vfcvt.x.f.v v8, v12, v0.t
96; CHECK-NEXT:    vfcvt.f.x.v v8, v8, v0.t
97; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, mu
98; CHECK-NEXT:    vfsgnj.vv v12, v8, v12, v0.t
99; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
100; CHECK-NEXT:    vfncvtbf16.f.f.w v8, v12
101; CHECK-NEXT:    fsflags a0
102; CHECK-NEXT:    ret
103  %a = call <vscale x 8 x bfloat> @llvm.nearbyint.nxv8bf16(<vscale x 8 x bfloat> %x)
104  ret <vscale x 8 x bfloat> %a
105}
106
107define <vscale x 16 x bfloat> @nearbyint_nxv16bf16(<vscale x 16 x bfloat> %x) {
108; CHECK-LABEL: nearbyint_nxv16bf16:
109; CHECK:       # %bb.0:
110; CHECK-NEXT:    vsetvli a0, zero, e16, m4, ta, ma
111; CHECK-NEXT:    vfwcvtbf16.f.f.v v16, v8
112; CHECK-NEXT:    lui a0, 307200
113; CHECK-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
114; CHECK-NEXT:    vfabs.v v8, v16
115; CHECK-NEXT:    fmv.w.x fa5, a0
116; CHECK-NEXT:    vmflt.vf v0, v8, fa5
117; CHECK-NEXT:    frflags a0
118; CHECK-NEXT:    vfcvt.x.f.v v8, v16, v0.t
119; CHECK-NEXT:    vfcvt.f.x.v v8, v8, v0.t
120; CHECK-NEXT:    vsetvli zero, zero, e32, m8, ta, mu
121; CHECK-NEXT:    vfsgnj.vv v16, v8, v16, v0.t
122; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
123; CHECK-NEXT:    vfncvtbf16.f.f.w v8, v16
124; CHECK-NEXT:    fsflags a0
125; CHECK-NEXT:    ret
126  %a = call <vscale x 16 x bfloat> @llvm.nearbyint.nxv16bf16(<vscale x 16 x bfloat> %x)
127  ret <vscale x 16 x bfloat> %a
128}
129
130define <vscale x 32 x bfloat> @nearbyint_nxv32bf16(<vscale x 32 x bfloat> %x) {
131; CHECK-LABEL: nearbyint_nxv32bf16:
132; CHECK:       # %bb.0:
133; CHECK-NEXT:    addi sp, sp, -16
134; CHECK-NEXT:    .cfi_def_cfa_offset 16
135; CHECK-NEXT:    csrr a0, vlenb
136; CHECK-NEXT:    slli a0, a0, 3
137; CHECK-NEXT:    sub sp, sp, a0
138; CHECK-NEXT:    .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 16 + 8 * vlenb
139; CHECK-NEXT:    vsetvli a0, zero, e16, m4, ta, ma
140; CHECK-NEXT:    vfwcvtbf16.f.f.v v16, v8
141; CHECK-NEXT:    lui a0, 307200
142; CHECK-NEXT:    fmv.w.x fa5, a0
143; CHECK-NEXT:    frflags a0
144; CHECK-NEXT:    vfwcvtbf16.f.f.v v24, v12
145; CHECK-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
146; CHECK-NEXT:    vfabs.v v8, v16
147; CHECK-NEXT:    vmflt.vf v0, v8, fa5
148; CHECK-NEXT:    vfabs.v v8, v24
149; CHECK-NEXT:    vmflt.vf v7, v8, fa5
150; CHECK-NEXT:    vfcvt.x.f.v v8, v16, v0.t
151; CHECK-NEXT:    vfcvt.f.x.v v8, v8, v0.t
152; CHECK-NEXT:    fsflags a0
153; CHECK-NEXT:    vsetvli zero, zero, e32, m8, ta, mu
154; CHECK-NEXT:    vfsgnj.vv v16, v8, v16, v0.t
155; CHECK-NEXT:    frflags a0
156; CHECK-NEXT:    vmv1r.v v0, v7
157; CHECK-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
158; CHECK-NEXT:    vfcvt.x.f.v v8, v24, v0.t
159; CHECK-NEXT:    addi a1, sp, 16
160; CHECK-NEXT:    vs8r.v v8, (a1) # Unknown-size Folded Spill
161; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
162; CHECK-NEXT:    vfncvtbf16.f.f.w v8, v16
163; CHECK-NEXT:    vl8r.v v16, (a1) # Unknown-size Folded Reload
164; CHECK-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
165; CHECK-NEXT:    vfcvt.f.x.v v16, v16, v0.t
166; CHECK-NEXT:    vsetvli zero, zero, e32, m8, ta, mu
167; CHECK-NEXT:    vfsgnj.vv v24, v16, v24, v0.t
168; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
169; CHECK-NEXT:    vfncvtbf16.f.f.w v12, v24
170; CHECK-NEXT:    fsflags a0
171; CHECK-NEXT:    csrr a0, vlenb
172; CHECK-NEXT:    slli a0, a0, 3
173; CHECK-NEXT:    add sp, sp, a0
174; CHECK-NEXT:    .cfi_def_cfa sp, 16
175; CHECK-NEXT:    addi sp, sp, 16
176; CHECK-NEXT:    .cfi_def_cfa_offset 0
177; CHECK-NEXT:    ret
178  %a = call <vscale x 32 x bfloat> @llvm.nearbyint.nxv32bf16(<vscale x 32 x bfloat> %x)
179  ret <vscale x 32 x bfloat> %a
180}
181
182define <vscale x 1 x half> @nearbyint_nxv1f16(<vscale x 1 x half> %x) {
183; ZVFH-LABEL: nearbyint_nxv1f16:
184; ZVFH:       # %bb.0:
185; ZVFH-NEXT:    lui a0, %hi(.LCPI6_0)
186; ZVFH-NEXT:    flh fa5, %lo(.LCPI6_0)(a0)
187; ZVFH-NEXT:    vsetvli a0, zero, e16, mf4, ta, ma
188; ZVFH-NEXT:    vfabs.v v9, v8
189; ZVFH-NEXT:    vmflt.vf v0, v9, fa5
190; ZVFH-NEXT:    frflags a0
191; ZVFH-NEXT:    vfcvt.x.f.v v9, v8, v0.t
192; ZVFH-NEXT:    vfcvt.f.x.v v9, v9, v0.t
193; ZVFH-NEXT:    vsetvli zero, zero, e16, mf4, ta, mu
194; ZVFH-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
195; ZVFH-NEXT:    fsflags a0
196; ZVFH-NEXT:    ret
197;
198; ZVFHMIN-LABEL: nearbyint_nxv1f16:
199; ZVFHMIN:       # %bb.0:
200; ZVFHMIN-NEXT:    vsetvli a0, zero, e16, mf4, ta, ma
201; ZVFHMIN-NEXT:    vfwcvt.f.f.v v9, v8
202; ZVFHMIN-NEXT:    lui a0, 307200
203; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
204; ZVFHMIN-NEXT:    vfabs.v v8, v9
205; ZVFHMIN-NEXT:    fmv.w.x fa5, a0
206; ZVFHMIN-NEXT:    vmflt.vf v0, v8, fa5
207; ZVFHMIN-NEXT:    frflags a0
208; ZVFHMIN-NEXT:    vfcvt.x.f.v v8, v9, v0.t
209; ZVFHMIN-NEXT:    vfcvt.f.x.v v8, v8, v0.t
210; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, mf2, ta, mu
211; ZVFHMIN-NEXT:    vfsgnj.vv v9, v8, v9, v0.t
212; ZVFHMIN-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
213; ZVFHMIN-NEXT:    vfncvt.f.f.w v8, v9
214; ZVFHMIN-NEXT:    fsflags a0
215; ZVFHMIN-NEXT:    ret
216  %a = call <vscale x 1 x half> @llvm.nearbyint.nxv1f16(<vscale x 1 x half> %x)
217  ret <vscale x 1 x half> %a
218}
219declare <vscale x 1 x half> @llvm.nearbyint.nxv1f16(<vscale x 1 x half>)
220
221define <vscale x 2 x half> @nearbyint_nxv2f16(<vscale x 2 x half> %x) {
222; ZVFH-LABEL: nearbyint_nxv2f16:
223; ZVFH:       # %bb.0:
224; ZVFH-NEXT:    lui a0, %hi(.LCPI7_0)
225; ZVFH-NEXT:    flh fa5, %lo(.LCPI7_0)(a0)
226; ZVFH-NEXT:    vsetvli a0, zero, e16, mf2, ta, ma
227; ZVFH-NEXT:    vfabs.v v9, v8
228; ZVFH-NEXT:    vmflt.vf v0, v9, fa5
229; ZVFH-NEXT:    frflags a0
230; ZVFH-NEXT:    vfcvt.x.f.v v9, v8, v0.t
231; ZVFH-NEXT:    vfcvt.f.x.v v9, v9, v0.t
232; ZVFH-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
233; ZVFH-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
234; ZVFH-NEXT:    fsflags a0
235; ZVFH-NEXT:    ret
236;
237; ZVFHMIN-LABEL: nearbyint_nxv2f16:
238; ZVFHMIN:       # %bb.0:
239; ZVFHMIN-NEXT:    vsetvli a0, zero, e16, mf2, ta, ma
240; ZVFHMIN-NEXT:    vfwcvt.f.f.v v9, v8
241; ZVFHMIN-NEXT:    lui a0, 307200
242; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m1, ta, ma
243; ZVFHMIN-NEXT:    vfabs.v v8, v9
244; ZVFHMIN-NEXT:    fmv.w.x fa5, a0
245; ZVFHMIN-NEXT:    vmflt.vf v0, v8, fa5
246; ZVFHMIN-NEXT:    frflags a0
247; ZVFHMIN-NEXT:    vfcvt.x.f.v v8, v9, v0.t
248; ZVFHMIN-NEXT:    vfcvt.f.x.v v8, v8, v0.t
249; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m1, ta, mu
250; ZVFHMIN-NEXT:    vfsgnj.vv v9, v8, v9, v0.t
251; ZVFHMIN-NEXT:    vsetvli zero, zero, e16, mf2, ta, ma
252; ZVFHMIN-NEXT:    vfncvt.f.f.w v8, v9
253; ZVFHMIN-NEXT:    fsflags a0
254; ZVFHMIN-NEXT:    ret
255  %a = call <vscale x 2 x half> @llvm.nearbyint.nxv2f16(<vscale x 2 x half> %x)
256  ret <vscale x 2 x half> %a
257}
258declare <vscale x 2 x half> @llvm.nearbyint.nxv2f16(<vscale x 2 x half>)
259
260define <vscale x 4 x half> @nearbyint_nxv4f16(<vscale x 4 x half> %x) {
261; ZVFH-LABEL: nearbyint_nxv4f16:
262; ZVFH:       # %bb.0:
263; ZVFH-NEXT:    lui a0, %hi(.LCPI8_0)
264; ZVFH-NEXT:    flh fa5, %lo(.LCPI8_0)(a0)
265; ZVFH-NEXT:    vsetvli a0, zero, e16, m1, ta, ma
266; ZVFH-NEXT:    vfabs.v v9, v8
267; ZVFH-NEXT:    vmflt.vf v0, v9, fa5
268; ZVFH-NEXT:    frflags a0
269; ZVFH-NEXT:    vfcvt.x.f.v v9, v8, v0.t
270; ZVFH-NEXT:    vfcvt.f.x.v v9, v9, v0.t
271; ZVFH-NEXT:    vsetvli zero, zero, e16, m1, ta, mu
272; ZVFH-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
273; ZVFH-NEXT:    fsflags a0
274; ZVFH-NEXT:    ret
275;
276; ZVFHMIN-LABEL: nearbyint_nxv4f16:
277; ZVFHMIN:       # %bb.0:
278; ZVFHMIN-NEXT:    vsetvli a0, zero, e16, m1, ta, ma
279; ZVFHMIN-NEXT:    vfwcvt.f.f.v v10, v8
280; ZVFHMIN-NEXT:    lui a0, 307200
281; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
282; ZVFHMIN-NEXT:    vfabs.v v8, v10
283; ZVFHMIN-NEXT:    fmv.w.x fa5, a0
284; ZVFHMIN-NEXT:    vmflt.vf v0, v8, fa5
285; ZVFHMIN-NEXT:    frflags a0
286; ZVFHMIN-NEXT:    vfcvt.x.f.v v8, v10, v0.t
287; ZVFHMIN-NEXT:    vfcvt.f.x.v v8, v8, v0.t
288; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m2, ta, mu
289; ZVFHMIN-NEXT:    vfsgnj.vv v10, v8, v10, v0.t
290; ZVFHMIN-NEXT:    vsetvli zero, zero, e16, m1, ta, ma
291; ZVFHMIN-NEXT:    vfncvt.f.f.w v8, v10
292; ZVFHMIN-NEXT:    fsflags a0
293; ZVFHMIN-NEXT:    ret
294  %a = call <vscale x 4 x half> @llvm.nearbyint.nxv4f16(<vscale x 4 x half> %x)
295  ret <vscale x 4 x half> %a
296}
297declare <vscale x 4 x half> @llvm.nearbyint.nxv4f16(<vscale x 4 x half>)
298
299define <vscale x 8 x half> @nearbyint_nxv8f16(<vscale x 8 x half> %x) {
300; ZVFH-LABEL: nearbyint_nxv8f16:
301; ZVFH:       # %bb.0:
302; ZVFH-NEXT:    lui a0, %hi(.LCPI9_0)
303; ZVFH-NEXT:    flh fa5, %lo(.LCPI9_0)(a0)
304; ZVFH-NEXT:    vsetvli a0, zero, e16, m2, ta, ma
305; ZVFH-NEXT:    vfabs.v v10, v8
306; ZVFH-NEXT:    vmflt.vf v0, v10, fa5
307; ZVFH-NEXT:    frflags a0
308; ZVFH-NEXT:    vfcvt.x.f.v v10, v8, v0.t
309; ZVFH-NEXT:    vfcvt.f.x.v v10, v10, v0.t
310; ZVFH-NEXT:    vsetvli zero, zero, e16, m2, ta, mu
311; ZVFH-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
312; ZVFH-NEXT:    fsflags a0
313; ZVFH-NEXT:    ret
314;
315; ZVFHMIN-LABEL: nearbyint_nxv8f16:
316; ZVFHMIN:       # %bb.0:
317; ZVFHMIN-NEXT:    vsetvli a0, zero, e16, m2, ta, ma
318; ZVFHMIN-NEXT:    vfwcvt.f.f.v v12, v8
319; ZVFHMIN-NEXT:    lui a0, 307200
320; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
321; ZVFHMIN-NEXT:    vfabs.v v8, v12
322; ZVFHMIN-NEXT:    fmv.w.x fa5, a0
323; ZVFHMIN-NEXT:    vmflt.vf v0, v8, fa5
324; ZVFHMIN-NEXT:    frflags a0
325; ZVFHMIN-NEXT:    vfcvt.x.f.v v8, v12, v0.t
326; ZVFHMIN-NEXT:    vfcvt.f.x.v v8, v8, v0.t
327; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m4, ta, mu
328; ZVFHMIN-NEXT:    vfsgnj.vv v12, v8, v12, v0.t
329; ZVFHMIN-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
330; ZVFHMIN-NEXT:    vfncvt.f.f.w v8, v12
331; ZVFHMIN-NEXT:    fsflags a0
332; ZVFHMIN-NEXT:    ret
333  %a = call <vscale x 8 x half> @llvm.nearbyint.nxv8f16(<vscale x 8 x half> %x)
334  ret <vscale x 8 x half> %a
335}
336declare <vscale x 8 x half> @llvm.nearbyint.nxv8f16(<vscale x 8 x half>)
337
338define <vscale x 16 x half> @nearbyint_nxv16f16(<vscale x 16 x half> %x) {
339; ZVFH-LABEL: nearbyint_nxv16f16:
340; ZVFH:       # %bb.0:
341; ZVFH-NEXT:    lui a0, %hi(.LCPI10_0)
342; ZVFH-NEXT:    flh fa5, %lo(.LCPI10_0)(a0)
343; ZVFH-NEXT:    vsetvli a0, zero, e16, m4, ta, ma
344; ZVFH-NEXT:    vfabs.v v12, v8
345; ZVFH-NEXT:    vmflt.vf v0, v12, fa5
346; ZVFH-NEXT:    frflags a0
347; ZVFH-NEXT:    vfcvt.x.f.v v12, v8, v0.t
348; ZVFH-NEXT:    vfcvt.f.x.v v12, v12, v0.t
349; ZVFH-NEXT:    vsetvli zero, zero, e16, m4, ta, mu
350; ZVFH-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
351; ZVFH-NEXT:    fsflags a0
352; ZVFH-NEXT:    ret
353;
354; ZVFHMIN-LABEL: nearbyint_nxv16f16:
355; ZVFHMIN:       # %bb.0:
356; ZVFHMIN-NEXT:    vsetvli a0, zero, e16, m4, ta, ma
357; ZVFHMIN-NEXT:    vfwcvt.f.f.v v16, v8
358; ZVFHMIN-NEXT:    lui a0, 307200
359; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
360; ZVFHMIN-NEXT:    vfabs.v v8, v16
361; ZVFHMIN-NEXT:    fmv.w.x fa5, a0
362; ZVFHMIN-NEXT:    vmflt.vf v0, v8, fa5
363; ZVFHMIN-NEXT:    frflags a0
364; ZVFHMIN-NEXT:    vfcvt.x.f.v v8, v16, v0.t
365; ZVFHMIN-NEXT:    vfcvt.f.x.v v8, v8, v0.t
366; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m8, ta, mu
367; ZVFHMIN-NEXT:    vfsgnj.vv v16, v8, v16, v0.t
368; ZVFHMIN-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
369; ZVFHMIN-NEXT:    vfncvt.f.f.w v8, v16
370; ZVFHMIN-NEXT:    fsflags a0
371; ZVFHMIN-NEXT:    ret
372  %a = call <vscale x 16 x half> @llvm.nearbyint.nxv16f16(<vscale x 16 x half> %x)
373  ret <vscale x 16 x half> %a
374}
375declare <vscale x 16 x half> @llvm.nearbyint.nxv16f16(<vscale x 16 x half>)
376
377define <vscale x 32 x half> @nearbyint_nxv32f16(<vscale x 32 x half> %x) {
378; ZVFH-LABEL: nearbyint_nxv32f16:
379; ZVFH:       # %bb.0:
380; ZVFH-NEXT:    lui a0, %hi(.LCPI11_0)
381; ZVFH-NEXT:    flh fa5, %lo(.LCPI11_0)(a0)
382; ZVFH-NEXT:    vsetvli a0, zero, e16, m8, ta, ma
383; ZVFH-NEXT:    vfabs.v v16, v8
384; ZVFH-NEXT:    vmflt.vf v0, v16, fa5
385; ZVFH-NEXT:    frflags a0
386; ZVFH-NEXT:    vfcvt.x.f.v v16, v8, v0.t
387; ZVFH-NEXT:    vfcvt.f.x.v v16, v16, v0.t
388; ZVFH-NEXT:    vsetvli zero, zero, e16, m8, ta, mu
389; ZVFH-NEXT:    vfsgnj.vv v8, v16, v8, v0.t
390; ZVFH-NEXT:    fsflags a0
391; ZVFH-NEXT:    ret
392;
393; ZVFHMIN-LABEL: nearbyint_nxv32f16:
394; ZVFHMIN:       # %bb.0:
395; ZVFHMIN-NEXT:    addi sp, sp, -16
396; ZVFHMIN-NEXT:    .cfi_def_cfa_offset 16
397; ZVFHMIN-NEXT:    csrr a0, vlenb
398; ZVFHMIN-NEXT:    slli a0, a0, 3
399; ZVFHMIN-NEXT:    sub sp, sp, a0
400; ZVFHMIN-NEXT:    .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 16 + 8 * vlenb
401; ZVFHMIN-NEXT:    vsetvli a0, zero, e16, m4, ta, ma
402; ZVFHMIN-NEXT:    vfwcvt.f.f.v v16, v8
403; ZVFHMIN-NEXT:    lui a0, 307200
404; ZVFHMIN-NEXT:    fmv.w.x fa5, a0
405; ZVFHMIN-NEXT:    frflags a0
406; ZVFHMIN-NEXT:    vfwcvt.f.f.v v24, v12
407; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
408; ZVFHMIN-NEXT:    vfabs.v v8, v16
409; ZVFHMIN-NEXT:    vmflt.vf v0, v8, fa5
410; ZVFHMIN-NEXT:    vfabs.v v8, v24
411; ZVFHMIN-NEXT:    vmflt.vf v7, v8, fa5
412; ZVFHMIN-NEXT:    vfcvt.x.f.v v8, v16, v0.t
413; ZVFHMIN-NEXT:    vfcvt.f.x.v v8, v8, v0.t
414; ZVFHMIN-NEXT:    fsflags a0
415; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m8, ta, mu
416; ZVFHMIN-NEXT:    vfsgnj.vv v16, v8, v16, v0.t
417; ZVFHMIN-NEXT:    frflags a0
418; ZVFHMIN-NEXT:    vmv1r.v v0, v7
419; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
420; ZVFHMIN-NEXT:    vfcvt.x.f.v v8, v24, v0.t
421; ZVFHMIN-NEXT:    addi a1, sp, 16
422; ZVFHMIN-NEXT:    vs8r.v v8, (a1) # Unknown-size Folded Spill
423; ZVFHMIN-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
424; ZVFHMIN-NEXT:    vfncvt.f.f.w v8, v16
425; ZVFHMIN-NEXT:    vl8r.v v16, (a1) # Unknown-size Folded Reload
426; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
427; ZVFHMIN-NEXT:    vfcvt.f.x.v v16, v16, v0.t
428; ZVFHMIN-NEXT:    vsetvli zero, zero, e32, m8, ta, mu
429; ZVFHMIN-NEXT:    vfsgnj.vv v24, v16, v24, v0.t
430; ZVFHMIN-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
431; ZVFHMIN-NEXT:    vfncvt.f.f.w v12, v24
432; ZVFHMIN-NEXT:    fsflags a0
433; ZVFHMIN-NEXT:    csrr a0, vlenb
434; ZVFHMIN-NEXT:    slli a0, a0, 3
435; ZVFHMIN-NEXT:    add sp, sp, a0
436; ZVFHMIN-NEXT:    .cfi_def_cfa sp, 16
437; ZVFHMIN-NEXT:    addi sp, sp, 16
438; ZVFHMIN-NEXT:    .cfi_def_cfa_offset 0
439; ZVFHMIN-NEXT:    ret
440  %a = call <vscale x 32 x half> @llvm.nearbyint.nxv32f16(<vscale x 32 x half> %x)
441  ret <vscale x 32 x half> %a
442}
443declare <vscale x 32 x half> @llvm.nearbyint.nxv32f16(<vscale x 32 x half>)
444
445define <vscale x 1 x float> @nearbyint_nxv1f32(<vscale x 1 x float> %x) {
446; CHECK-LABEL: nearbyint_nxv1f32:
447; CHECK:       # %bb.0:
448; CHECK-NEXT:    vsetvli a0, zero, e32, mf2, ta, ma
449; CHECK-NEXT:    vfabs.v v9, v8
450; CHECK-NEXT:    lui a0, 307200
451; CHECK-NEXT:    fmv.w.x fa5, a0
452; CHECK-NEXT:    vmflt.vf v0, v9, fa5
453; CHECK-NEXT:    frflags a0
454; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
455; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
456; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, mu
457; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
458; CHECK-NEXT:    fsflags a0
459; CHECK-NEXT:    ret
460  %a = call <vscale x 1 x float> @llvm.nearbyint.nxv1f32(<vscale x 1 x float> %x)
461  ret <vscale x 1 x float> %a
462}
463declare <vscale x 1 x float> @llvm.nearbyint.nxv1f32(<vscale x 1 x float>)
464
465define <vscale x 2 x float> @nearbyint_nxv2f32(<vscale x 2 x float> %x) {
466; CHECK-LABEL: nearbyint_nxv2f32:
467; CHECK:       # %bb.0:
468; CHECK-NEXT:    vsetvli a0, zero, e32, m1, ta, ma
469; CHECK-NEXT:    vfabs.v v9, v8
470; CHECK-NEXT:    lui a0, 307200
471; CHECK-NEXT:    fmv.w.x fa5, a0
472; CHECK-NEXT:    vmflt.vf v0, v9, fa5
473; CHECK-NEXT:    frflags a0
474; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
475; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
476; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, mu
477; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
478; CHECK-NEXT:    fsflags a0
479; CHECK-NEXT:    ret
480  %a = call <vscale x 2 x float> @llvm.nearbyint.nxv2f32(<vscale x 2 x float> %x)
481  ret <vscale x 2 x float> %a
482}
483declare <vscale x 2 x float> @llvm.nearbyint.nxv2f32(<vscale x 2 x float>)
484
485define <vscale x 4 x float> @nearbyint_nxv4f32(<vscale x 4 x float> %x) {
486; CHECK-LABEL: nearbyint_nxv4f32:
487; CHECK:       # %bb.0:
488; CHECK-NEXT:    vsetvli a0, zero, e32, m2, ta, ma
489; CHECK-NEXT:    vfabs.v v10, v8
490; CHECK-NEXT:    lui a0, 307200
491; CHECK-NEXT:    fmv.w.x fa5, a0
492; CHECK-NEXT:    vmflt.vf v0, v10, fa5
493; CHECK-NEXT:    frflags a0
494; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
495; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
496; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, mu
497; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
498; CHECK-NEXT:    fsflags a0
499; CHECK-NEXT:    ret
500  %a = call <vscale x 4 x float> @llvm.nearbyint.nxv4f32(<vscale x 4 x float> %x)
501  ret <vscale x 4 x float> %a
502}
503declare <vscale x 4 x float> @llvm.nearbyint.nxv4f32(<vscale x 4 x float>)
504
505define <vscale x 8 x float> @nearbyint_nxv8f32(<vscale x 8 x float> %x) {
506; CHECK-LABEL: nearbyint_nxv8f32:
507; CHECK:       # %bb.0:
508; CHECK-NEXT:    vsetvli a0, zero, e32, m4, ta, ma
509; CHECK-NEXT:    vfabs.v v12, v8
510; CHECK-NEXT:    lui a0, 307200
511; CHECK-NEXT:    fmv.w.x fa5, a0
512; CHECK-NEXT:    vmflt.vf v0, v12, fa5
513; CHECK-NEXT:    frflags a0
514; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
515; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
516; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, mu
517; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
518; CHECK-NEXT:    fsflags a0
519; CHECK-NEXT:    ret
520  %a = call <vscale x 8 x float> @llvm.nearbyint.nxv8f32(<vscale x 8 x float> %x)
521  ret <vscale x 8 x float> %a
522}
523declare <vscale x 8 x float> @llvm.nearbyint.nxv8f32(<vscale x 8 x float>)
524
525define <vscale x 16 x float> @nearbyint_nxv16f32(<vscale x 16 x float> %x) {
526; CHECK-LABEL: nearbyint_nxv16f32:
527; CHECK:       # %bb.0:
528; CHECK-NEXT:    vsetvli a0, zero, e32, m8, ta, ma
529; CHECK-NEXT:    vfabs.v v16, v8
530; CHECK-NEXT:    lui a0, 307200
531; CHECK-NEXT:    fmv.w.x fa5, a0
532; CHECK-NEXT:    vmflt.vf v0, v16, fa5
533; CHECK-NEXT:    frflags a0
534; CHECK-NEXT:    vfcvt.x.f.v v16, v8, v0.t
535; CHECK-NEXT:    vfcvt.f.x.v v16, v16, v0.t
536; CHECK-NEXT:    vsetvli zero, zero, e32, m8, ta, mu
537; CHECK-NEXT:    vfsgnj.vv v8, v16, v8, v0.t
538; CHECK-NEXT:    fsflags a0
539; CHECK-NEXT:    ret
540  %a = call <vscale x 16 x float> @llvm.nearbyint.nxv16f32(<vscale x 16 x float> %x)
541  ret <vscale x 16 x float> %a
542}
543declare <vscale x 16 x float> @llvm.nearbyint.nxv16f32(<vscale x 16 x float>)
544
545define <vscale x 1 x double> @nearbyint_nxv1f64(<vscale x 1 x double> %x) {
546; CHECK-LABEL: nearbyint_nxv1f64:
547; CHECK:       # %bb.0:
548; CHECK-NEXT:    lui a0, %hi(.LCPI17_0)
549; CHECK-NEXT:    fld fa5, %lo(.LCPI17_0)(a0)
550; CHECK-NEXT:    vsetvli a0, zero, e64, m1, ta, ma
551; CHECK-NEXT:    vfabs.v v9, v8
552; CHECK-NEXT:    vmflt.vf v0, v9, fa5
553; CHECK-NEXT:    frflags a0
554; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
555; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
556; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, mu
557; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
558; CHECK-NEXT:    fsflags a0
559; CHECK-NEXT:    ret
560  %a = call <vscale x 1 x double> @llvm.nearbyint.nxv1f64(<vscale x 1 x double> %x)
561  ret <vscale x 1 x double> %a
562}
563declare <vscale x 1 x double> @llvm.nearbyint.nxv1f64(<vscale x 1 x double>)
564
565define <vscale x 2 x double> @nearbyint_nxv2f64(<vscale x 2 x double> %x) {
566; CHECK-LABEL: nearbyint_nxv2f64:
567; CHECK:       # %bb.0:
568; CHECK-NEXT:    lui a0, %hi(.LCPI18_0)
569; CHECK-NEXT:    fld fa5, %lo(.LCPI18_0)(a0)
570; CHECK-NEXT:    vsetvli a0, zero, e64, m2, ta, ma
571; CHECK-NEXT:    vfabs.v v10, v8
572; CHECK-NEXT:    vmflt.vf v0, v10, fa5
573; CHECK-NEXT:    frflags a0
574; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
575; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
576; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, mu
577; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
578; CHECK-NEXT:    fsflags a0
579; CHECK-NEXT:    ret
580  %a = call <vscale x 2 x double> @llvm.nearbyint.nxv2f64(<vscale x 2 x double> %x)
581  ret <vscale x 2 x double> %a
582}
583declare <vscale x 2 x double> @llvm.nearbyint.nxv2f64(<vscale x 2 x double>)
584
585define <vscale x 4 x double> @nearbyint_nxv4f64(<vscale x 4 x double> %x) {
586; CHECK-LABEL: nearbyint_nxv4f64:
587; CHECK:       # %bb.0:
588; CHECK-NEXT:    lui a0, %hi(.LCPI19_0)
589; CHECK-NEXT:    fld fa5, %lo(.LCPI19_0)(a0)
590; CHECK-NEXT:    vsetvli a0, zero, e64, m4, ta, ma
591; CHECK-NEXT:    vfabs.v v12, v8
592; CHECK-NEXT:    vmflt.vf v0, v12, fa5
593; CHECK-NEXT:    frflags a0
594; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
595; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
596; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, mu
597; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
598; CHECK-NEXT:    fsflags a0
599; CHECK-NEXT:    ret
600  %a = call <vscale x 4 x double> @llvm.nearbyint.nxv4f64(<vscale x 4 x double> %x)
601  ret <vscale x 4 x double> %a
602}
603declare <vscale x 4 x double> @llvm.nearbyint.nxv4f64(<vscale x 4 x double>)
604
605define <vscale x 8 x double> @nearbyint_nxv8f64(<vscale x 8 x double> %x) {
606; CHECK-LABEL: nearbyint_nxv8f64:
607; CHECK:       # %bb.0:
608; CHECK-NEXT:    lui a0, %hi(.LCPI20_0)
609; CHECK-NEXT:    fld fa5, %lo(.LCPI20_0)(a0)
610; CHECK-NEXT:    vsetvli a0, zero, e64, m8, ta, ma
611; CHECK-NEXT:    vfabs.v v16, v8
612; CHECK-NEXT:    vmflt.vf v0, v16, fa5
613; CHECK-NEXT:    frflags a0
614; CHECK-NEXT:    vfcvt.x.f.v v16, v8, v0.t
615; CHECK-NEXT:    vfcvt.f.x.v v16, v16, v0.t
616; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, mu
617; CHECK-NEXT:    vfsgnj.vv v8, v16, v8, v0.t
618; CHECK-NEXT:    fsflags a0
619; CHECK-NEXT:    ret
620  %a = call <vscale x 8 x double> @llvm.nearbyint.nxv8f64(<vscale x 8 x double> %x)
621  ret <vscale x 8 x double> %a
622}
623declare <vscale x 8 x double> @llvm.nearbyint.nxv8f64(<vscale x 8 x double>)
624