1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc -mtriple=riscv64 -mattr=+zve32x \
3 ; RUN: -verify-machineinstrs < %s | FileCheck %s
5 ; Make sure we don't select a 0 vl to X0 in the custom isel handlers we use
6 ; for these intrinsics.
8 declare target("riscv.vector.tuple", <vscale x 32 x i8>, 2) @llvm.riscv.vlseg2.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2), ptr, i64, i64)
9 declare target("riscv.vector.tuple", <vscale x 32 x i8>, 2) @llvm.riscv.vlseg2.mask.nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2), ptr, <vscale x 16 x i1>, i64, i64, i64)
11 define <vscale x 16 x i16> @test_vlseg2_mask_nxv16i16(ptr %base, <vscale x 16 x i1> %mask) {
12 ; CHECK-LABEL: test_vlseg2_mask_nxv16i16:
13 ; CHECK: # %bb.0: # %entry
14 ; CHECK-NEXT: vsetivli zero, 0, e16, m4, ta, ma
15 ; CHECK-NEXT: vlseg2e16.v v4, (a0)
18 %0 = tail call target("riscv.vector.tuple", <vscale x 32 x i8>, 2) @llvm.riscv.vlseg2.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) undef, ptr %base, i64 0, i64 4)
19 %1 = tail call target("riscv.vector.tuple", <vscale x 32 x i8>, 2) @llvm.riscv.vlseg2.mask.nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %0, ptr %base, <vscale x 16 x i1> %mask, i64 0, i64 1, i64 4)
20 %2 = call <vscale x 16 x i16> @llvm.riscv.tuple.extract.nxv16i16.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %0, i32 1)
21 ret <vscale x 16 x i16> %2
24 declare target("riscv.vector.tuple", <vscale x 32 x i8>, 2) @llvm.riscv.vlsseg2.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2), ptr, i64, i64, i64)
25 declare target("riscv.vector.tuple", <vscale x 32 x i8>, 2) @llvm.riscv.vlsseg2.mask.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2), ptr, i64, <vscale x 16 x i1>, i64, i64, i64)
27 define <vscale x 16 x i16> @test_vlsseg2_mask_nxv16i16(ptr %base, i64 %offset, <vscale x 16 x i1> %mask) {
28 ; CHECK-LABEL: test_vlsseg2_mask_nxv16i16:
29 ; CHECK: # %bb.0: # %entry
30 ; CHECK-NEXT: vsetivli zero, 0, e16, m4, ta, mu
31 ; CHECK-NEXT: vlsseg2e16.v v4, (a0), a1
32 ; CHECK-NEXT: vlsseg2e16.v v4, (a0), a1, v0.t
35 %0 = tail call target("riscv.vector.tuple", <vscale x 32 x i8>, 2) @llvm.riscv.vlsseg2.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) undef, ptr %base, i64 %offset, i64 0, i64 4)
36 %1 = tail call target("riscv.vector.tuple", <vscale x 32 x i8>, 2) @llvm.riscv.vlsseg2.mask.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %0, ptr %base, i64 %offset, <vscale x 16 x i1> %mask, i64 0, i64 1, i64 4)
37 %2 = call <vscale x 16 x i16> @llvm.riscv.tuple.extract.nxv16i16.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %1, i32 1)
38 ret <vscale x 16 x i16> %2
40 declare target("riscv.vector.tuple", <vscale x 32 x i8>, 2) @llvm.riscv.vloxseg2.triscv.vector.tuple_nxv32i8_2t.nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2), ptr, <vscale x 16 x i16>, i64, i64)
41 declare target("riscv.vector.tuple", <vscale x 32 x i8>, 2) @llvm.riscv.vloxseg2.mask.triscv.vector.tuple_nxv32i8_2t.nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2), ptr, <vscale x 16 x i16>, <vscale x 16 x i1>, i64, i64, i64)
43 define <vscale x 16 x i16> @test_vloxseg2_mask_nxv16i16_nxv16i16(ptr %base, <vscale x 16 x i16> %index, <vscale x 16 x i1> %mask) {
44 ; CHECK-LABEL: test_vloxseg2_mask_nxv16i16_nxv16i16:
45 ; CHECK: # %bb.0: # %entry
46 ; CHECK-NEXT: vsetivli zero, 0, e16, m4, ta, mu
47 ; CHECK-NEXT: vloxseg2ei16.v v12, (a0), v8
48 ; CHECK-NEXT: vloxseg2ei16.v v12, (a0), v8, v0.t
49 ; CHECK-NEXT: vmv4r.v v8, v16
52 %0 = tail call target("riscv.vector.tuple", <vscale x 32 x i8>, 2) @llvm.riscv.vloxseg2.triscv.vector.tuple_nxv32i8_2t.nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) undef, ptr %base, <vscale x 16 x i16> %index, i64 0, i64 4)
53 %1 = tail call target("riscv.vector.tuple", <vscale x 32 x i8>, 2) @llvm.riscv.vloxseg2.mask.triscv.vector.tuple_nxv32i8_2t.nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %0, ptr %base, <vscale x 16 x i16> %index, <vscale x 16 x i1> %mask, i64 0, i64 1, i64 4)
54 %2 = call <vscale x 16 x i16> @llvm.riscv.tuple.extract.nxv16i16.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %1, i32 1)
55 ret <vscale x 16 x i16> %2
58 declare target("riscv.vector.tuple", <vscale x 32 x i8>, 2) @llvm.riscv.vluxseg2.triscv.vector.tuple_nxv32i8_2t.nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2), ptr, <vscale x 16 x i16>, i64, i64)
59 declare target("riscv.vector.tuple", <vscale x 32 x i8>, 2) @llvm.riscv.vluxseg2.mask.triscv.vector.tuple_nxv32i8_2t.nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2), ptr, <vscale x 16 x i16>, <vscale x 16 x i1>, i64, i64, i64)
61 define <vscale x 16 x i16> @test_vluxseg2_mask_nxv16i16_nxv16i16(ptr %base, <vscale x 16 x i16> %index, <vscale x 16 x i1> %mask) {
62 ; CHECK-LABEL: test_vluxseg2_mask_nxv16i16_nxv16i16:
63 ; CHECK: # %bb.0: # %entry
64 ; CHECK-NEXT: vsetivli zero, 0, e16, m4, ta, mu
65 ; CHECK-NEXT: vluxseg2ei16.v v12, (a0), v8
66 ; CHECK-NEXT: vluxseg2ei16.v v12, (a0), v8, v0.t
67 ; CHECK-NEXT: vmv4r.v v8, v16
70 %0 = tail call target("riscv.vector.tuple", <vscale x 32 x i8>, 2) @llvm.riscv.vluxseg2.triscv.vector.tuple_nxv32i8_2t.nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) undef, ptr %base, <vscale x 16 x i16> %index, i64 0, i64 4)
71 %1 = tail call target("riscv.vector.tuple", <vscale x 32 x i8>, 2) @llvm.riscv.vluxseg2.mask.triscv.vector.tuple_nxv32i8_2t.nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %0, ptr %base, <vscale x 16 x i16> %index, <vscale x 16 x i1> %mask, i64 0, i64 1, i64 4)
72 %2 = call <vscale x 16 x i16> @llvm.riscv.tuple.extract.nxv16i16.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %1, i32 1)
73 ret <vscale x 16 x i16> %2
76 declare {target("riscv.vector.tuple", <vscale x 32 x i8>, 2), i64} @llvm.riscv.vlseg2ff.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2), ptr , i64, i64)
77 declare {target("riscv.vector.tuple", <vscale x 32 x i8>, 2), i64} @llvm.riscv.vlseg2ff.mask.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2), ptr, <vscale x 16 x i1>, i64, i64, i64)
79 define <vscale x 16 x i16> @test_vlseg2ff_nxv16i16(ptr %base, ptr %outvl) {
80 ; CHECK-LABEL: test_vlseg2ff_nxv16i16:
81 ; CHECK: # %bb.0: # %entry
82 ; CHECK-NEXT: vsetivli zero, 0, e16, m4, ta, ma
83 ; CHECK-NEXT: vlseg2e16ff.v v8, (a0)
84 ; CHECK-NEXT: csrr a0, vl
85 ; CHECK-NEXT: sd a0, 0(a1)
88 %0 = tail call {target("riscv.vector.tuple", <vscale x 32 x i8>, 2), i64} @llvm.riscv.vlseg2ff.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) undef, ptr %base, i64 0, i64 4)
89 %1 = extractvalue {target("riscv.vector.tuple", <vscale x 32 x i8>, 2), i64} %0, 0
90 %2 = call <vscale x 16 x i16> @llvm.riscv.tuple.extract.nxv16i16.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %1, i32 0)
91 %3 = extractvalue {target("riscv.vector.tuple", <vscale x 32 x i8>, 2), i64} %0, 1
92 store i64 %3, ptr %outvl
93 ret <vscale x 16 x i16> %2
96 define <vscale x 16 x i16> @test_vlseg2ff_mask_nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %val, ptr %base, <vscale x 16 x i1> %mask, ptr %outvl) {
97 ; CHECK-LABEL: test_vlseg2ff_mask_nxv16i16:
98 ; CHECK: # %bb.0: # %entry
99 ; CHECK-NEXT: vsetivli zero, 0, e16, m4, ta, mu
100 ; CHECK-NEXT: vlseg2e16ff.v v8, (a0), v0.t
101 ; CHECK-NEXT: csrr a0, vl
102 ; CHECK-NEXT: sd a0, 0(a1)
105 %0 = tail call {target("riscv.vector.tuple", <vscale x 32 x i8>, 2), i64} @llvm.riscv.vlseg2ff.mask.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %val, ptr %base, <vscale x 16 x i1> %mask, i64 0, i64 1, i64 4)
106 %1 = extractvalue {target("riscv.vector.tuple", <vscale x 32 x i8>, 2), i64} %0, 0
107 %2 = call <vscale x 16 x i16> @llvm.riscv.tuple.extract.nxv16i16.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %1, i32 0)
108 %3 = extractvalue {target("riscv.vector.tuple", <vscale x 32 x i8>, 2), i64} %0, 1
109 store i64 %3, ptr %outvl
110 ret <vscale x 16 x i16> %2
113 declare void @llvm.riscv.vsseg2.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2), ptr , i64, i64)
114 declare void @llvm.riscv.vsseg2.mask.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2), ptr, <vscale x 16 x i1>, i64, i64)
116 define void @test_vsseg2_nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %val, ptr %base) {
117 ; CHECK-LABEL: test_vsseg2_nxv16i16:
118 ; CHECK: # %bb.0: # %entry
119 ; CHECK-NEXT: vsetivli zero, 0, e16, m4, ta, ma
120 ; CHECK-NEXT: vsseg2e16.v v8, (a0)
123 tail call void @llvm.riscv.vsseg2.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %val, ptr %base, i64 0, i64 4)
127 define void @test_vsseg2_mask_nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %val, ptr %base, <vscale x 16 x i1> %mask) {
128 ; CHECK-LABEL: test_vsseg2_mask_nxv16i16:
129 ; CHECK: # %bb.0: # %entry
130 ; CHECK-NEXT: vsetivli zero, 0, e16, m4, ta, ma
131 ; CHECK-NEXT: vsseg2e16.v v8, (a0), v0.t
134 tail call void @llvm.riscv.vsseg2.mask.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %val, ptr %base, <vscale x 16 x i1> %mask, i64 0, i64 4)
138 declare void @llvm.riscv.vssseg2.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2), ptr, i64, i64, i64)
139 declare void @llvm.riscv.vssseg2.mask.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2), ptr, i64, <vscale x 16 x i1>, i64, i64)
141 define void @test_vssseg2_nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %val, ptr %base, i64 %offset) {
142 ; CHECK-LABEL: test_vssseg2_nxv16i16:
143 ; CHECK: # %bb.0: # %entry
144 ; CHECK-NEXT: vsetivli zero, 0, e16, m4, ta, ma
145 ; CHECK-NEXT: vssseg2e16.v v8, (a0), a1
148 tail call void @llvm.riscv.vssseg2.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %val, ptr %base, i64 %offset, i64 0, i64 4)
152 define void @test_vssseg2_mask_nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %val, ptr %base, i64 %offset, <vscale x 16 x i1> %mask) {
153 ; CHECK-LABEL: test_vssseg2_mask_nxv16i16:
154 ; CHECK: # %bb.0: # %entry
155 ; CHECK-NEXT: vsetivli zero, 0, e16, m4, ta, ma
156 ; CHECK-NEXT: vssseg2e16.v v8, (a0), a1, v0.t
159 tail call void @llvm.riscv.vssseg2.mask.triscv.vector.tuple_nxv32i8_2t(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %val, ptr %base, i64 %offset, <vscale x 16 x i1> %mask, i64 0, i64 4)
163 declare void @llvm.riscv.vsoxseg2.triscv.vector.tuple_nxv32i8_2t.nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2), ptr, <vscale x 16 x i16>, i64, i64)
164 declare void @llvm.riscv.vsoxseg2.mask.triscv.vector.tuple_nxv32i8_2t.nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2), ptr, <vscale x 16 x i16>, <vscale x 16 x i1>, i64, i64)
166 define void @test_vsoxseg2_nxv16i16_nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %val, ptr %base, <vscale x 16 x i16> %index) {
167 ; CHECK-LABEL: test_vsoxseg2_nxv16i16_nxv16i16:
168 ; CHECK: # %bb.0: # %entry
169 ; CHECK-NEXT: vsetivli zero, 0, e16, m4, ta, ma
170 ; CHECK-NEXT: vsoxseg2ei16.v v8, (a0), v16
173 tail call void @llvm.riscv.vsoxseg2.triscv.vector.tuple_nxv32i8_2t.nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %val, ptr %base, <vscale x 16 x i16> %index, i64 0, i64 4)
177 define void @test_vsoxseg2_mask_nxv16i16_nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %val, ptr %base, <vscale x 16 x i16> %index, <vscale x 16 x i1> %mask) {
178 ; CHECK-LABEL: test_vsoxseg2_mask_nxv16i16_nxv16i16:
179 ; CHECK: # %bb.0: # %entry
180 ; CHECK-NEXT: vsetivli zero, 0, e16, m4, ta, ma
181 ; CHECK-NEXT: vsoxseg2ei16.v v8, (a0), v16, v0.t
184 tail call void @llvm.riscv.vsoxseg2.mask.triscv.vector.tuple_nxv32i8_2t.nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %val, ptr %base, <vscale x 16 x i16> %index, <vscale x 16 x i1> %mask, i64 0, i64 4)
188 declare void @llvm.riscv.vsuxseg2.triscv.vector.tuple_nxv32i8_2t.nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2), ptr, <vscale x 16 x i16>, i64, i64)
189 declare void @llvm.riscv.vsuxseg2.mask.triscv.vector.tuple_nxv32i8_2t.nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2), ptr, <vscale x 16 x i16>, <vscale x 16 x i1>, i64, i64)
191 define void @test_vsuxseg2_nxv16i16_nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %val, ptr %base, <vscale x 16 x i16> %index) {
192 ; CHECK-LABEL: test_vsuxseg2_nxv16i16_nxv16i16:
193 ; CHECK: # %bb.0: # %entry
194 ; CHECK-NEXT: vsetivli zero, 0, e16, m4, ta, ma
195 ; CHECK-NEXT: vsuxseg2ei16.v v8, (a0), v16
198 tail call void @llvm.riscv.vsuxseg2.triscv.vector.tuple_nxv32i8_2t.nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %val, ptr %base, <vscale x 16 x i16> %index, i64 0, i64 4)
202 define void @test_vsuxseg2_mask_nxv16i16_nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %val, ptr %base, <vscale x 16 x i16> %index, <vscale x 16 x i1> %mask) {
203 ; CHECK-LABEL: test_vsuxseg2_mask_nxv16i16_nxv16i16:
204 ; CHECK: # %bb.0: # %entry
205 ; CHECK-NEXT: vsetivli zero, 0, e16, m4, ta, ma
206 ; CHECK-NEXT: vsuxseg2ei16.v v8, (a0), v16, v0.t
209 tail call void @llvm.riscv.vsuxseg2.mask.triscv.vector.tuple_nxv32i8_2t.nxv16i16(target("riscv.vector.tuple", <vscale x 32 x i8>, 2) %val, ptr %base, <vscale x 16 x i16> %index, <vscale x 16 x i1> %mask, i64 0, i64 4)