xref: /llvm-project/llvm/test/CodeGen/VE/Vector/loadvr.ll (revision b006b60dc993b2e0ba3e412c80709477241b6be6)
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc < %s -mtriple=ve-unknown-unknown -mattr=+vpu | FileCheck %s
3
4@v256i64 = common dso_local local_unnamed_addr global <256 x i64> zeroinitializer, align 16
5
6; Function Attrs: norecurse nounwind readonly
7define fastcc <256 x i64> @loadv256i64(ptr nocapture readonly) {
8; CHECK-LABEL: loadv256i64:
9; CHECK:       # %bb.0:
10; CHECK-NEXT:    lea %s1, 256
11; CHECK-NEXT:    lvl %s1
12; CHECK-NEXT:    vld %v0, 8, %s0
13; CHECK-NEXT:    b.l.t (, %s10)
14  %2 = load <256 x i64>, ptr %0, align 16
15  ret <256 x i64> %2
16}
17
18; Function Attrs: norecurse nounwind readonly
19define fastcc <256 x double> @loadv256f64(ptr nocapture readonly) {
20; CHECK-LABEL: loadv256f64:
21; CHECK:       # %bb.0:
22; CHECK-NEXT:    lea %s1, 256
23; CHECK-NEXT:    lvl %s1
24; CHECK-NEXT:    vld %v0, 8, %s0
25; CHECK-NEXT:    b.l.t (, %s10)
26  %2 = load <256 x double>, ptr %0, align 16
27  ret <256 x double> %2
28}
29
30; Function Attrs: norecurse nounwind readonly
31define fastcc <256 x i32> @loadv256i32(ptr nocapture readonly) {
32; CHECK-LABEL: loadv256i32:
33; CHECK:       # %bb.0:
34; CHECK-NEXT:    lea %s1, 256
35; CHECK-NEXT:    lvl %s1
36; CHECK-NEXT:    vldl.zx %v0, 4, %s0
37; CHECK-NEXT:    b.l.t (, %s10)
38  %2 = load <256 x i32>, ptr %0, align 16
39  ret <256 x i32> %2
40}
41
42; Function Attrs: norecurse nounwind readonly
43define fastcc <256 x float> @loadv256f32(ptr nocapture readonly) {
44; CHECK-LABEL: loadv256f32:
45; CHECK:       # %bb.0:
46; CHECK-NEXT:    lea %s1, 256
47; CHECK-NEXT:    lvl %s1
48; CHECK-NEXT:    vldu %v0, 4, %s0
49; CHECK-NEXT:    b.l.t (, %s10)
50  %2 = load <256 x float>, ptr %0, align 16
51  ret <256 x float> %2
52}
53
54; Function Attrs: norecurse nounwind readonly
55define fastcc <256 x i64> @loadv256i64stk() {
56; CHECK-LABEL: loadv256i64stk:
57; CHECK:       # %bb.0:
58; CHECK-NEXT:    lea %s11, -2048(, %s11)
59; CHECK-NEXT:    brge.l.t %s11, %s8, .LBB4_2
60; CHECK-NEXT:  # %bb.1:
61; CHECK-NEXT:    ld %s61, 24(, %s14)
62; CHECK-NEXT:    or %s62, 0, %s0
63; CHECK-NEXT:    lea %s63, 315
64; CHECK-NEXT:    shm.l %s63, (%s61)
65; CHECK-NEXT:    shm.l %s8, 8(%s61)
66; CHECK-NEXT:    shm.l %s11, 16(%s61)
67; CHECK-NEXT:    monc
68; CHECK-NEXT:    or %s0, 0, %s62
69; CHECK-NEXT:  .LBB4_2:
70; CHECK-NEXT:    lea %s0, 256
71; CHECK-NEXT:    lea %s1, (, %s11)
72; CHECK-NEXT:    lvl %s0
73; CHECK-NEXT:    vld %v0, 8, %s1
74; CHECK-NEXT:    lea %s11, 2048(, %s11)
75; CHECK-NEXT:    b.l.t (, %s10)
76  %addr = alloca <256 x i64>, align 16
77  %1 = load <256 x i64>, ptr %addr, align 16
78  ret <256 x i64> %1
79}
80
81; Function Attrs: norecurse nounwind readonly
82define fastcc <256 x i64> @loadv256i64com() {
83; CHECK-LABEL: loadv256i64com:
84; CHECK:       # %bb.0:
85; CHECK-NEXT:    lea %s0, v256i64@lo
86; CHECK-NEXT:    and %s0, %s0, (32)0
87; CHECK-NEXT:    lea.sl %s0, v256i64@hi(, %s0)
88; CHECK-NEXT:    lea %s1, 256
89; CHECK-NEXT:    lvl %s1
90; CHECK-NEXT:    vld %v0, 8, %s0
91; CHECK-NEXT:    b.l.t (, %s10)
92  %1 = load <256 x i64>, ptr @v256i64, align 16
93  ret <256 x i64> %1
94}
95