1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s -march=ve -mattr=+vpu | FileCheck %s
4 declare <512 x float> @llvm.vp.fdiv.v512f32(<512 x float>, <512 x float>, <512 x i1>, i32)
6 define fastcc <512 x float> @test_vp_fdiv_v512f32_vv(<512 x float> %i0, <512 x float> %i1, <512 x i1> %m, i32 %n) {
7 ; CHECK-LABEL: test_vp_fdiv_v512f32_vv:
9 ; CHECK-NEXT: and %s1, %s0, (32)0
10 ; CHECK-NEXT: srl %s1, %s1, 1
12 ; CHECK-NEXT: vshf %v2, %v1, %v1, 4
13 ; CHECK-NEXT: vshf %v3, %v0, %v0, 4
14 ; CHECK-NEXT: vfdiv.s %v2, %v3, %v2, %vm3
15 ; CHECK-NEXT: adds.w.sx %s0, 1, %s0
16 ; CHECK-NEXT: and %s0, %s0, (32)0
17 ; CHECK-NEXT: srl %s0, %s0, 1
19 ; CHECK-NEXT: vfdiv.s %v0, %v0, %v1, %vm2
20 ; CHECK-NEXT: vshf %v0, %v2, %v0, 8
21 ; CHECK-NEXT: b.l.t (, %s10)
22 %r0 = call <512 x float> @llvm.vp.fdiv.v512f32(<512 x float> %i0, <512 x float> %i1, <512 x i1> %m, i32 %n)
26 define fastcc <512 x float> @test_vp_fdiv_v512f32_rv(float %s0, <512 x float> %i1, <512 x i1> %m, i32 %n) {
27 ; CHECK-LABEL: test_vp_fdiv_v512f32_rv:
29 ; CHECK-NEXT: and %s2, %s0, (32)1
30 ; CHECK-NEXT: srl %s0, %s0, 32
31 ; CHECK-NEXT: or %s0, %s0, %s2
32 ; CHECK-NEXT: lea %s2, 256
34 ; CHECK-NEXT: vbrd %v1, %s0
35 ; CHECK-NEXT: adds.w.sx %s0, 1, %s1
36 ; CHECK-NEXT: and %s0, %s0, (32)0
37 ; CHECK-NEXT: srl %s0, %s0, 1
39 ; CHECK-NEXT: vfdiv.s %v2, %v1, %v0, %vm2
40 ; CHECK-NEXT: and %s1, %s1, (32)0
41 ; CHECK-NEXT: srl %s1, %s1, 1
43 ; CHECK-NEXT: vshf %v1, %v1, %v1, 4
44 ; CHECK-NEXT: vshf %v0, %v0, %v0, 4
45 ; CHECK-NEXT: vfdiv.s %v0, %v1, %v0, %vm3
47 ; CHECK-NEXT: vshf %v0, %v0, %v2, 8
48 ; CHECK-NEXT: b.l.t (, %s10)
49 %xins = insertelement <512 x float> undef, float %s0, i32 0
50 %i0 = shufflevector <512 x float> %xins, <512 x float> undef, <512 x i32> zeroinitializer
51 %r0 = call <512 x float> @llvm.vp.fdiv.v512f32(<512 x float> %i0, <512 x float> %i1, <512 x i1> %m, i32 %n)
55 define fastcc <512 x float> @test_vp_fdiv_v512f32_vr(<512 x float> %i0, float %s1, <512 x i1> %m, i32 %n) {
56 ; CHECK-LABEL: test_vp_fdiv_v512f32_vr:
58 ; CHECK-NEXT: and %s2, %s0, (32)1
59 ; CHECK-NEXT: srl %s0, %s0, 32
60 ; CHECK-NEXT: or %s0, %s0, %s2
61 ; CHECK-NEXT: lea %s2, 256
63 ; CHECK-NEXT: vbrd %v1, %s0
64 ; CHECK-NEXT: adds.w.sx %s0, 1, %s1
65 ; CHECK-NEXT: and %s0, %s0, (32)0
66 ; CHECK-NEXT: srl %s0, %s0, 1
68 ; CHECK-NEXT: vfdiv.s %v2, %v0, %v1, %vm2
69 ; CHECK-NEXT: and %s1, %s1, (32)0
70 ; CHECK-NEXT: srl %s1, %s1, 1
72 ; CHECK-NEXT: vshf %v1, %v1, %v1, 4
73 ; CHECK-NEXT: vshf %v0, %v0, %v0, 4
74 ; CHECK-NEXT: vfdiv.s %v0, %v0, %v1, %vm3
76 ; CHECK-NEXT: vshf %v0, %v0, %v2, 8
77 ; CHECK-NEXT: b.l.t (, %s10)
78 %yins = insertelement <512 x float> undef, float %s1, i32 0
79 %i1 = shufflevector <512 x float> %yins, <512 x float> undef, <512 x i32> zeroinitializer
80 %r0 = call <512 x float> @llvm.vp.fdiv.v512f32(<512 x float> %i0, <512 x float> %i1, <512 x i1> %m, i32 %n)