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 <256 x float> @llvm.vp.merge.v256f32(<256 x i1>, <256 x float>, <256 x float>, i32)
5 declare <256 x float> @llvm.vp.fma.v256f32(<256 x float>, <256 x float>, <256 x float>, <256 x i1>, i32)
7 define fastcc <256 x float> @test_vp_fma_v256f32_vvv_merge(<256 x float> %i0, <256 x float> %i1, <256 x float> %i2, <256 x i1> %m, i32 %n, <256 x float> %passthru) {
8 ; CHECK-LABEL: test_vp_fma_v256f32_vvv_merge:
10 ; CHECK-NEXT: and %s0, %s0, (32)0
12 ; CHECK-NEXT: vfmad.s %v3, %v2, %v0, %v1, %vm1
13 ; CHECK-NEXT: lea %s16, 256
14 ; CHECK-NEXT: lvl %s16
15 ; CHECK-NEXT: vor %v0, (0)1, %v3
16 ; CHECK-NEXT: b.l.t (, %s10)
17 %vr = call <256 x float> @llvm.vp.fma.v256f32(<256 x float> %i0, <256 x float> %i1, <256 x float> %i2, <256 x i1> %m, i32 %n)
18 %r0 = call <256 x float> @llvm.vp.merge.v256f32(<256 x i1> %m, <256 x float> %vr, <256 x float> %passthru, i32 %n)
22 define fastcc <256 x float> @test_vp_fma_v256f32_rvv_merge(float %s0, <256 x float> %i1, <256 x float> %i2, <256 x i1> %m, i32 %n, <256 x float> %passthru) {
23 ; CHECK-LABEL: test_vp_fma_v256f32_rvv_merge:
25 ; CHECK-NEXT: and %s1, %s1, (32)0
27 ; CHECK-NEXT: vfmad.s %v2, %v1, %s0, %v0, %vm1
28 ; CHECK-NEXT: lea %s16, 256
29 ; CHECK-NEXT: lvl %s16
30 ; CHECK-NEXT: vor %v0, (0)1, %v2
31 ; CHECK-NEXT: b.l.t (, %s10)
32 %xins = insertelement <256 x float> undef, float %s0, i32 0
33 %i0 = shufflevector <256 x float> %xins, <256 x float> undef, <256 x i32> zeroinitializer
34 %vr = call <256 x float> @llvm.vp.fma.v256f32(<256 x float> %i0, <256 x float> %i1, <256 x float> %i2, <256 x i1> %m, i32 %n)
35 %r0 = call <256 x float> @llvm.vp.merge.v256f32(<256 x i1> %m, <256 x float> %vr, <256 x float> %passthru, i32 %n)
39 define fastcc <256 x float> @test_vp_fma_v256f32_vrv_merge(<256 x float> %i0, float %s1, <256 x float> %i2, <256 x i1> %m, i32 %n, <256 x float> %passthru) {
40 ; CHECK-LABEL: test_vp_fma_v256f32_vrv_merge:
42 ; CHECK-NEXT: and %s1, %s1, (32)0
44 ; CHECK-NEXT: vfmad.s %v2, %v1, %s0, %v0, %vm1
45 ; CHECK-NEXT: lea %s16, 256
46 ; CHECK-NEXT: lvl %s16
47 ; CHECK-NEXT: vor %v0, (0)1, %v2
48 ; CHECK-NEXT: b.l.t (, %s10)
49 %yins = insertelement <256 x float> undef, float %s1, i32 0
50 %i1 = shufflevector <256 x float> %yins, <256 x float> undef, <256 x i32> zeroinitializer
51 %vr = call <256 x float> @llvm.vp.fma.v256f32(<256 x float> %i0, <256 x float> %i1, <256 x float> %i2, <256 x i1> %m, i32 %n)
52 %r0 = call <256 x float> @llvm.vp.merge.v256f32(<256 x i1> %m, <256 x float> %vr, <256 x float> %passthru, i32 %n)
56 define fastcc <256 x float> @test_vp_fma_v256f32_vvr_merge(<256 x float> %i0, <256 x float> %i1, float %s2, <256 x i1> %m, i32 %n, <256 x float> %passthru) {
57 ; CHECK-LABEL: test_vp_fma_v256f32_vvr_merge:
59 ; CHECK-NEXT: and %s1, %s1, (32)0
61 ; CHECK-NEXT: vfmad.s %v2, %s0, %v0, %v1, %vm1
62 ; CHECK-NEXT: lea %s16, 256
63 ; CHECK-NEXT: lvl %s16
64 ; CHECK-NEXT: vor %v0, (0)1, %v2
65 ; CHECK-NEXT: b.l.t (, %s10)
66 %zins = insertelement <256 x float> undef, float %s2, i32 0
67 %i2 = shufflevector <256 x float> %zins, <256 x float> undef, <256 x i32> zeroinitializer
68 %vr = call <256 x float> @llvm.vp.fma.v256f32(<256 x float> %i0, <256 x float> %i1, <256 x float> %i2, <256 x i1> %m, i32 %n)
69 %r0 = call <256 x float> @llvm.vp.merge.v256f32(<256 x i1> %m, <256 x float> %vr, <256 x float> %passthru, i32 %n)