1 ; RUN: opt -S -dxil-op-lower < %s | FileCheck %s
3 ; Make sure dxil operation function calls for lerp are generated for float and half.
5 ; CHECK-LABEL: lerp_half
6 ; CHECK: fsub half %{{.*}}, %{{.*}}
7 ; CHECK: fmul half %{{.*}}, %{{.*}}
8 ; CHECK: fadd half %{{.*}}, %{{.*}}
9 define noundef half @lerp_half(half noundef %p0) {
11 %p0.addr = alloca half, align 2
12 store half %p0, ptr %p0.addr, align 2
13 %0 = load half, ptr %p0.addr, align 2
14 %1 = load half, ptr %p0.addr, align 2
15 %2 = load half, ptr %p0.addr, align 2
16 %dx.lerp = call half @llvm.dx.lerp.f16(half %0, half %1, half %2)
20 ; CHECK-LABEL: lerp_float
21 ; CHECK: fsub float %{{.*}}, %{{.*}}
22 ; CHECK: fmul float %{{.*}}, %{{.*}}
23 ; CHECK: fadd float %{{.*}}, %{{.*}}
24 define noundef float @lerp_float(float noundef %p0, float noundef %p1) {
26 %p1.addr = alloca float, align 4
27 %p0.addr = alloca float, align 4
28 store float %p1, ptr %p1.addr, align 4
29 store float %p0, ptr %p0.addr, align 4
30 %0 = load float, ptr %p0.addr, align 4
31 %1 = load float, ptr %p0.addr, align 4
32 %2 = load float, ptr %p0.addr, align 4
33 %dx.lerp = call float @llvm.dx.lerp.f32(float %0, float %1, float %2)
37 ; CHECK-LABEL: lerp_float4
38 ; CHECK: fsub <4 x float> %{{.*}}, %{{.*}}
39 ; CHECK: fmul <4 x float> %{{.*}}, %{{.*}}
40 ; CHECK: fadd <4 x float> %{{.*}}, %{{.*}}
41 define noundef <4 x float> @lerp_float4(<4 x float> noundef %p0, <4 x float> noundef %p1) {
43 %p1.addr = alloca <4 x float>, align 16
44 %p0.addr = alloca <4 x float>, align 16
45 store <4 x float> %p1, ptr %p1.addr, align 16
46 store <4 x float> %p0, ptr %p0.addr, align 16
47 %0 = load <4 x float>, ptr %p0.addr, align 16
48 %1 = load <4 x float>, ptr %p0.addr, align 16
49 %2 = load <4 x float>, ptr %p0.addr, align 16
50 %dx.lerp = call <4 x float> @llvm.dx.lerp.v4f32(<4 x float> %0, <4 x float> %1, <4 x float> %2)
51 ret <4 x float> %dx.lerp
54 declare half @llvm.dx.lerp.f16(half, half, half)
55 declare float @llvm.dx.lerp.f32(float, float, float)
56 declare <4 x float> @llvm.dx.lerp.v4f32(<4 x float>, <4 x float>, <4 x float>)