1 ; RUN: llc -march=amdgcn -mcpu=gfx906 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX9,GFX906
2 ; RUN: llc -march=amdgcn -mcpu=gfx940 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX9,GFX940
3 ; RUN: llc -march=amdgcn -mcpu=gfx1011 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX10
4 ; RUN: llc -march=amdgcn -mcpu=gfx1012 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX10
6 declare float @llvm.amdgcn.fdot2(<2 x half> %a, <2 x half> %b, float %c, i1 %clamp)
8 ; GCN-LABEL: {{^}}test_llvm_amdgcn_fdot2_clamp
9 ; GFX9: v_dot2_f32_f16 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} clamp{{$}}
10 ; GFX10: v_dot2_f32_f16 v{{[0-9]+}}, s{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}} clamp{{$}}
11 define amdgpu_kernel void @test_llvm_amdgcn_fdot2_clamp(
12 float addrspace(1)* %r,
13 <2 x half> addrspace(1)* %a,
14 <2 x half> addrspace(1)* %b,
15 float addrspace(1)* %c) {
17 %a.val = load <2 x half>, <2 x half> addrspace(1)* %a
18 %b.val = load <2 x half>, <2 x half> addrspace(1)* %b
19 %c.val = load float, float addrspace(1)* %c
20 %r.val = call float @llvm.amdgcn.fdot2(<2 x half> %a.val, <2 x half> %b.val, float %c.val, i1 1)
21 store float %r.val, float addrspace(1)* %r
25 ; GCN-LABEL: {{^}}test_llvm_amdgcn_fdot2_no_clamp
26 ; GFX906: v_dot2_f32_f16 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}{{$}}
27 ; GFX940: v_dot2c_f32_f16_e32 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}{{$}}
28 ; GFX10: v_dot2c_f32_f16_e32 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}{{$}}
29 define amdgpu_kernel void @test_llvm_amdgcn_fdot2_no_clamp(
30 float addrspace(1)* %r,
31 <2 x half> addrspace(1)* %a,
32 <2 x half> addrspace(1)* %b,
33 float addrspace(1)* %c) {
35 %a.val = load <2 x half>, <2 x half> addrspace(1)* %a
36 %b.val = load <2 x half>, <2 x half> addrspace(1)* %b
37 %c.val = load float, float addrspace(1)* %c
38 %r.val = call float @llvm.amdgcn.fdot2(<2 x half> %a.val, <2 x half> %b.val, float %c.val, i1 0)
39 store float %r.val, float addrspace(1)* %r
43 ; GFX9-LABEL: {{^}}fdot2_inline_literal
44 ; GFX906: v_dot2_f32_f16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}, 1.0
45 ; GFX940: v_dot2c_f32_f16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}{{$}}
46 define float @fdot2_inline_literal(<2 x half> %a, <2 x half> %b) {
47 %ret = tail call float @llvm.amdgcn.fdot2(<2 x half> %a, <2 x half> %b, float 1.0, i1 false)