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
5 ; RUN: llc -march=amdgcn -mcpu=gfx1100 -amdgpu-enable-vopd=0 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX10
7 declare float @llvm.amdgcn.fdot2(<2 x half> %a, <2 x half> %b, float %c, i1 %clamp)
9 ; GCN-LABEL: {{^}}test_llvm_amdgcn_fdot2_clamp
10 ; GFX9: v_dot2_f32_f16 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} clamp{{$}}
11 ; GFX10: v_dot2_f32_f16 v{{[0-9]+}}, s{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}} clamp{{$}}
12 define amdgpu_kernel void @test_llvm_amdgcn_fdot2_clamp(
16 ptr addrspace(1) %c) {
18 %a.val = load <2 x half>, ptr addrspace(1) %a
19 %b.val = load <2 x half>, ptr addrspace(1) %b
20 %c.val = load float, ptr addrspace(1) %c
21 %r.val = call float @llvm.amdgcn.fdot2(<2 x half> %a.val, <2 x half> %b.val, float %c.val, i1 1)
22 store float %r.val, ptr addrspace(1) %r
26 ; GCN-LABEL: {{^}}test_llvm_amdgcn_fdot2_no_clamp
27 ; GFX906: v_dot2_f32_f16 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}{{$}}
28 ; GFX940: v_dot2c_f32_f16_e32 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}{{$}}
29 ; GFX10: {{v_dot2c_f32_f16_e32|v_dot2acc_f32_f16}} v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}{{$}}
30 define amdgpu_kernel void @test_llvm_amdgcn_fdot2_no_clamp(
34 ptr addrspace(1) %c) {
36 %a.val = load <2 x half>, ptr addrspace(1) %a
37 %b.val = load <2 x half>, ptr addrspace(1) %b
38 %c.val = load float, ptr addrspace(1) %c
39 %r.val = call float @llvm.amdgcn.fdot2(<2 x half> %a.val, <2 x half> %b.val, float %c.val, i1 0)
40 store float %r.val, ptr addrspace(1) %r
44 ; GFX9-LABEL: {{^}}fdot2_inline_literal
45 ; GFX906: v_dot2_f32_f16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}, 1.0
46 ; GFX940: v_dot2c_f32_f16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}{{$}}
47 define float @fdot2_inline_literal(<2 x half> %a, <2 x half> %b) {
48 %ret = tail call float @llvm.amdgcn.fdot2(<2 x half> %a, <2 x half> %b, float 1.0, i1 false)