1; RUN: llc -march=amdgcn -mcpu=gfx906 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX906
2; RUN: llc -march=amdgcn -mcpu=gfx908 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX908
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
6declare i32 @llvm.amdgcn.sdot2(<2 x i16> %a, <2 x i16> %b, i32 %c, i1 %clamp)
7
8; GCN-LABEL: {{^}}test_llvm_amdgcn_sdot2_clamp
9; GFX906: v_dot2_i32_i16 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} clamp{{$}}
10; GFX908: v_dot2_i32_i16 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} clamp{{$}}
11; GFX10:  v_dot2_i32_i16 v{{[0-9]+}}, s{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}} clamp{{$}}
12define amdgpu_kernel void @test_llvm_amdgcn_sdot2_clamp(
13    i32 addrspace(1)* %r,
14    <2 x i16> addrspace(1)* %a,
15    <2 x i16> addrspace(1)* %b,
16    i32 addrspace(1)* %c) {
17entry:
18  %a.val = load <2 x i16>, <2 x i16> addrspace(1)* %a
19  %b.val = load <2 x i16>, <2 x i16> addrspace(1)* %b
20  %c.val = load i32, i32 addrspace(1)* %c
21  %r.val = call i32 @llvm.amdgcn.sdot2(<2 x i16> %a.val, <2 x i16> %b.val, i32 %c.val, i1 1)
22  store i32 %r.val, i32 addrspace(1)* %r
23  ret void
24}
25
26; GCN-LABEL: {{^}}test_llvm_amdgcn_sdot2_no_clamp
27; GFX906: v_dot2_i32_i16 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}{{$}}
28; GFX908: v_dot2c_i32_i16_e32 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}{{$}}
29; GFX10:  v_dot2_i32_i16 v{{[0-9]+}}, s{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}{{$}}
30define amdgpu_kernel void @test_llvm_amdgcn_sdot2_no_clamp(
31    i32 addrspace(1)* %r,
32    <2 x i16> addrspace(1)* %a,
33    <2 x i16> addrspace(1)* %b,
34    i32 addrspace(1)* %c) {
35entry:
36  %a.val = load <2 x i16>, <2 x i16> addrspace(1)* %a
37  %b.val = load <2 x i16>, <2 x i16> addrspace(1)* %b
38  %c.val = load i32, i32 addrspace(1)* %c
39  %r.val = call i32 @llvm.amdgcn.sdot2(<2 x i16> %a.val, <2 x i16> %b.val, i32 %c.val, i1 0)
40  store i32 %r.val, i32 addrspace(1)* %r
41  ret void
42}
43