1 ; RUN: llc -mtriple=amdgcn -mcpu=gfx906 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX9
2 ; RUN: llc -mtriple=amdgcn -mcpu=gfx940 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX9
3 ; RUN: llc -mtriple=amdgcn -mcpu=gfx1011 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX10
4 ; RUN: llc -mtriple=amdgcn -mcpu=gfx1012 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX10
5 ; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX10
7 declare i32 @llvm.amdgcn.udot8(i32 %a, i32 %b, i32 %c, i1 %clamp)
9 ; GCN-LABEL: {{^}}test_llvm_amdgcn_udot8_clamp
10 ; GFX9: v_dot8_u32_u4 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} clamp{{$}}
11 ; GFX10: v_dot8_u32_u4 v{{[0-9]+}}, s{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}} clamp{{$}}
12 define amdgpu_kernel void @test_llvm_amdgcn_udot8_clamp(
16 ptr addrspace(1) %c) {
18 %a.val = load <8 x i4>, ptr addrspace(1) %a
19 %b.val = load <8 x i4>, ptr addrspace(1) %b
20 %a.val.cast = bitcast <8 x i4> %a.val to i32
21 %b.val.cast = bitcast <8 x i4> %b.val to i32
22 %c.val = load i32, ptr addrspace(1) %c
23 %r.val = call i32 @llvm.amdgcn.udot8(i32 %a.val.cast, i32 %b.val.cast, i32 %c.val, i1 1)
24 store i32 %r.val, ptr addrspace(1) %r
28 ; GCN-LABEL: {{^}}test_llvm_amdgcn_udot8_no_clamp
29 ; GFX9: v_dot8_u32_u4 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}{{$}}
30 ; GFX10: v_dot8_u32_u4 v{{[0-9]+}}, s{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}{{$}}
31 define amdgpu_kernel void @test_llvm_amdgcn_udot8_no_clamp(
35 ptr addrspace(1) %c) {
37 %a.val = load <8 x i4>, ptr addrspace(1) %a
38 %b.val = load <8 x i4>, ptr addrspace(1) %b
39 %a.val.cast = bitcast <8 x i4> %a.val to i32
40 %b.val.cast = bitcast <8 x i4> %b.val to i32
41 %c.val = load i32, ptr addrspace(1) %c
42 %r.val = call i32 @llvm.amdgcn.udot8(i32 %a.val.cast, i32 %b.val.cast, i32 %c.val, i1 0)
43 store i32 %r.val, ptr addrspace(1) %r