summaryrefslogtreecommitdiff
path: root/test/CodeGen/AMDGPU/llvm.amdgcn.cvt.pk.u16.ll
blob: 8d5c9aa952196054a1b859144be75794f410a8d0 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=fiji -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=VI %s
; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=VI %s

; GCN-LABEL: {{^}}s_cvt_pk_u16_u32:
; GCN-DAG: s_load_dword [[X:s[0-9]+]], s[0:1], 0x{{b|2c}}
; GCN-DAG: s_load_dword [[SY:s[0-9]+]], s[0:1], 0x{{c|30}}
; GCN: v_mov_b32_e32 [[VY:v[0-9]+]], [[SY]]
; SI: v_cvt_pk_u16_u32_e32 v{{[0-9]+}}, [[X]], [[VY]]
; VI: v_cvt_pk_u16_u32 v{{[0-9]+}}, [[X]], [[VY]]
define amdgpu_kernel void @s_cvt_pk_u16_u32(i32 addrspace(1)* %out, i32 %x, i32 %y) #0 {
  %result = call <2 x i16> @llvm.amdgcn.cvt.pk.u16(i32 %x, i32 %y)
  %r = bitcast <2 x i16> %result to i32
  store i32 %r, i32 addrspace(1)* %out
  ret void
}

; GCN-LABEL: {{^}}s_cvt_pk_u16_samereg_i32:
; GCN: s_load_dword [[X:s[0-9]+]]
; GCN: v_cvt_pk_u16_u32{{(_e64)*}} v{{[0-9]+}}, [[X]], [[X]]
define amdgpu_kernel void @s_cvt_pk_u16_samereg_i32(i32 addrspace(1)* %out, i32 %x) #0 {
  %result = call <2 x i16> @llvm.amdgcn.cvt.pk.u16(i32 %x, i32 %x)
  %r = bitcast <2 x i16> %result to i32
  store i32 %r, i32 addrspace(1)* %out
  ret void
}

; GCN-LABEL: {{^}}v_cvt_pk_u16_u32:
; GCN: {{buffer|flat|global}}_load_dword [[A:v[0-9]+]]
; GCN: {{buffer|flat|global}}_load_dword [[B:v[0-9]+]]
; SI: v_cvt_pk_u16_u32_e32 v{{[0-9]+}}, [[A]], [[B]]
; VI: v_cvt_pk_u16_u32 v{{[0-9]+}}, [[A]], [[B]]
define amdgpu_kernel void @v_cvt_pk_u16_u32(i32 addrspace(1)* %out, i32 addrspace(1)* %a.ptr, i32 addrspace(1)* %b.ptr) #0 {
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %tid.ext = sext i32 %tid to i64
  %a.gep = getelementptr inbounds i32, i32 addrspace(1)* %a.ptr, i64 %tid.ext
  %b.gep = getelementptr inbounds i32, i32 addrspace(1)* %b.ptr, i64 %tid.ext
  %out.gep = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 %tid.ext
  %a = load volatile i32, i32 addrspace(1)* %a.gep
  %b = load volatile i32, i32 addrspace(1)* %b.gep
  %cvt = call <2 x i16> @llvm.amdgcn.cvt.pk.u16(i32 %a, i32 %b)
  %r = bitcast <2 x i16> %cvt to i32
  store i32 %r, i32 addrspace(1)* %out.gep
  ret void
}

; GCN-LABEL: {{^}}v_cvt_pk_u16_u32_reg_imm:
; GCN: {{buffer|flat|global}}_load_dword [[A:v[0-9]+]]
; GCN: v_cvt_pk_u16_u32{{(_e64)*}} v{{[0-9]+}}, [[A]], 1
define amdgpu_kernel void @v_cvt_pk_u16_u32_reg_imm(i32 addrspace(1)* %out, i32 addrspace(1)* %a.ptr) #0 {
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %tid.ext = sext i32 %tid to i64
  %a.gep = getelementptr inbounds i32, i32 addrspace(1)* %a.ptr, i64 %tid.ext
  %out.gep = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 %tid.ext
  %a = load volatile i32, i32 addrspace(1)* %a.gep
  %cvt = call <2 x i16> @llvm.amdgcn.cvt.pk.u16(i32 %a, i32 1)
  %r = bitcast <2 x i16> %cvt to i32
  store i32 %r, i32 addrspace(1)* %out.gep
  ret void
}

; GCN-LABEL: {{^}}v_cvt_pk_u16_u32_imm_reg:
; GCN: {{buffer|flat|global}}_load_dword [[A:v[0-9]+]]
; SI: v_cvt_pk_u16_u32_e32 v{{[0-9]+}}, 1, [[A]]
; VI: v_cvt_pk_u16_u32 v{{[0-9]+}}, 1, [[A]]
define amdgpu_kernel void @v_cvt_pk_u16_u32_imm_reg(i32 addrspace(1)* %out, i32 addrspace(1)* %a.ptr) #0 {
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %tid.ext = sext i32 %tid to i64
  %a.gep = getelementptr inbounds i32, i32 addrspace(1)* %a.ptr, i64 %tid.ext
  %out.gep = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 %tid.ext
  %a = load volatile i32, i32 addrspace(1)* %a.gep
  %cvt = call <2 x i16> @llvm.amdgcn.cvt.pk.u16(i32 1, i32 %a)
  %r = bitcast <2 x i16> %cvt to i32
  store i32 %r, i32 addrspace(1)* %out.gep
  ret void
}

declare <2 x i16> @llvm.amdgcn.cvt.pk.u16(i32, i32) #1
declare i32 @llvm.amdgcn.workitem.id.x() #1


attributes #0 = { nounwind }
attributes #1 = { nounwind readnone }