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
85
86
87
88
89
90
91
92
|
; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tahiti -verify-machineinstrs< %s | FileCheck -check-prefixes=GCN,SI,FUNC %s
; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tonga -verify-machineinstrs< %s | FileCheck -check-prefixes=GCN,VI,FUNC %s
; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx900 -verify-machineinstrs< %s | FileCheck -check-prefixes=GCN,GFX9,FUNC %s
declare { i32, i1 } @llvm.ssub.with.overflow.i32(i32, i32) nounwind readnone
declare { i64, i1 } @llvm.ssub.with.overflow.i64(i64, i64) nounwind readnone
declare { <2 x i32>, <2 x i1> } @llvm.ssub.with.overflow.v2i32(<2 x i32>, <2 x i32>) nounwind readnone
; FUNC-LABEL: {{^}}ssubo_i64_zext:
define amdgpu_kernel void @ssubo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 %b) nounwind {
%ssub = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 %a, i64 %b) nounwind
%val = extractvalue { i64, i1 } %ssub, 0
%carry = extractvalue { i64, i1 } %ssub, 1
%ext = zext i1 %carry to i64
%add2 = add i64 %val, %ext
store i64 %add2, ptr addrspace(1) %out, align 8
ret void
}
; FUNC-LABEL: {{^}}s_ssubo_i32:
define amdgpu_kernel void @s_ssubo_i32(ptr addrspace(1) %out, ptr addrspace(1) %carryout, i32 %a, i32 %b) nounwind {
%ssub = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 %a, i32 %b) nounwind
%val = extractvalue { i32, i1 } %ssub, 0
%carry = extractvalue { i32, i1 } %ssub, 1
store i32 %val, ptr addrspace(1) %out, align 4
store i1 %carry, ptr addrspace(1) %carryout
ret void
}
; FUNC-LABEL: {{^}}v_ssubo_i32:
define amdgpu_kernel void @v_ssubo_i32(ptr addrspace(1) %out, ptr addrspace(1) %carryout, ptr addrspace(1) %aptr, ptr addrspace(1) %bptr) nounwind {
%a = load i32, ptr addrspace(1) %aptr, align 4
%b = load i32, ptr addrspace(1) %bptr, align 4
%ssub = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 %a, i32 %b) nounwind
%val = extractvalue { i32, i1 } %ssub, 0
%carry = extractvalue { i32, i1 } %ssub, 1
store i32 %val, ptr addrspace(1) %out, align 4
store i1 %carry, ptr addrspace(1) %carryout
ret void
}
; FUNC-LABEL: {{^}}s_ssubo_i64:
; GCN: s_sub_u32
; GCN: s_subb_u32
define amdgpu_kernel void @s_ssubo_i64(ptr addrspace(1) %out, ptr addrspace(1) %carryout, i64 %a, i64 %b) nounwind {
%ssub = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 %a, i64 %b) nounwind
%val = extractvalue { i64, i1 } %ssub, 0
%carry = extractvalue { i64, i1 } %ssub, 1
store i64 %val, ptr addrspace(1) %out, align 8
store i1 %carry, ptr addrspace(1) %carryout
ret void
}
; FUNC-LABEL: {{^}}v_ssubo_i64:
; SI: v_sub_i32_e32 v{{[0-9]+}}, vcc,
; SI: v_subb_u32_e32 v{{[0-9]+}}, vcc,
; VI: v_sub_u32_e32 v{{[0-9]+}}, vcc,
; VI: v_subb_u32_e32 v{{[0-9]+}}, vcc,
; GFX9: v_sub_co_u32_e32 v{{[0-9]+}}, vcc,
; GFX9: v_subb_co_u32_e32 v{{[0-9]+}}, vcc,
define amdgpu_kernel void @v_ssubo_i64(ptr addrspace(1) %out, ptr addrspace(1) %carryout, ptr addrspace(1) %aptr, ptr addrspace(1) %bptr) nounwind {
%a = load i64, ptr addrspace(1) %aptr, align 4
%b = load i64, ptr addrspace(1) %bptr, align 4
%ssub = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 %a, i64 %b) nounwind
%val = extractvalue { i64, i1 } %ssub, 0
%carry = extractvalue { i64, i1 } %ssub, 1
store i64 %val, ptr addrspace(1) %out, align 8
store i1 %carry, ptr addrspace(1) %carryout
ret void
}
; FUNC-LABEL: {{^}}v_ssubo_v2i32:
; SICIVI: v_cmp_lt_i32
; SICIVI: v_cmp_lt_i32
; SICIVI: v_sub_{{[iu]}}32
; SICIVI: v_cmp_lt_i32
; SICIVI: v_cmp_lt_i32
; SICIVI: v_sub_{{[iu]}}32
define amdgpu_kernel void @v_ssubo_v2i32(ptr addrspace(1) %out, ptr addrspace(1) %carryout, ptr addrspace(1) %aptr, ptr addrspace(1) %bptr) nounwind {
%a = load <2 x i32>, ptr addrspace(1) %aptr, align 4
%b = load <2 x i32>, ptr addrspace(1) %bptr, align 4
%sadd = call { <2 x i32>, <2 x i1> } @llvm.ssub.with.overflow.v2i32(<2 x i32> %a, <2 x i32> %b) nounwind
%val = extractvalue { <2 x i32>, <2 x i1> } %sadd, 0
%carry = extractvalue { <2 x i32>, <2 x i1> } %sadd, 1
store <2 x i32> %val, ptr addrspace(1) %out, align 4
%carry.ext = zext <2 x i1> %carry to <2 x i32>
store <2 x i32> %carry.ext, ptr addrspace(1) %carryout
ret void
}
|