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
93
94
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s
; In the following 4 tests, the existing call to VZU/VZA ensures clean state before
; the call to the unknown, so we don't need to insert a second VZU at that point.
define <4 x float> @zeroupper_v4f32(<8 x float> *%x, <8 x float> %y) nounwind {
; CHECK-LABEL: zeroupper_v4f32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: subq $48, %rsp
; CHECK-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; CHECK-NEXT: movq %rdi, %rbx
; CHECK-NEXT: vzeroupper
; CHECK-NEXT: callq the_unknown
; CHECK-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
; CHECK-NEXT: vaddps (%rbx), %ymm0, %ymm0
; CHECK-NEXT: vextractf128 $1, %ymm0, %xmm1
; CHECK-NEXT: vaddps %xmm1, %xmm0, %xmm0
; CHECK-NEXT: addq $48, %rsp
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: vzeroupper
; CHECK-NEXT: retq
call void @llvm.x86.avx.vzeroupper()
call void @the_unknown()
%loadx = load <8 x float>, <8 x float> *%x, align 32
%sum = fadd <8 x float> %loadx, %y
%lo = shufflevector <8 x float> %sum, <8 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
%hi = shufflevector <8 x float> %sum, <8 x float> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
%res = fadd <4 x float> %lo, %hi
ret <4 x float> %res
}
define <8 x float> @zeroupper_v8f32(<8 x float> %x) nounwind {
; CHECK-LABEL: zeroupper_v8f32:
; CHECK: # %bb.0:
; CHECK-NEXT: subq $56, %rsp
; CHECK-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; CHECK-NEXT: vzeroupper
; CHECK-NEXT: callq the_unknown
; CHECK-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
; CHECK-NEXT: addq $56, %rsp
; CHECK-NEXT: retq
call void @llvm.x86.avx.vzeroupper()
call void @the_unknown()
ret <8 x float> %x
}
define <4 x float> @zeroall_v4f32(<8 x float> *%x, <8 x float> %y) nounwind {
; CHECK-LABEL: zeroall_v4f32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: subq $48, %rsp
; CHECK-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; CHECK-NEXT: movq %rdi, %rbx
; CHECK-NEXT: vzeroall
; CHECK-NEXT: callq the_unknown
; CHECK-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
; CHECK-NEXT: vaddps (%rbx), %ymm0, %ymm0
; CHECK-NEXT: vextractf128 $1, %ymm0, %xmm1
; CHECK-NEXT: vaddps %xmm1, %xmm0, %xmm0
; CHECK-NEXT: addq $48, %rsp
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: vzeroupper
; CHECK-NEXT: retq
call void @llvm.x86.avx.vzeroall()
call void @the_unknown()
%loadx = load <8 x float>, <8 x float> *%x, align 32
%sum = fadd <8 x float> %loadx, %y
%lo = shufflevector <8 x float> %sum, <8 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
%hi = shufflevector <8 x float> %sum, <8 x float> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
%res = fadd <4 x float> %lo, %hi
ret <4 x float> %res
}
define <8 x float> @zeroall_v8f32(<8 x float> %x) nounwind {
; CHECK-LABEL: zeroall_v8f32:
; CHECK: # %bb.0:
; CHECK-NEXT: subq $56, %rsp
; CHECK-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; CHECK-NEXT: vzeroall
; CHECK-NEXT: callq the_unknown
; CHECK-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
; CHECK-NEXT: addq $56, %rsp
; CHECK-NEXT: retq
call void @llvm.x86.avx.vzeroall()
call void @the_unknown()
ret <8 x float> %x
}
declare void @llvm.x86.avx.vzeroupper() nounwind readnone
declare void @llvm.x86.avx.vzeroall() nounwind readnone
declare void @the_unknown() nounwind
|
