File: amx_tile_pair_O2_to_O0.ll

package info (click to toggle)
llvm-toolchain-20 1%3A20.1.6-1~exp1
  • links: PTS, VCS
  • area: main
  • in suites: experimental
  • size: 2,111,304 kB
  • sloc: cpp: 7,438,677; ansic: 1,393,822; asm: 1,012,926; python: 241,650; f90: 86,635; objc: 75,479; lisp: 42,144; pascal: 17,286; sh: 10,027; ml: 5,082; perl: 4,730; awk: 3,523; makefile: 3,349; javascript: 2,251; xml: 892; fortran: 672
file content (136 lines) | stat: -rw-r--r-- 6,208 bytes parent folder | download | duplicates (6)
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
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -O0 -mtriple=x86_64-unknown-unknown -mattr=+amx-tile,+amx-bf16,+avx512f, \
; RUN: -mattr=+amx-transpose -verify-machineinstrs | FileCheck %s

@buf = dso_local global [2048 x i8] zeroinitializer, align 16
@buf2 = dso_local global [2048 x i8] zeroinitializer, align 16

define dso_local void @test_tile_2rpntlvwz0(i16 noundef signext %row, i16 noundef signext %col0, i16 noundef signext %col1) local_unnamed_addr #0 {
; CHECK-LABEL: test_tile_2rpntlvwz0:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    .cfi_offset %rbp, -16
; CHECK-NEXT:    movq %rsp, %rbp
; CHECK-NEXT:    .cfi_def_cfa_register %rbp
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    andq $-1024, %rsp # imm = 0xFC00
; CHECK-NEXT:    subq $8192, %rsp # imm = 0x2000
; CHECK-NEXT:    .cfi_offset %rbx, -24
; CHECK-NEXT:    vxorps %xmm0, %xmm0, %xmm0
; CHECK-NEXT:    vmovups %zmm0, {{[0-9]+}}(%rsp)
; CHECK-NEXT:    movb $1, {{[0-9]+}}(%rsp)
; CHECK-NEXT:    # kill: def $dx killed $dx killed $edx
; CHECK-NEXT:    movw %si, %cx
; CHECK-NEXT:    movw %di, %ax
; CHECK-NEXT:    # implicit-def: $al
; CHECK-NEXT:    movb %al, {{[0-9]+}}(%rsp)
; CHECK-NEXT:    movw %dx, {{[0-9]+}}(%rsp)
; CHECK-NEXT:    # implicit-def: $al
; CHECK-NEXT:    movb %al, {{[0-9]+}}(%rsp)
; CHECK-NEXT:    movw %dx, {{[0-9]+}}(%rsp)
; CHECK-NEXT:    # implicit-def: $al
; CHECK-NEXT:    movb %al, {{[0-9]+}}(%rsp)
; CHECK-NEXT:    movw %cx, {{[0-9]+}}(%rsp)
; CHECK-NEXT:    # implicit-def: $cl
; CHECK-NEXT:    movb %cl, {{[0-9]+}}(%rsp)
; CHECK-NEXT:    movw %dx, {{[0-9]+}}(%rsp)
; CHECK-NEXT:    # implicit-def: $al
; CHECK-NEXT:    movb %al, {{[0-9]+}}(%rsp)
; CHECK-NEXT:    movw %cx, {{[0-9]+}}(%rsp)
; CHECK-NEXT:    # implicit-def: $al
; CHECK-NEXT:    movb %al, {{[0-9]+}}(%rsp)
; CHECK-NEXT:    movw %cx, {{[0-9]+}}(%rsp)
; CHECK-NEXT:    # implicit-def: $al
; CHECK-NEXT:    movb %al, {{[0-9]+}}(%rsp)
; CHECK-NEXT:    movw %cx, {{[0-9]+}}(%rsp)
; CHECK-NEXT:    # implicit-def: $al
; CHECK-NEXT:    movb %al, {{[0-9]+}}(%rsp)
; CHECK-NEXT:    movw %dx, {{[0-9]+}}(%rsp)
; CHECK-NEXT:    ldtilecfg {{[0-9]+}}(%rsp)
; CHECK-NEXT:    movl $buf, %esi
; CHECK-NEXT:    movl $32, %edi
; CHECK-NEXT:    t2rpntlvwz0 (%rsi,%rdi), %tmm4
; CHECK-NEXT:    movabsq $64, %rbx
; CHECK-NEXT:    tilestored %tmm5, (%rsp,%rbx) # 1024-byte Folded Spill
; CHECK-NEXT:    tileloadd (%rsp,%rbx), %tmm0 # 1024-byte Folded Reload
; CHECK-NEXT:    movabsq $64, %rbx
; CHECK-NEXT:    tilestored %tmm4, 1024(%rsp,%rbx) # 1024-byte Folded Spill
; CHECK-NEXT:    tileloadd 1024(%rsp,%rbx), %tmm1 # 1024-byte Folded Reload
; CHECK-NEXT:    movl $64, %edi
; CHECK-NEXT:    leaq {{[0-9]+}}(%rsp), %rsi
; CHECK-NEXT:    tilestored %tmm1, (%rsi,%rdi)
; CHECK-NEXT:    movl $64, %edi
; CHECK-NEXT:    leaq {{[0-9]+}}(%rsp), %rsi
; CHECK-NEXT:    tilestored %tmm0, (%rsi,%rdi)
; CHECK-NEXT:    tilezero %tmm0
; CHECK-NEXT:    movl $64, %edi
; CHECK-NEXT:    leaq {{[0-9]+}}(%rsp), %rsi
; CHECK-NEXT:    tilestored %tmm0, (%rsi,%rdi)
; CHECK-NEXT:    movl $64, %edi
; CHECK-NEXT:    leaq {{[0-9]+}}(%rsp), %rsi
; CHECK-NEXT:    tileloadd (%rsi,%rdi), %tmm1
; CHECK-NEXT:    movl $64, %edi
; CHECK-NEXT:    leaq {{[0-9]+}}(%rsp), %rsi
; CHECK-NEXT:    tileloadd (%rsi,%rdi), %tmm2
; CHECK-NEXT:    movl $64, %edi
; CHECK-NEXT:    leaq {{[0-9]+}}(%rsp), %rsi
; CHECK-NEXT:    tileloadd (%rsi,%rdi), %tmm0
; CHECK-NEXT:    tdpbssd %tmm2, %tmm1, %tmm0
; CHECK-NEXT:    movl $64, %edi
; CHECK-NEXT:    leaq {{[0-9]+}}(%rsp), %rsi
; CHECK-NEXT:    tilestored %tmm0, (%rsi,%rdi)
; CHECK-NEXT:    movl $64, %edi
; CHECK-NEXT:    leaq {{[0-9]+}}(%rsp), %rsi
; CHECK-NEXT:    tileloadd (%rsi,%rdi), %tmm0
; CHECK-NEXT:    movl $buf2, %edx
; CHECK-NEXT:    movl $32, %esi
; CHECK-NEXT:    tilestored %tmm0, (%rdx,%rsi)
; CHECK-NEXT:    leaq -8(%rbp), %rsp
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    popq %rbp
; CHECK-NEXT:    .cfi_def_cfa %rsp, 8
; CHECK-NEXT:    tilerelease
; CHECK-NEXT:    vzeroupper
; CHECK-NEXT:    retq
entry:
  %0 = tail call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0.internal(i16 %row, i16 %col0, i16 %col1, ptr @buf, i64 32) #3
  %1 = extractvalue { x86_amx, x86_amx } %0, 0
  %2 = tail call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx %1) #3
  %3 = extractvalue { x86_amx, x86_amx } %0, 1
  %4 = tail call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx %3) #3
  %5 = tail call x86_amx @llvm.x86.tilezero.internal(i16 %row, i16 %col0) #3
  %6 = tail call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx %5) #3
  %7 = tail call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> %6) #3
  %8 = tail call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> %2) #3
  %9 = tail call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> %4) #3
  %10 = tail call x86_amx @llvm.x86.tdpbssd.internal(i16 %row, i16 %col1, i16 %col0, x86_amx %7, x86_amx %8, x86_amx %9) #3
  %11 = tail call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx %10) #3
  %12 = tail call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> %11) #3
  tail call void @llvm.x86.tilestored64.internal(i16 %row, i16 %col0, ptr @buf2, i64 32, x86_amx %12) #3
  ret void
}

declare { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0.internal(i16, i16, i16, ptr, i64) #1

declare <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx) #2

declare x86_amx @llvm.x86.tilezero.internal(i16, i16) #3

declare x86_amx @llvm.x86.tdpbssd.internal(i16, i16, i16, x86_amx, x86_amx, x86_amx) #3

declare x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32>) #2

declare void @llvm.x86.tilestored64.internal(i16, i16, ptr, i64, x86_amx) #4

attributes #0 = { nounwind uwtable "target-cpu"="x86-64" "target-features"="+amx-bf16,+amx-int8,+amx-tile,+amx-transpose" }
attributes #1 = { argmemonly nofree nounwind readonly }
attributes #2 = { nofree nosync nounwind readnone }
attributes #3 = { nounwind }
attributes #4 = { argmemonly nounwind writeonly }

!llvm.module.flags = !{!0, !1, !2}

!0 = !{i32 1, !"wchar_size", i32 4}
!1 = !{i32 7, !"uwtable", i32 2}
!2 = !{i32 7, !"frame-pointer", i32 2}