docs/experiments/matmul/matmul.normalopt.s

   1         .file   "matmul.normalopt.ll"
   2         .section        .rodata.cst8,"aM",@progbits,8
   3         .align  8
   4 .LCPI0_0:
   5         .quad   4602678819172646912     # double 0.5
   6         .text
   7         .globl  init_array
   8         .align  16, 0x90
   9         .type   init_array,@function
  10 init_array:                             # @init_array
  11         .cfi_startproc
  12 # BB#0:                                 # %entry
  13         pushq   %rbp
  14 .Ltmp2:
  15         .cfi_def_cfa_offset 16
  16 .Ltmp3:
  17         .cfi_offset %rbp, -16
  18         movq    %rsp, %rbp
  19 .Ltmp4:
  20         .cfi_def_cfa_register %rbp
  21         xorl    %r8d, %r8d
  22         vmovsd  .LCPI0_0(%rip), %xmm0
  23         .align  16, 0x90
  24 .LBB0_1:                                # %for.cond1.preheader
  25                                         # =>This Loop Header: Depth=1
  26                                         #     Child Loop BB0_2 Depth 2
  27         xorl    %ecx, %ecx
  28         .align  16, 0x90
  29 .LBB0_2:                                # %for.body3
  30                                         #   Parent Loop BB0_1 Depth=1
  31                                         # =>  This Inner Loop Header: Depth=2
  32         movl    %ecx, %edx
  33         imull   %r8d, %edx
  34         movl    %edx, %esi
  35         sarl    $31, %esi
  36         shrl    $22, %esi
  37         addl    %edx, %esi
  38         andl    $-1024, %esi            # imm = 0xFFFFFFFFFFFFFC00
  39         negl    %esi
  40         movq    %r8, %rax
  41         shlq    $11, %rax
  42         leal    1(%rdx,%rsi), %edi
  43         leaq    (%rax,%rax,2), %rsi
  44         leaq    1(%rcx), %rdx
  45         cmpq    $1536, %rdx             # imm = 0x600
  46         vcvtsi2sdl      %edi, %xmm0, %xmm1
  47         vmulsd  %xmm0, %xmm1, %xmm1
  48         vcvtsd2ss       %xmm1, %xmm1, %xmm1
  49         vmovss  %xmm1, A(%rsi,%rcx,4)
  50         vmovss  %xmm1, B(%rsi,%rcx,4)
  51         movq    %rdx, %rcx
  52         jne     .LBB0_2
  53 # BB#3:                                 # %for.inc17
  54                                         #   in Loop: Header=BB0_1 Depth=1
  55         incq    %r8
  56         cmpq    $1536, %r8              # imm = 0x600
  57         jne     .LBB0_1
  58 # BB#4:                                 # %for.end19
  59         popq    %rbp
  60         ret
  61 .Ltmp5:
  62         .size   init_array, .Ltmp5-init_array
  63         .cfi_endproc
  64
  65         .globl  print_array
  66         .align  16, 0x90
  67         .type   print_array,@function
  68 print_array:                            # @print_array
  69         .cfi_startproc
  70 # BB#0:                                 # %entry
  71         pushq   %rbp
  72 .Ltmp9:
  73         .cfi_def_cfa_offset 16
  74 .Ltmp10:
  75         .cfi_offset %rbp, -16
  76         movq    %rsp, %rbp
  77 .Ltmp11:
  78         .cfi_def_cfa_register %rbp
  79         pushq   %r15
  80         pushq   %r14
  81         pushq   %r12
  82         pushq   %rbx
  83 .Ltmp12:
  84         .cfi_offset %rbx, -48
  85 .Ltmp13:
  86         .cfi_offset %r12, -40
  87 .Ltmp14:
  88         .cfi_offset %r14, -32
  89 .Ltmp15:
  90         .cfi_offset %r15, -24
  91         xorl    %r14d, %r14d
  92         movl    $C, %r15d
  93         .align  16, 0x90
  94 .LBB1_1:                                # %for.cond1.preheader
  95                                         # =>This Loop Header: Depth=1
  96                                         #     Child Loop BB1_2 Depth 2
  97         movq    stdout(%rip), %rax
  98         movq    %r15, %r12
  99         xorl    %ebx, %ebx
 100         .align  16, 0x90
 101 .LBB1_2:                                # %for.body3
 102                                         #   Parent Loop BB1_1 Depth=1
 103                                         # =>  This Inner Loop Header: Depth=2
 104         vmovss  (%r12), %xmm0
 105         vcvtss2sd       %xmm0, %xmm0, %xmm0
 106         movq    %rax, %rdi
 107         movl    $.L.str, %esi
 108         movb    $1, %al
 109         callq   fprintf
 110         movslq  %ebx, %rax
 111         imulq   $1717986919, %rax, %rcx # imm = 0x66666667
 112         movq    %rcx, %rdx
 113         shrq    $63, %rdx
 114         sarq    $37, %rcx
 115         addl    %edx, %ecx
 116         imull   $80, %ecx, %ecx
 117         subl    %ecx, %eax
 118         cmpl    $79, %eax
 119         jne     .LBB1_4
 120 # BB#3:                                 # %if.then
 121                                         #   in Loop: Header=BB1_2 Depth=2
 122         movq    stdout(%rip), %rsi
 123         movl    $10, %edi
 124         callq   fputc
 125 .LBB1_4:                                # %for.inc
 126                                         #   in Loop: Header=BB1_2 Depth=2
 127         addq    $4, %r12
 128         incq    %rbx
 129         movq    stdout(%rip), %rax
 130         cmpq    $1536, %rbx             # imm = 0x600
 131         jne     .LBB1_2
 132 # BB#5:                                 # %for.end
 133                                         #   in Loop: Header=BB1_1 Depth=1
 134         movl    $10, %edi
 135         movq    %rax, %rsi
 136         callq   fputc
 137         addq    $6144, %r15             # imm = 0x1800
 138         incq    %r14
 139         cmpq    $1536, %r14             # imm = 0x600
 140         jne     .LBB1_1
 141 # BB#6:                                 # %for.end12
 142         popq    %rbx
 143         popq    %r12
 144         popq    %r14
 145         popq    %r15
 146         popq    %rbp
 147         ret
 148 .Ltmp16:
 149         .size   print_array, .Ltmp16-print_array
 150         .cfi_endproc
 151
 152         .section        .rodata.cst8,"aM",@progbits,8
 153         .align  8
 154 .LCPI2_0:
 155         .quad   4602678819172646912     # double 0.5
 156         .text
 157         .globl  main
 158         .align  16, 0x90
 159         .type   main,@function
 160 main:                                   # @main
 161         .cfi_startproc
 162 # BB#0:                                 # %entry
 163         pushq   %rbp
 164 .Ltmp19:
 165         .cfi_def_cfa_offset 16
 166 .Ltmp20:
 167         .cfi_offset %rbp, -16
 168         movq    %rsp, %rbp
 169 .Ltmp21:
 170         .cfi_def_cfa_register %rbp
 171         xorl    %r8d, %r8d
 172         vmovsd  .LCPI2_0(%rip), %xmm0
 173         .align  16, 0x90
 174 .LBB2_1:                                # %for.cond1.preheader.i
 175                                         # =>This Loop Header: Depth=1
 176                                         #     Child Loop BB2_2 Depth 2
 177         xorl    %ecx, %ecx
 178         .align  16, 0x90
 179 .LBB2_2:                                # %for.body3.i
 180                                         #   Parent Loop BB2_1 Depth=1
 181                                         # =>  This Inner Loop Header: Depth=2
 182         movl    %ecx, %edx
 183         imull   %r8d, %edx
 184         movl    %edx, %esi
 185         sarl    $31, %esi
 186         shrl    $22, %esi
 187         addl    %edx, %esi
 188         andl    $-1024, %esi            # imm = 0xFFFFFFFFFFFFFC00
 189         negl    %esi
 190         movq    %r8, %rax
 191         shlq    $11, %rax
 192         leal    1(%rdx,%rsi), %edi
 193         leaq    (%rax,%rax,2), %rsi
 194         leaq    1(%rcx), %rdx
 195         cmpq    $1536, %rdx             # imm = 0x600
 196         vcvtsi2sdl      %edi, %xmm0, %xmm1
 197         vmulsd  %xmm0, %xmm1, %xmm1
 198         vcvtsd2ss       %xmm1, %xmm1, %xmm1
 199         vmovss  %xmm1, A(%rsi,%rcx,4)
 200         vmovss  %xmm1, B(%rsi,%rcx,4)
 201         movq    %rdx, %rcx
 202         jne     .LBB2_2
 203 # BB#3:                                 # %for.inc17.i
 204                                         #   in Loop: Header=BB2_1 Depth=1
 205         incq    %r8
 206         cmpq    $1536, %r8              # imm = 0x600
 207         jne     .LBB2_1
 208 # BB#4:
 209         xorl    %r8d, %r8d
 210         movl    $A, %r9d
 211         .align  16, 0x90
 212 .LBB2_5:                                # %for.cond1.preheader
 213                                         # =>This Loop Header: Depth=1
 214                                         #     Child Loop BB2_6 Depth 2
 215                                         #       Child Loop BB2_7 Depth 3
 216         leaq    (%r8,%r8,2), %rdx
 217         shlq    $11, %rdx
 218         leaq    C(%rdx), %rsi
 219         xorl    %edi, %edi
 220         .align  16, 0x90
 221 .LBB2_6:                                # %for.body3
 222                                         #   Parent Loop BB2_5 Depth=1
 223                                         # =>  This Loop Header: Depth=2
 224                                         #       Child Loop BB2_7 Depth 3
 225         movl    $0, (%rsi)
 226         vxorps  %xmm0, %xmm0, %xmm0
 227         movq    $-9437184, %rax         # imm = 0xFFFFFFFFFF700000
 228         movq    %r9, %rcx
 229         .align  16, 0x90
 230 .LBB2_7:                                # %for.body8
 231                                         #   Parent Loop BB2_5 Depth=1
 232                                         #     Parent Loop BB2_6 Depth=2
 233                                         # =>    This Inner Loop Header: Depth=3
 234         vmovss  (%rcx), %xmm1
 235         vmulss  B+9437184(%rax,%rdi,4), %xmm1, %xmm1
 236         vaddss  %xmm1, %xmm0, %xmm0
 237         addq    $4, %rcx
 238         addq    $6144, %rax             # imm = 0x1800
 239         jne     .LBB2_7
 240 # BB#8:                                 # %for.inc25
 241                                         #   in Loop: Header=BB2_6 Depth=2
 242         vmovss  %xmm0, (%rsi)
 243         leaq    C+4(%rdx,%rdi,4), %rsi
 244         incq    %rdi
 245         cmpq    $1536, %rdi             # imm = 0x600
 246         jne     .LBB2_6
 247 # BB#9:                                 # %for.inc28
 248                                         #   in Loop: Header=BB2_5 Depth=1
 249         addq    $6144, %r9              # imm = 0x1800
 250         incq    %r8
 251         cmpq    $1536, %r8              # imm = 0x600
 252         jne     .LBB2_5
 253 # BB#10:                                # %for.end30
 254         xorl    %eax, %eax
 255         popq    %rbp
 256         ret
 257 .Ltmp22:
 258         .size   main, .Ltmp22-main
 259         .cfi_endproc
 260
 261         .type   A,@object               # @A
 262         .comm   A,9437184,16
 263         .type   B,@object               # @B
 264         .comm   B,9437184,16
 265         .type   .L.str,@object          # @.str
 266         .section        .rodata.str1.1,"aMS",@progbits,1
 267 .L.str:
 268         .asciz   "%lf "
 269         .size   .L.str, 5
 270
 271         .type   C,@object               # @C
 272         .comm   C,9437184,16
 273
 274         .section        ".note.GNU-stack","",@progbits