test/tools/llvm-mca/X86/BtVer2/dot-product.s

   1 # NOTE: Assertions have been autogenerated by utils/update_mca_test_checks.py
   2 # RUN: llvm-mca -mtriple=x86_64-unknown-unknown -mcpu=btver2 -iterations=300 -timeline -timeline-max-iterations=3 < %s | FileCheck %s
   3
   4 vmulps   %xmm0, %xmm1, %xmm2
   5 vhaddps  %xmm2, %xmm2, %xmm3
   6 vhaddps  %xmm3, %xmm3, %xmm4
   7
   8 # CHECK:      Iterations:        300
   9 # CHECK-NEXT: Instructions:      900
  10 # CHECK-NEXT: Total Cycles:      611
  11 # CHECK-NEXT: Total uOps:        900
  12
  13 # CHECK:      Dispatch Width:    2
  14 # CHECK-NEXT: uOps Per Cycle:    1.47
  15 # CHECK-NEXT: IPC:               1.47
  16 # CHECK-NEXT: Block RThroughput: 2.0
  17
  18 # CHECK:      Instruction Info:
  19 # CHECK-NEXT: [1]: #uOps
  20 # CHECK-NEXT: [2]: Latency
  21 # CHECK-NEXT: [3]: RThroughput
  22 # CHECK-NEXT: [4]: MayLoad
  23 # CHECK-NEXT: [5]: MayStore
  24 # CHECK-NEXT: [6]: HasSideEffects (U)
  25
  26 # CHECK:      [1]    [2]    [3]    [4]    [5]    [6]    Instructions:
  27 # CHECK-NEXT:  1      2     1.00                        vmulps  %xmm0, %xmm1, %xmm2
  28 # CHECK-NEXT:  1      4     1.00                        vhaddps %xmm2, %xmm2, %xmm3
  29 # CHECK-NEXT:  1      4     1.00                        vhaddps %xmm3, %xmm3, %xmm4
  30
  31 # CHECK:      Resources:
  32 # CHECK-NEXT: [0]   - JALU0
  33 # CHECK-NEXT: [1]   - JALU1
  34 # CHECK-NEXT: [2]   - JDiv
  35 # CHECK-NEXT: [3]   - JFPA
  36 # CHECK-NEXT: [4]   - JFPM
  37 # CHECK-NEXT: [5]   - JFPU0
  38 # CHECK-NEXT: [6]   - JFPU1
  39 # CHECK-NEXT: [7]   - JLAGU
  40 # CHECK-NEXT: [8]   - JMul
  41 # CHECK-NEXT: [9]   - JSAGU
  42 # CHECK-NEXT: [10]  - JSTC
  43 # CHECK-NEXT: [11]  - JVALU0
  44 # CHECK-NEXT: [12]  - JVALU1
  45 # CHECK-NEXT: [13]  - JVIMUL
  46
  47 # CHECK:      Resource pressure per iteration:
  48 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    [10]   [11]   [12]   [13]
  49 # CHECK-NEXT:  -      -      -     2.00   1.00   2.00   1.00    -      -      -      -      -      -      -
  50
  51 # CHECK:      Resource pressure by instruction:
  52 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    [10]   [11]   [12]   [13]   Instructions:
  53 # CHECK-NEXT:  -      -      -      -     1.00    -     1.00    -      -      -      -      -      -      -     vmulps  %xmm0, %xmm1, %xmm2
  54 # CHECK-NEXT:  -      -      -     1.00    -     1.00    -      -      -      -      -      -      -      -     vhaddps %xmm2, %xmm2, %xmm3
  55 # CHECK-NEXT:  -      -      -     1.00    -     1.00    -      -      -      -      -      -      -      -     vhaddps %xmm3, %xmm3, %xmm4
  56
  57 # CHECK:      Timeline view:
  58 # CHECK-NEXT:                     012345
  59 # CHECK-NEXT: Index     0123456789
  60
  61 # CHECK:      [0,0]     DeeER.    .    .   vmulps       %xmm0, %xmm1, %xmm2
  62 # CHECK-NEXT: [0,1]     D==eeeeER .    .   vhaddps      %xmm2, %xmm2, %xmm3
  63 # CHECK-NEXT: [0,2]     .D=====eeeeER  .   vhaddps      %xmm3, %xmm3, %xmm4
  64 # CHECK-NEXT: [1,0]     .DeeE-------R  .   vmulps       %xmm0, %xmm1, %xmm2
  65 # CHECK-NEXT: [1,1]     . D=eeeeE----R .   vhaddps      %xmm2, %xmm2, %xmm3
  66 # CHECK-NEXT: [1,2]     . D=====eeeeER .   vhaddps      %xmm3, %xmm3, %xmm4
  67 # CHECK-NEXT: [2,0]     .  DeeE-------R.   vmulps       %xmm0, %xmm1, %xmm2
  68 # CHECK-NEXT: [2,1]     .  D==eeeeE---R.   vhaddps      %xmm2, %xmm2, %xmm3
  69 # CHECK-NEXT: [2,2]     .   D=====eeeeER   vhaddps      %xmm3, %xmm3, %xmm4
  70
  71 # CHECK:      Average Wait times (based on the timeline view):
  72 # CHECK-NEXT: [0]: Executions
  73 # CHECK-NEXT: [1]: Average time spent waiting in a scheduler's queue
  74 # CHECK-NEXT: [2]: Average time spent waiting in a scheduler's queue while ready
  75 # CHECK-NEXT: [3]: Average time elapsed from WB until retire stage
  76
  77 # CHECK:            [0]    [1]    [2]    [3]
  78 # CHECK-NEXT: 0.     3     1.0    1.0    4.7       vmulps       %xmm0, %xmm1, %xmm2
  79 # CHECK-NEXT: 1.     3     2.7    0.0    2.3       vhaddps      %xmm2, %xmm2, %xmm3
  80 # CHECK-NEXT: 2.     3     6.0    0.0    0.0       vhaddps      %xmm3, %xmm3, %xmm4