test/tools/llvm-mca/X86/BtVer2/bottleneck-hints-4.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 -bottleneck-analysis < %s | FileCheck %s
   3
   4 vmulps   %xmm0, %xmm1, %xmm2
   5 vhaddps  %xmm2, %xmm2, %xmm3
   6 vhaddps  %xmm3, %xmm3, %xmm4
   7
   8 # CHECK:      Iterations:        100
   9 # CHECK-NEXT: Instructions:      300
  10 # CHECK-NEXT: Total Cycles:      211
  11 # CHECK-NEXT: Total uOps:        300
  12
  13 # CHECK:      Dispatch Width:    2
  14 # CHECK-NEXT: uOps Per Cycle:    1.42
  15 # CHECK-NEXT: IPC:               1.42
  16 # CHECK-NEXT: Block RThroughput: 2.0
  17
  18 # CHECK:      Cycles with backend pressure increase [ 40.76% ]
  19 # CHECK-NEXT: Throughput Bottlenecks:
  20 # CHECK-NEXT:   Resource Pressure       [ 39.34% ]
  21 # CHECK-NEXT:   - JFPA  [ 39.34% ]
  22 # CHECK-NEXT:   - JFPU0  [ 39.34% ]
  23 # CHECK-NEXT:   Data Dependencies:      [ 1.42% ]
  24 # CHECK-NEXT:   - Register Dependencies [ 1.42% ]
  25 # CHECK-NEXT:   - Memory Dependencies   [ 0.00% ]
  26
  27 # CHECK:      Critical sequence based on the simulation:
  28
  29 # CHECK:                    Instruction                                 Dependency Information
  30 # CHECK-NEXT:  +----< 2.    vhaddps     %xmm3, %xmm3, %xmm4
  31 # CHECK-NEXT:  |
  32 # CHECK-NEXT:  |    < loop carried >
  33 # CHECK-NEXT:  |
  34 # CHECK-NEXT:  |      0.    vmulps      %xmm0, %xmm1, %xmm2
  35 # CHECK-NEXT:  +----> 1.    vhaddps     %xmm2, %xmm2, %xmm3               ## RESOURCE interference:  JFPA [ probability: 73% ]
  36 # CHECK-NEXT:  +----> 2.    vhaddps     %xmm3, %xmm3, %xmm4               ## REGISTER dependency:  %xmm3
  37 # CHECK-NEXT:  |
  38 # CHECK-NEXT:  |    < loop carried >
  39 # CHECK-NEXT:  |
  40 # CHECK-NEXT:  +----> 1.    vhaddps     %xmm2, %xmm2, %xmm3               ## RESOURCE interference:  JFPA [ probability: 73% ]
  41
  42 # CHECK:      Instruction Info:
  43 # CHECK-NEXT: [1]: #uOps
  44 # CHECK-NEXT: [2]: Latency
  45 # CHECK-NEXT: [3]: RThroughput
  46 # CHECK-NEXT: [4]: MayLoad
  47 # CHECK-NEXT: [5]: MayStore
  48 # CHECK-NEXT: [6]: HasSideEffects (U)
  49
  50 # CHECK:      [1]    [2]    [3]    [4]    [5]    [6]    Instructions:
  51 # CHECK-NEXT:  1      2     1.00                        vmulps  %xmm0, %xmm1, %xmm2
  52 # CHECK-NEXT:  1      4     1.00                        vhaddps %xmm2, %xmm2, %xmm3
  53 # CHECK-NEXT:  1      4     1.00                        vhaddps %xmm3, %xmm3, %xmm4
  54
  55 # CHECK:      Resources:
  56 # CHECK-NEXT: [0]   - JALU0
  57 # CHECK-NEXT: [1]   - JALU1
  58 # CHECK-NEXT: [2]   - JDiv
  59 # CHECK-NEXT: [3]   - JFPA
  60 # CHECK-NEXT: [4]   - JFPM
  61 # CHECK-NEXT: [5]   - JFPU0
  62 # CHECK-NEXT: [6]   - JFPU1
  63 # CHECK-NEXT: [7]   - JLAGU
  64 # CHECK-NEXT: [8]   - JMul
  65 # CHECK-NEXT: [9]   - JSAGU
  66 # CHECK-NEXT: [10]  - JSTC
  67 # CHECK-NEXT: [11]  - JVALU0
  68 # CHECK-NEXT: [12]  - JVALU1
  69 # CHECK-NEXT: [13]  - JVIMUL
  70
  71 # CHECK:      Resource pressure per iteration:
  72 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    [10]   [11]   [12]   [13]
  73 # CHECK-NEXT:  -      -      -     2.00   1.00   2.00   1.00    -      -      -      -      -      -      -
  74
  75 # CHECK:      Resource pressure by instruction:
  76 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    [10]   [11]   [12]   [13]   Instructions:
  77 # CHECK-NEXT:  -      -      -      -     1.00    -     1.00    -      -      -      -      -      -      -     vmulps  %xmm0, %xmm1, %xmm2
  78 # CHECK-NEXT:  -      -      -     1.00    -     1.00    -      -      -      -      -      -      -      -     vhaddps %xmm2, %xmm2, %xmm3
  79 # CHECK-NEXT:  -      -      -     1.00    -     1.00    -      -      -      -      -      -      -      -     vhaddps %xmm3, %xmm3, %xmm4