test/tools/llvm-mca/X86/BdVer2/dependency-breaking-pcmpeq.s

   1 # NOTE: Assertions have been autogenerated by utils/update_mca_test_checks.py
   2 # RUN: llvm-mca -mtriple=x86_64-unknown-unknown -mcpu=bdver2 -timeline -timeline-max-iterations=3 -iterations=1500 < %s | FileCheck %s
   3
   4 # perf stat reports an IPC of 2.00 for this block of code.
   5
   6 # All of the vector packed compares from this test are dependency breaking
   7 # instructions. That means, there is no RAW dependency between any of the
   8 # instructions, and the code can be fully parallelized in hardware.
   9
  10 vpcmpeqb %xmm0, %xmm0, %xmm1
  11 vpcmpeqw %xmm1, %xmm1, %xmm2
  12 vpcmpeqd %xmm2, %xmm2, %xmm3
  13 vpcmpeqq %xmm3, %xmm3, %xmm0
  14
  15 # CHECK:      Iterations:        1500
  16 # CHECK-NEXT: Instructions:      6000
  17 # CHECK-NEXT: Total Cycles:      6003
  18 # CHECK-NEXT: Total uOps:        6000
  19
  20 # CHECK:      Dispatch Width:    4
  21 # CHECK-NEXT: uOps Per Cycle:    1.00
  22 # CHECK-NEXT: IPC:               1.00
  23 # CHECK-NEXT: Block RThroughput: 4.0
  24
  25 # CHECK:      Instruction Info:
  26 # CHECK-NEXT: [1]: #uOps
  27 # CHECK-NEXT: [2]: Latency
  28 # CHECK-NEXT: [3]: RThroughput
  29 # CHECK-NEXT: [4]: MayLoad
  30 # CHECK-NEXT: [5]: MayStore
  31 # CHECK-NEXT: [6]: HasSideEffects (U)
  32
  33 # CHECK:      [1]    [2]    [3]    [4]    [5]    [6]    Instructions:
  34 # CHECK-NEXT:  1      2     1.00                        vpcmpeqb        %xmm0, %xmm0, %xmm1
  35 # CHECK-NEXT:  1      2     1.00                        vpcmpeqw        %xmm1, %xmm1, %xmm2
  36 # CHECK-NEXT:  1      2     1.00                        vpcmpeqd        %xmm2, %xmm2, %xmm3
  37 # CHECK-NEXT:  1      2     1.00                        vpcmpeqq        %xmm3, %xmm3, %xmm0
  38
  39 # CHECK:      Resources:
  40 # CHECK-NEXT: [0.0] - PdAGLU01
  41 # CHECK-NEXT: [0.1] - PdAGLU01
  42 # CHECK-NEXT: [1]   - PdBranch
  43 # CHECK-NEXT: [2]   - PdCount
  44 # CHECK-NEXT: [3]   - PdDiv
  45 # CHECK-NEXT: [4]   - PdEX0
  46 # CHECK-NEXT: [5]   - PdEX1
  47 # CHECK-NEXT: [6]   - PdFPCVT
  48 # CHECK-NEXT: [7.0] - PdFPFMA
  49 # CHECK-NEXT: [7.1] - PdFPFMA
  50 # CHECK-NEXT: [8.0] - PdFPMAL
  51 # CHECK-NEXT: [8.1] - PdFPMAL
  52 # CHECK-NEXT: [9]   - PdFPMMA
  53 # CHECK-NEXT: [10]  - PdFPSTO
  54 # CHECK-NEXT: [11]  - PdFPU0
  55 # CHECK-NEXT: [12]  - PdFPU1
  56 # CHECK-NEXT: [13]  - PdFPU2
  57 # CHECK-NEXT: [14]  - PdFPU3
  58 # CHECK-NEXT: [15]  - PdFPXBR
  59 # CHECK-NEXT: [16.0] - PdLoad
  60 # CHECK-NEXT: [16.1] - PdLoad
  61 # CHECK-NEXT: [17]  - PdMul
  62 # CHECK-NEXT: [18]  - PdStore
  63
  64 # CHECK:      Resource pressure per iteration:
  65 # CHECK-NEXT: [0.0]  [0.1]  [1]    [2]    [3]    [4]    [5]    [6]    [7.0]  [7.1]  [8.0]  [8.1]  [9]    [10]   [11]   [12]   [13]   [14]   [15]   [16.0] [16.1] [17]   [18]
  66 # CHECK-NEXT:  -      -      -      -      -      -      -      -      -      -     4.00   4.00    -      -     2.00   2.00    -      -      -      -      -      -      -
  67
  68 # CHECK:      Resource pressure by instruction:
  69 # CHECK-NEXT: [0.0]  [0.1]  [1]    [2]    [3]    [4]    [5]    [6]    [7.0]  [7.1]  [8.0]  [8.1]  [9]    [10]   [11]   [12]   [13]   [14]   [15]   [16.0] [16.1] [17]   [18]   Instructions:
  70 # CHECK-NEXT:  -      -      -      -      -      -      -      -      -      -      -     2.00    -      -      -     1.00    -      -      -      -      -      -      -     vpcmpeqb %xmm0, %xmm0, %xmm1
  71 # CHECK-NEXT:  -      -      -      -      -      -      -      -      -      -      -     2.00    -      -      -     1.00    -      -      -      -      -      -      -     vpcmpeqw %xmm1, %xmm1, %xmm2
  72 # CHECK-NEXT:  -      -      -      -      -      -      -      -      -      -     2.00    -      -      -     1.00    -      -      -      -      -      -      -      -     vpcmpeqd %xmm2, %xmm2, %xmm3
  73 # CHECK-NEXT:  -      -      -      -      -      -      -      -      -      -     2.00    -      -      -     1.00    -      -      -      -      -      -      -      -     vpcmpeqq %xmm3, %xmm3, %xmm0
  74
  75 # CHECK:      Timeline view:
  76 # CHECK-NEXT:                     01234
  77 # CHECK-NEXT: Index     0123456789
  78
  79 # CHECK:      [0,0]     DeeER.    .   .   vpcmpeqb      %xmm0, %xmm0, %xmm1
  80 # CHECK-NEXT: [0,1]     D==eeER   .   .   vpcmpeqw      %xmm1, %xmm1, %xmm2
  81 # CHECK-NEXT: [0,2]     DeeE--R   .   .   vpcmpeqd      %xmm2, %xmm2, %xmm3
  82 # CHECK-NEXT: [0,3]     D==eeER   .   .   vpcmpeqq      %xmm3, %xmm3, %xmm0
  83 # CHECK-NEXT: [1,0]     .D===eeER .   .   vpcmpeqb      %xmm0, %xmm0, %xmm1
  84 # CHECK-NEXT: [1,1]     .D=====eeER   .   vpcmpeqw      %xmm1, %xmm1, %xmm2
  85 # CHECK-NEXT: [1,2]     .D===eeE--R   .   vpcmpeqd      %xmm2, %xmm2, %xmm3
  86 # CHECK-NEXT: [1,3]     .D=====eeER   .   vpcmpeqq      %xmm3, %xmm3, %xmm0
  87 # CHECK-NEXT: [2,0]     . D======eeER .   vpcmpeqb      %xmm0, %xmm0, %xmm1
  88 # CHECK-NEXT: [2,1]     . D========eeER   vpcmpeqw      %xmm1, %xmm1, %xmm2
  89 # CHECK-NEXT: [2,2]     . D======eeE--R   vpcmpeqd      %xmm2, %xmm2, %xmm3
  90 # CHECK-NEXT: [2,3]     . D========eeER   vpcmpeqq      %xmm3, %xmm3, %xmm0
  91
  92 # CHECK:      Average Wait times (based on the timeline view):
  93 # CHECK-NEXT: [0]: Executions
  94 # CHECK-NEXT: [1]: Average time spent waiting in a scheduler's queue
  95 # CHECK-NEXT: [2]: Average time spent waiting in a scheduler's queue while ready
  96 # CHECK-NEXT: [3]: Average time elapsed from WB until retire stage
  97
  98 # CHECK:            [0]    [1]    [2]    [3]
  99 # CHECK-NEXT: 0.     3     4.0    4.0    0.0       vpcmpeqb     %xmm0, %xmm0, %xmm1
 100 # CHECK-NEXT: 1.     3     6.0    6.0    0.0       vpcmpeqw     %xmm1, %xmm1, %xmm2
 101 # CHECK-NEXT: 2.     3     4.0    4.0    2.0       vpcmpeqd     %xmm2, %xmm2, %xmm3
 102 # CHECK-NEXT: 3.     3     6.0    0.0    0.0       vpcmpeqq     %xmm3, %xmm3, %xmm0