llvm/test/tools/llvm-mca/AMDGPU/gfx10-add-sequence.s

   1 # NOTE: Assertions have been autogenerated by utils/update_mca_test_checks.py
   2 # RUN: llvm-mca -mtriple=amdgcn -mcpu=gfx1010 --timeline --iterations=1 --timeline-max-cycles=0 < %s | FileCheck %s
   3
   4 v_add_f32 v0, v0, v0
   5 v_add_f32 v1, v1, v1
   6 v_add_f32 v2, v1, v0
   7
   8 # CHECK:      Iterations:        1
   9 # CHECK-NEXT: Instructions:      3
  10 # CHECK-NEXT: Total Cycles:      12
  11 # CHECK-NEXT: Total uOps:        3
  12
  13 # CHECK:      Dispatch Width:    1
  14 # CHECK-NEXT: uOps Per Cycle:    0.25
  15 # CHECK-NEXT: IPC:               0.25
  16 # CHECK-NEXT: Block RThroughput: 3.0
  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      5     1.00                  U     v_add_f32_e32 v0, v0, v0
  28 # CHECK-NEXT:  1      5     1.00                  U     v_add_f32_e32 v1, v1, v1
  29 # CHECK-NEXT:  1      5     1.00                  U     v_add_f32_e32 v2, v1, v0
  30
  31 # CHECK:      Resources:
  32 # CHECK-NEXT: [0]   - HWBranch
  33 # CHECK-NEXT: [1]   - HWExport
  34 # CHECK-NEXT: [2]   - HWLGKM
  35 # CHECK-NEXT: [3]   - HWRC
  36 # CHECK-NEXT: [4]   - HWSALU
  37 # CHECK-NEXT: [5]   - HWTransVALU
  38 # CHECK-NEXT: [6]   - HWVALU
  39 # CHECK-NEXT: [7]   - HWVMEM
  40
  41 # CHECK:      Resource pressure per iteration:
  42 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]
  43 # CHECK-NEXT:  -      -      -     3.00    -      -     3.00    -
  44
  45 # CHECK:      Resource pressure by instruction:
  46 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    Instructions:
  47 # CHECK-NEXT:  -      -      -     1.00    -      -     1.00    -     v_add_f32_e32 v0, v0, v0
  48 # CHECK-NEXT:  -      -      -     1.00    -      -     1.00    -     v_add_f32_e32 v1, v1, v1
  49 # CHECK-NEXT:  -      -      -     1.00    -      -     1.00    -     v_add_f32_e32 v2, v1, v0
  50
  51 # CHECK:      Timeline view:
  52 # CHECK-NEXT:                     01
  53 # CHECK-NEXT: Index     0123456789
  54
  55 # CHECK:      [0,0]     DeeeeE    ..   v_add_f32_e32 v0, v0, v0
  56 # CHECK-NEXT: [0,1]     .DeeeeE   ..   v_add_f32_e32 v1, v1, v1
  57 # CHECK-NEXT: [0,2]     .    .DeeeeE   v_add_f32_e32 v2, v1, v0
  58
  59 # CHECK:      Average Wait times (based on the timeline view):
  60 # CHECK-NEXT: [0]: Executions
  61 # CHECK-NEXT: [1]: Average time spent waiting in a scheduler's queue
  62 # CHECK-NEXT: [2]: Average time spent waiting in a scheduler's queue while ready
  63 # CHECK-NEXT: [3]: Average time elapsed from WB until retire stage
  64
  65 # CHECK:            [0]    [1]    [2]    [3]
  66 # CHECK-NEXT: 0.     1     0.0    0.0    0.0       v_add_f32_e32 v0, v0, v0
  67 # CHECK-NEXT: 1.     1     0.0    0.0    0.0       v_add_f32_e32 v1, v1, v1
  68 # CHECK-NEXT: 2.     1     0.0    0.0    0.0       v_add_f32_e32 v2, v1, v0
  69 # CHECK-NEXT:        1     0.0    0.0    0.0       <total>