|
| 1 | +# NOTE: Assertions have been autogenerated by utils/update_mca_test_checks.py |
| 2 | +# RUN: llvm-mca -mtriple=riscv64 -mcpu=mips-p8700 -timeline -iterations=1 < %s | FileCheck %s |
| 3 | + |
| 4 | +# A few instructions to test the pipeline: |
| 5 | +# - Integer division (IDiv) exercises the p8700GpDiv resource. |
| 6 | +# - Integer multiplication (IMul) uses p8700GpMul. |
| 7 | +# - Floating-point multiplication uses the FPUL pipeline. |
| 8 | +# - Load/Store instructions use the LSU pipeline. |
| 9 | +# - Simple ALU instructions test the p8700WriteEitherALU and p8700IssueAL2 resources. |
| 10 | +# - A jump instruction to test the CTI pipeline. |
| 11 | + |
| 12 | + .text |
| 13 | + .globl _start |
| 14 | +_start: |
| 15 | + |
| 16 | +# Integer division: a0 = a1 / a2 |
| 17 | +# Exercises p8700GpDiv resource. |
| 18 | + div a0, a1, a2 |
| 19 | + |
| 20 | +# Integer multiplication: a4 = a1 * a2 |
| 21 | +# Exercises p8700GpMul resource. |
| 22 | + mul a4, a1, a2 |
| 23 | + |
| 24 | +# Floating-point multiply: f1 = f2 * f3 (single precision) |
| 25 | +# Exercises p8700FpuLong + p8700FpuApu resources. |
| 26 | + fmul.s f1, f2, f3 |
| 27 | + |
| 28 | +# Load/Store: load word from a0 into a3, then store a3 into a1 |
| 29 | +# Exercises p8700IssueLSU resource. |
| 30 | + lw a3, 0(a0) |
| 31 | + sw a3, 0(a1) |
| 32 | + |
| 33 | +# Simple ALU operations (adding two registers, rotating bits) |
| 34 | +# Exercises p8700WriteEitherALU. |
| 35 | + add a5, a1, a2 |
| 36 | + ror a6, a5, a2 |
| 37 | + |
| 38 | +# A jump instruction: a simple forward jump |
| 39 | +# Exercises p8700IssueCTI resource. |
| 40 | + jal x0, .Lend |
| 41 | + |
| 42 | + add a7, a4, a0 # Instruction after jump (won't execute) |
| 43 | +.Lend: |
| 44 | + nop |
| 45 | + |
| 46 | +# CHECK: Iterations: 1 |
| 47 | +# CHECK-NEXT: Instructions: 10 |
| 48 | +# CHECK-NEXT: Total Cycles: 17 |
| 49 | +# CHECK-NEXT: Total uOps: 10 |
| 50 | + |
| 51 | +# CHECK: Dispatch Width: 4 |
| 52 | +# CHECK-NEXT: uOps Per Cycle: 0.59 |
| 53 | +# CHECK-NEXT: IPC: 0.59 |
| 54 | +# CHECK-NEXT: Block RThroughput: 7.0 |
| 55 | + |
| 56 | +# CHECK: Instruction Info: |
| 57 | +# CHECK-NEXT: [1]: #uOps |
| 58 | +# CHECK-NEXT: [2]: Latency |
| 59 | +# CHECK-NEXT: [3]: RThroughput |
| 60 | +# CHECK-NEXT: [4]: MayLoad |
| 61 | +# CHECK-NEXT: [5]: MayStore |
| 62 | +# CHECK-NEXT: [6]: HasSideEffects (U) |
| 63 | + |
| 64 | +# CHECK: [1] [2] [3] [4] [5] [6] Instructions: |
| 65 | +# CHECK-NEXT: 1 7 7.00 div a0, a1, a2 |
| 66 | +# CHECK-NEXT: 1 4 1.00 mul a4, a1, a2 |
| 67 | +# CHECK-NEXT: 1 5 1.00 fmul.s ft1, ft2, ft3 |
| 68 | +# CHECK-NEXT: 1 4 1.00 * lw a3, 0(a0) |
| 69 | +# CHECK-NEXT: 1 3 1.00 * sw a3, 0(a1) |
| 70 | +# CHECK-NEXT: 1 1 0.50 add a5, a1, a2 |
| 71 | +# CHECK-NEXT: 1 1 0.50 ror a6, a5, a2 |
| 72 | +# CHECK-NEXT: 1 1 1.00 j .Lend |
| 73 | +# CHECK-NEXT: 1 1 0.50 add a7, a4, a0 |
| 74 | +# CHECK-NEXT: 1 0 0.50 nop |
| 75 | + |
| 76 | +# CHECK: Resources: |
| 77 | +# CHECK-NEXT: [0.0] - p8700AGQ |
| 78 | +# CHECK-NEXT: [0.1] - p8700AGQ |
| 79 | +# CHECK-NEXT: [0.2] - p8700AGQ |
| 80 | +# CHECK-NEXT: [1] - p8700ALQ |
| 81 | +# CHECK-NEXT: [2.0] - p8700FPQ |
| 82 | +# CHECK-NEXT: [2.1] - p8700FPQ |
| 83 | +# CHECK-NEXT: [2.2] - p8700FPQ |
| 84 | +# CHECK-NEXT: [3] - p8700FpuApu |
| 85 | +# CHECK-NEXT: [4] - p8700FpuLong |
| 86 | +# CHECK-NEXT: [5] - p8700GpDiv |
| 87 | +# CHECK-NEXT: [6] - p8700GpMul |
| 88 | +# CHECK-NEXT: [7] - p8700IssueAL2 |
| 89 | +# CHECK-NEXT: [8] - p8700IssueCTI |
| 90 | +# CHECK-NEXT: [9] - p8700IssueFPUL |
| 91 | +# CHECK-NEXT: [10] - p8700IssueFPUS |
| 92 | +# CHECK-NEXT: [11] - p8700IssueLSU |
| 93 | + |
| 94 | +# CHECK: Resource pressure per iteration: |
| 95 | +# CHECK-NEXT: [0.0] [0.1] [0.2] [1] [2.0] [2.1] [2.2] [3] [4] [5] [6] [7] [8] [9] [10] [11] |
| 96 | +# CHECK-NEXT: 1.00 1.00 1.00 2.00 - - - 1.00 1.00 7.00 1.00 2.00 1.00 - - 2.00 |
| 97 | + |
| 98 | +# CHECK: Resource pressure by instruction: |
| 99 | +# CHECK-NEXT: [0.0] [0.1] [0.2] [1] [2.0] [2.1] [2.2] [3] [4] [5] [6] [7] [8] [9] [10] [11] Instructions: |
| 100 | +# CHECK-NEXT: - - - - - - - - - 7.00 - - - - - - div a0, a1, a2 |
| 101 | +# CHECK-NEXT: - - - - - - - - - - 1.00 - - - - - mul a4, a1, a2 |
| 102 | +# CHECK-NEXT: - - - - - - - 1.00 1.00 - - - - - - - fmul.s ft1, ft2, ft3 |
| 103 | +# CHECK-NEXT: - 1.00 - - - - - - - - - - - - - 1.00 lw a3, 0(a0) |
| 104 | +# CHECK-NEXT: 1.00 - - - - - - - - - - - - - - 1.00 sw a3, 0(a1) |
| 105 | +# CHECK-NEXT: - - - - - - - - - - - 1.00 - - - - add a5, a1, a2 |
| 106 | +# CHECK-NEXT: - - - 1.00 - - - - - - - - - - - - ror a6, a5, a2 |
| 107 | +# CHECK-NEXT: - - 1.00 - - - - - - - - - 1.00 - - - j .Lend |
| 108 | +# CHECK-NEXT: - - - 1.00 - - - - - - - - - - - - add a7, a4, a0 |
| 109 | +# CHECK-NEXT: - - - - - - - - - - - 1.00 - - - - nop |
| 110 | + |
| 111 | +# CHECK: Timeline view: |
| 112 | +# CHECK-NEXT: 0123456 |
| 113 | +# CHECK-NEXT: Index 0123456789 |
| 114 | + |
| 115 | +# CHECK: [0,0] DeeeeeeeER. .. div a0, a1, a2 |
| 116 | +# CHECK-NEXT: [0,1] DeeeeE---R. .. mul a4, a1, a2 |
| 117 | +# CHECK-NEXT: [0,2] DeeeeeE--R. .. fmul.s ft1, ft2, ft3 |
| 118 | +# CHECK-NEXT: [0,3] D=======eeeeER .. lw a3, 0(a0) |
| 119 | +# CHECK-NEXT: [0,4] .D==========eeeER sw a3, 0(a1) |
| 120 | +# CHECK-NEXT: [0,5] .DeE------------R add a5, a1, a2 |
| 121 | +# CHECK-NEXT: [0,6] .D=eE-----------R ror a6, a5, a2 |
| 122 | +# CHECK-NEXT: [0,7] .DeE------------R j .Lend |
| 123 | +# CHECK-NEXT: [0,8] . D=====eE------R add a7, a4, a0 |
| 124 | +# CHECK-NEXT: [0,9] . DE------------R nop |
| 125 | + |
| 126 | +# CHECK: Average Wait times (based on the timeline view): |
| 127 | +# CHECK-NEXT: [0]: Executions |
| 128 | +# CHECK-NEXT: [1]: Average time spent waiting in a scheduler's queue |
| 129 | +# CHECK-NEXT: [2]: Average time spent waiting in a scheduler's queue while ready |
| 130 | +# CHECK-NEXT: [3]: Average time elapsed from WB until retire stage |
| 131 | + |
| 132 | +# CHECK: [0] [1] [2] [3] |
| 133 | +# CHECK-NEXT: 0. 1 1.0 1.0 0.0 div a0, a1, a2 |
| 134 | +# CHECK-NEXT: 1. 1 1.0 1.0 3.0 mul a4, a1, a2 |
| 135 | +# CHECK-NEXT: 2. 1 1.0 1.0 2.0 fmul.s ft1, ft2, ft3 |
| 136 | +# CHECK-NEXT: 3. 1 8.0 0.0 0.0 lw a3, 0(a0) |
| 137 | +# CHECK-NEXT: 4. 1 11.0 0.0 0.0 sw a3, 0(a1) |
| 138 | +# CHECK-NEXT: 5. 1 1.0 1.0 12.0 add a5, a1, a2 |
| 139 | +# CHECK-NEXT: 6. 1 2.0 0.0 11.0 ror a6, a5, a2 |
| 140 | +# CHECK-NEXT: 7. 1 1.0 1.0 12.0 j .Lend |
| 141 | +# CHECK-NEXT: 8. 1 6.0 0.0 6.0 add a7, a4, a0 |
| 142 | +# CHECK-NEXT: 9. 1 1.0 1.0 12.0 nop |
| 143 | +# CHECK-NEXT: 1 3.3 0.6 5.8 <total> |
0 commit comments