| 1 |
# NOTE: Assertions have been autogenerated by utils/update_mca_test_checks.py |
| 2 |
# RUN: llvm-mca -mtriple=aarch64 -mcpu=cortex-a55 --all-views=false --summary-view --iterations=1000 < %s | FileCheck %s |
| 3 |
|
| 4 |
# It appears that ADD and MUL fuse together, if both can be issued in |
| 5 |
# one cycle: |
| 6 |
# |
| 7 |
# add w12, w8, #1 |
| 8 |
# mul w10, w12, w10 |
| 9 |
# |
| 10 |
# FIXME: MCA (and LLVM scheduling model) do not support this. The test |
| 11 |
# case uses different registers to break the pattern. |
| 12 |
|
| 13 |
add w8, w8, #1 |
| 14 |
add w13, w8, #1 |
| 15 |
mul w10, w12, w10 |
| 16 |
|
| 17 |
# CHECK: Iterations: 1000 |
| 18 |
# CHECK-NEXT: Instructions: 3000 |
| 19 |
# CHECK-NEXT: Total Cycles: 3003 |
| 20 |
# CHECK-NEXT: Total uOps: 3000 |
| 21 |
|
| 22 |
# CHECK: Dispatch Width: 2 |
| 23 |
# CHECK-NEXT: uOps Per Cycle: 1.00 |
| 24 |
# CHECK-NEXT: IPC: 1.00 |
| 25 |
# CHECK-NEXT: Block RThroughput: 1.5 |