maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

[MCA] Fix CarryOver check in the DispatchStage (PR50174). 

Early exit from method DispatchStage::isAvailable() if the dispatch group is
already full. Not all instructions declare at least one uOP.
Fixes PR50174.
This commit is contained in:
Andrea Di Biagio 2021-04-30 13:53:33 +01:00
parent 6c31295493
commit 8bd4f3d547
2 changed files with 89 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
llvm/lib/MCA/Stages/DispatchStage.cpp
View File

@ -156,6 +156,10 @@ Error DispatchStage::cycleStart() {
}
bool DispatchStage::isAvailable(const InstRef &IR) const {
// Conservatively bail out if there are no available dispatch entries.
if (!AvailableEntries)
return false;
const Instruction &Inst = *IR.getInstruction();
unsigned NumMicroOps = Inst.getNumMicroOps();
const InstrDesc &Desc = Inst.getDesc();

85
llvm/test/tools/llvm-mca/ARM/cortex-a57-carryover.s Normal file
View File

@ -0,0 +1,85 @@
# NOTE: Assertions have been autogenerated by utils/update_mca_test_checks.py
# RUN: llvm-mca -mtriple=thumbv7-gnueabihf-linux -mcpu=cortex-a57 -timeline -iterations=10 < %s | FileCheck %s
# Test case from PR50174.
# NOP consumes zero resources, and it is 0 uOPs.
pop {r3, r4, r5, r6, r7, pc}
nop
# CHECK: Iterations: 10
# CHECK-NEXT: Instructions: 20
# CHECK-NEXT: Total Cycles: 43
# CHECK-NEXT: Total uOps: 80
# CHECK: Dispatch Width: 3
# CHECK-NEXT: uOps Per Cycle: 1.86
# CHECK-NEXT: IPC: 0.47
# CHECK-NEXT: Block RThroughput: 4.0
# CHECK: Instruction Info:
# CHECK-NEXT: [1]: #uOps
# CHECK-NEXT: [2]: Latency
# CHECK-NEXT: [3]: RThroughput
# CHECK-NEXT: [4]: MayLoad
# CHECK-NEXT: [5]: MayStore
# CHECK-NEXT: [6]: HasSideEffects (U)
# CHECK: [1] [2] [3] [4] [5] [6] Instructions:
# CHECK-NEXT: 8 4 4.00 * U pop {r3, r4, r5, r6, r7, pc}
# CHECK-NEXT: 0 0 0.00 * * U nop
# CHECK: Resources:
# CHECK-NEXT: [0] - A57UnitB
# CHECK-NEXT: [1.0] - A57UnitI
# CHECK-NEXT: [1.1] - A57UnitI
# CHECK-NEXT: [2] - A57UnitL
# CHECK-NEXT: [3] - A57UnitM
# CHECK-NEXT: [4] - A57UnitS
# CHECK-NEXT: [5] - A57UnitW
# CHECK-NEXT: [6] - A57UnitX
# CHECK: Resource pressure per iteration:
# CHECK-NEXT: [0] [1.0] [1.1] [2] [3] [4] [5] [6]
# CHECK-NEXT: - 2.00 2.00 4.00 - - - -
# CHECK: Resource pressure by instruction:
# CHECK-NEXT: [0] [1.0] [1.1] [2] [3] [4] [5] [6] Instructions:
# CHECK-NEXT: - 2.00 2.00 4.00 - - - - pop {r3, r4, r5, r6, r7, pc}
# CHECK-NEXT: - - - - - - - - nop
# CHECK: Timeline view:
# CHECK-NEXT: 0123456789 0123456789
# CHECK-NEXT: Index 0123456789 0123456789 012
# CHECK: [0,0] DeeeeER . . . . . . . . pop {r3, r4, r5, r6, r7, pc}
# CHECK-NEXT: [0,1] . D---R . . . . . . . . nop
# CHECK-NEXT: [1,0] . D=eeeeER . . . . . . . pop {r3, r4, r5, r6, r7, pc}
# CHECK-NEXT: [1,1] . D----R . . . . . . . nop
# CHECK-NEXT: [2,0] . .D==eeeeER. . . . . . . pop {r3, r4, r5, r6, r7, pc}
# CHECK-NEXT: [2,1] . . DE----R. . . . . . . nop
# CHECK-NEXT: [3,0] . . D===eeeeER . . . . . . pop {r3, r4, r5, r6, r7, pc}
# CHECK-NEXT: [3,1] . . .D=E----R . . . . . . nop
# CHECK-NEXT: [4,0] . . . D====eeeeER . . . . . pop {r3, r4, r5, r6, r7, pc}
# CHECK-NEXT: [4,1] . . . D==E----R . . . . . nop
# CHECK-NEXT: [5,0] . . . D=====eeeeER . . . . pop {r3, r4, r5, r6, r7, pc}
# CHECK-NEXT: [5,1] . . . . D===E----R . . . . nop
# CHECK-NEXT: [6,0] . . . . D======eeeeER . . . pop {r3, r4, r5, r6, r7, pc}
# CHECK-NEXT: [6,1] . . . . D====E----R . . . nop
# CHECK-NEXT: [7,0] . . . . .D=======eeeeER. . . pop {r3, r4, r5, r6, r7, pc}
# CHECK-NEXT: [7,1] . . . . . D=====E----R. . . nop
# CHECK-NEXT: [8,0] . . . . . D========eeeeER . . pop {r3, r4, r5, r6, r7, pc}
# CHECK-NEXT: [8,1] . . . . . .D======E----R . . nop
# CHECK-NEXT: [9,0] . . . . . . D=========eeeeER pop {r3, r4, r5, r6, r7, pc}
# CHECK-NEXT: [9,1] . . . . . . D=======E----R nop
# CHECK: Average Wait times (based on the timeline view):
# CHECK-NEXT: [0]: Executions
# CHECK-NEXT: [1]: Average time spent waiting in a scheduler's queue
# CHECK-NEXT: [2]: Average time spent waiting in a scheduler's queue while ready
# CHECK-NEXT: [3]: Average time elapsed from WB until retire stage
# CHECK: [0] [1] [2] [3]
# CHECK-NEXT: 0. 10 5.5 0.1 0.0 pop {r3, r4, r5, r6, r7, pc}
# CHECK-NEXT: 1. 10 3.6 0.0 3.9 nop
# CHECK-NEXT: 10 4.6 0.1 2.0 <total>