llvm-project/llvm/test/CodeGen/Hexagon/swp-alias-cross-iteration.mir
Ryotaro Kasuga 3cd6b86cc1
[MachinePipeliner] Use AliasAnalysis properly when analyzing loop-carried dependencies (#136691)
MachinePipeliner uses AliasAnalysis to collect loop-carried memory
dependencies. To analyze loop-carried dependencies, we need to
explicitly tell AliasAnalysis that the values may come from different
iterations. Before this patch, MachinePipeliner didn't do this, so some
loop-carried dependencies might be missed. For example, in the following
case, there is a loop-carried dependency from the load to the store, but
it wasn't considered.

```
def @f(ptr noalias %p0, ptr noalias %p1) {
entry:
  br label %body

loop:
  %idx0 = phi ptr [ %p0, %entry ], [ %p1, %body ]
  %idx1 = phi ptr [ %p1, %entry ], [ %p0, %body ]
  %v0 = load %idx0
  ...
  store %v1, %idx1
  ...
}
```

Further, the handling of the underlying objects was not sound. If there
is no information about memory operands (i.e., `memoperands()` is
empty), it must be handled conservatively. However, Machinepipeliner
uses a dummy value (namely `UnknownValue`). It is distinguished from
other "known" objects, causing necessary dependencies to be missed.
(NOTE: in such cases, `buildSchedGraph` adds non-loop-carried
dependencies correctly, so perhaps a critical problem has not occurred.)

This patch fixes the above problems. This change has increased false
dependencies that didn't exist before. Therefore, this patch also
introduces additional alias checks with the underlying objects.

Split off from #135148
2025-04-23 18:11:34 +09:00

73 lines
2.5 KiB
YAML

# RUN: llc -mtriple=hexagon -run-pass pipeliner -debug-only=pipeliner %s -o /dev/null 2>&1 | FileCheck %s
# REQUIRES: asserts
# Test that pipeliner correctly detects the loop-carried dependency between the
# load and the store, which is indicated by `Ord` dependency from SU(2) to
# SU(4). Note that there is no dependency within a single iteration.
# CHECK: SU(2): %7:intregs = L2_loadri_io %5:intregs, 0 :: (load (s32) from %ir.ptr.load)
# CHECK-NEXT: # preds left
# CHECK-NEXT: # succs left
# CHECK-NEXT: # rdefs left
# CHECK-NEXT: Latency
# CHECK-NEXT: Depth
# CHECK-NEXT: Height
# CHECK-NEXT: Predecessors:
# CHECK-NEXT: SU(0): Data Latency=0 Reg=%5
# CHECK-NEXT: Successors:
# CHECK-DAG: SU(3): Data Latency=2 Reg=%7
# CHECK-DAG: SU(4): Ord Latency=1 Barrier
# CHECK-NEXT: SU(3): %8:intregs = F2_sfadd %7:intregs, %3:intregs, implicit $usr
# CHECK: SU(4): S2_storeri_io %6:intregs, 0, %8:intregs :: (store (s32) into %ir.ptr.store)
--- |
define void @foo(ptr noalias %p0, ptr noalias %p1, i32 %n) {
entry:
br label %body
body: ; preds = %body, %entry
%i = phi i32 [ 0, %entry ], [ %i.next, %body ]
%ptr.load = phi ptr [ %p0, %entry ], [ %p1, %body ]
%ptr.store = phi ptr [ %p1, %entry ], [ %p0, %body ]
%v = load float, ptr %ptr.load, align 4
%add = fadd float %v, 1.000000e+00
store float %add, ptr %ptr.store, align 4
%i.next = add i32 %i, 1
%cond = icmp slt i32 %i.next, %n
br i1 %cond, label %body, label %exit
exit: ; preds = %body
ret void
}
...
---
name: foo
tracksRegLiveness: true
body: |
bb.0.entry:
successors: %bb.1(0x80000000)
liveins: $r0, $r1, $r2
%6:intregs = COPY $r2
%5:intregs = COPY $r1
%4:intregs = COPY $r0
%9:intregs = A2_tfrsi 1065353216
%12:intregs = COPY %6
J2_loop0r %bb.1, %12, implicit-def $lc0, implicit-def $sa0, implicit-def $usr
bb.1.body (machine-block-address-taken):
successors: %bb.1(0x7c000000), %bb.2(0x04000000)
%1:intregs = PHI %4, %bb.0, %5, %bb.1
%2:intregs = PHI %5, %bb.0, %4, %bb.1
%8:intregs = L2_loadri_io %1, 0 :: (load (s32) from %ir.ptr.load)
%10:intregs = F2_sfadd killed %8, %9, implicit $usr
S2_storeri_io %2, 0, killed %10 :: (store (s32) into %ir.ptr.store)
ENDLOOP0 %bb.1, implicit-def $pc, implicit-def $lc0, implicit $sa0, implicit $lc0
J2_jump %bb.2, implicit-def dead $pc
bb.2.exit:
PS_jmpret $r31, implicit-def dead $pc
...