Add a regression test for mlir-stage-13-v1

This adds one LIT test covering 6 lines and 4 branch outcomes that the existing suite does not reach, across 1 source file, at 48615bc5925ef4b9db8b4550b5d4322933cf4b7b.

Coverage this test adds

Source file	Newly covered lines	Newly covered branch outcomes
lib/Frontend/Lowering/LowerGraphMemoryPass.cpp	459–460, 736, 739–741	459:3–459:45 true; 476:43–476:65 true; 735:11–735:27 true; 738:34–738:35 true

lib/Frontend/Lowering/LowerGraphMemoryPass.cpp

 457    case ::mlir::LLVM::AtomicOrdering::monotonic: 458      return ::dataflow::AtomicOrdering::Monotonic;+459    case ::mlir::LLVM::AtomicOrdering::acquire:    [true branch at 459:3]+460      return ::dataflow::AtomicOrdering::Acquire; 461    case ::mlir::LLVM::AtomicOrdering::release: 462      return ::dataflow::AtomicOrdering::Release;@@ 474  convertSyncScope(::mlir::MLIRContext *context, 475                   std::optional<::llvm::StringRef> syncscope) {+476    if (!syncscope || syncscope->empty() || *syncscope == "system")    [true branch at 476:43] 477      return ::dataflow::SyncScopeRefAttr::get( 478          context, ::dataflow::SyncScopeKind::System);@@ 733      ::llvm::SmallVector<::mlir::LLVM::GEPOp, 8> dead; 734      graph.getBody().walk([&](::mlir::LLVM::GEPOp gep) {+735        if (gep->use_empty())    [true branch at 735:11]+736          dead.push_back(gep); 737      });+738      for (::mlir::LLVM::GEPOp gep : dead) {    [true branch at 738:34]+739        gep.erase();+740        changed = true;+741      } 742    } 743  }

The test

The input was reduced from 20 to 18 lines while keeping every covered line and branch outcome above, and while the compiler still accepted it. The CHECK lines are the compiler's exact output on this input at 48615bc5925ef4b9db8b4550b5d4322933cf4b7b.

// RUN: loom-raise-opt --loom-lower-graph-memory --mlir-print-op-generic %s | FileCheck %s

module attributes {
} {
  dataflow.graph private @source_memory_graph(
      %start: none, %base: !llvm.ptr, %expected: i32, %desired: i32,
      %index: i64, %cond: i1) -> ()
      attributes {input_segments = array<i32: 5, 0, 0>,
                  result_segments = array<i32: 0, 0, 0>} {
    %ptr = llvm.getelementptr inbounds %base[%index]
        : (!llvm.ptr, i64) -> !llvm.ptr, !llvm.array<4 x i8>
    %lower = arith.constant 0 : index
    %upper = arith.constant 4 : index
    %step = arith.constant 1 : index
    scf.for %iv = %lower to %upper step %step {
    llvm.fence syncscope("system") acquire
    }
    dataflow.graph.return %start : none
  }
}

// CHECK: "builtin.module"() ({
// CHECK-NEXT:   "dataflow.graph"() <{function_type = (!llvm.ptr, i32, i32, i64, i1, memref<?xi8>) -> (), input_segments = array<i32: 5, 0, 1>, result_segments = array<i32: 0, 0, 0>, sym_name = "source_memory_graph", sym_visibility = "private"}> ({
// CHECK-NEXT:   ^bb0(%arg0: none, %arg1: !llvm.ptr, %arg2: i32, %arg3: i32, %arg4: i64, %arg5: i1, %arg6: memref<?xi8>):
// CHECK-NEXT:     %0 = "arith.constant"() <{value = 0 : index}> : () -> index
// CHECK-NEXT:     %1 = "arith.constant"() <{value = 4 : index}> : () -> index
// CHECK-NEXT:     %2 = "arith.constant"() <{value = 1 : index}> : () -> index
// CHECK-NEXT:     %3 = "arith.index_cast"(%0) : (index) -> i32
// CHECK-NEXT:     %4 = "arith.index_cast"(%1) : (index) -> i32
// CHECK-NEXT:     %5 = "arith.index_cast"(%2) : (index) -> i32
// CHECK-NEXT:     %6:2 = "dataflow.stream"(%3, %4, %5) <{predicate = 2 : i64, step_kind = 0 : i32}> : (i32, i32, i32) -> (i32, i1)
// CHECK-NEXT:     %7 = "dataflow.carry"(%6#1, %arg0, %9) : (i1, none, none) -> none
// CHECK-NEXT:     %8:2 = "dataflow.demux"(%6#1, %7) : (i1, none) -> (none, none)
// CHECK-NEXT:     %9 = "dataflow.fence"(%8#1) <{contract = #dataflow.fence_contract<ordering = acquire, sync_scope = <system>>}> : (none) -> none
// CHECK-NEXT:     "dataflow.graph.return"(%8#0) <{operandSegmentSizes = array<i32: 0, 0, 0, 1>}> : (none) -> ()
// CHECK-NEXT:   }) : () -> ()
// CHECK-NEXT: }) : () -> ()
// CHECK-EMPTY:

Validation

• Native compiler and FileCheck replay: passed at 48615bc5925ef4b9db8b4550b5d4322933cf4b7b.
• Property checker verdict on the reduced pair: PASS (postcondition.spct, SHA-256 840ba12c591c37fe0a0bf28ef27f6f8a496505d00629390bb6209f57e770e9bb).
• Reduction criterion: accepted checker PASS and every original line/branch increment over the suite baseline.

Where this test came from

PBT mlir-stage-13-v1, run 20260911-083853, seed 31.

The property under test is anchored on documentation:

• selected output: docs/spec-compiler-part-3-mem.md lines 482–486 (output side)
• linked input 29: docs/spec-compiler-part-3-mem.md lines 482–486 (input side)
• linked input 32: docs/spec-compiler-part-3-mem.md lines 92–97 (input side)
• linked input 50: docs/spec-compiler-part-3-mem.md lines 148–152 (input side)
• linked input 59: docs/spec-compiler-part-3-mem.md lines 136–140 (input side)
• linked input 66: docs/spec-compiler-part-3-mem.md lines 489–493 (input side)
• linked input 67: docs/spec-compiler-part-3-mem.md lines 3–6 (input side)
• linked input 68: docs/spec-compiler-part-3-mem.md lines 79–90 (input side)
• linked input 80: docs/spec-compiler-part-3-mem.md lines 99–106 (input side)
• linked input 102: docs/spec-compiler-part-3-mem.md lines 389–393 (input side)
• linked input 104: docs/spec-compiler-part-3-mem.md lines 470–480 (input side)
• linked input 115: docs/spec-compiler-part-3-mem.md lines 486–489 (input side)
• linked input 126: docs/spec-compiler-part-3-mem.md lines 142–146 (input side)
• linked input 149: docs/spec-compiler-part-3-mem.md lines 43–46 (input side)
• linked input 170: docs/spec-compiler-part-3-mem.md lines 154–157 (input side)
• linked input 182: docs/spec-compiler-part-3-mem.md lines 382–387 (input side)
• linked input 184: docs/spec-compiler-part-3-mem.md lines 34–35 (input side)
• linked input 195: docs/spec-compiler-part-3-mem.md lines 159–161 (input side)
• linked input 200: docs/spec-compiler-part-3-mem.md lines 40–41 (input side)
• linked input 202: docs/spec-compiler-part-3-mem.md lines 31–32 (input side)
• linked input 211: docs/spec-compiler-part-3-mem.md lines 245–251 (input side)

---

Coverage is measured against the recorded baseline suite at this revision, and counts unique mapped file locations including generated code. Reduction retains every added line or branch outcome and a passing property check; it is not a claim of global minimality, nor of a defect.