[core] Don't let DQE eliminate an entry point's last allocation

[mitchdz]

A kernel that allocates qubits but applies no operation to them loses its register entirely. sample() returns no counts at all and get_state() segfaults dereferencing a null SimulationState:

@cudaq.kernel
def allocate_only(n: int):
    q = cudaq.qvector(n)

cudaq.sample(allocate_only, 3)     # {} instead of { 000:1000 }
cudaq.get_state(allocate_only, 3)  # SIGSEGV

DQE erases an alloca whose only use is a dealloc, and a null_wire whose only use is a sink. That is sound when nothing observes the register, but an entry point's register is observable: sample() implicitly measures every allocated qubit and get_state() returns the whole register. Since #4993 enabled value semantics for local simulators as well, such a kernel now reaches the runtime with no qubits at all.

Keep the allocation when it is the last one in a cudaq-entrypoint function. Partially used registers are unaffected, and dead qubits in non-entry-point functions are still eliminated.

Adds a FileCheck test for both halves of that rule and an execution test asserting the observable contract. Both fail without the fix.

[github-actions]

CI Summary (push) — ❌ failed

Run #31724679744 · ✅ 5 · ⏩ 7 · ❌ 1 · ⛔ 0

❌ Failed or cancelled

Job	Result	Link
build_and_test	❌ failure	view

Top-level jobs (13)

Job	Result
binaries	⏩ skipped
build_and_test	❌ failure
config_devdeps	✅ success
config_source_build	⏩ skipped
config_wheeldeps	✅ success
devdeps	✅ success
docker_image	⏩ skipped
gen_code_coverage	⏩ skipped
metadata	✅ success
python_metapackages	⏩ skipped
python_wheels	⏩ skipped
source_build	⏩ skipped
wheeldeps	✅ success

⏩ Skipped jobs (7) — intentionally skipped on PR builds; run on merge_group / workflow_dispatch

Job
binaries
config_source_build
docker_image
gen_code_coverage
python_metapackages
python_wheels
source_build

All sub-jobs (42) — every matrix leg, with links

Job	Status	Link
Build and test (amd64, gcc12, openmpi) / Dev environment (Debug)	❌ failure	view
Build and test (amd64, gcc12, openmpi) / Dev environment (Python)	❌ failure	view
Build and test (amd64, llvm, openmpi) / Dev environment (Debug)	❌ failure	view
Build and test (amd64, llvm, openmpi) / Dev environment (Python)	❌ failure	view
Build and test (arm64, llvm, openmpi) / Dev environment (Debug)	❌ failure	view
Build and test (arm64, llvm, openmpi) / Dev environment (Python)	❌ failure	view
CI Summary	❔ in_progress	view
Configure build (devdeps)	✅ success	view
Configure build (source_build)	⏩ skipped	view
Configure build (wheeldeps)	✅ success	view
Create CUDA Quantum installer	⏩ skipped	view
Create Docker images	⏩ skipped	view
Create Python metapackages	⏩ skipped	view
Create Python wheels	⏩ skipped	view
Gen code coverage	⏩ skipped	view
Load dependencies (amd64, gcc12) / Caching	✅ success	view
Load dependencies (amd64, gcc12) / Finalize	✅ success	view
Load dependencies (amd64, gcc12) / Metadata	✅ success	view
Load dependencies (amd64, llvm) / Caching	✅ success	view
Load dependencies (amd64, llvm) / Finalize	✅ success	view
Load dependencies (amd64, llvm) / Metadata	✅ success	view
Load dependencies (arm64, gcc12) / Caching	✅ success	view
Load dependencies (arm64, gcc12) / Finalize	✅ success	view
Load dependencies (arm64, gcc12) / Metadata	✅ success	view
Load dependencies (arm64, llvm) / Caching	✅ success	view
Load dependencies (arm64, llvm) / Finalize	✅ success	view
Load dependencies (arm64, llvm) / Metadata	✅ success	view
Load source build cache	⏩ skipped	view
Load wheel dependencies (amd64, 12.6) / Caching	✅ success	view
Load wheel dependencies (amd64, 12.6) / Finalize	✅ success	view
Load wheel dependencies (amd64, 12.6) / Metadata	✅ success	view
Load wheel dependencies (amd64, 13.0) / Caching	✅ success	view
Load wheel dependencies (amd64, 13.0) / Finalize	✅ success	view
Load wheel dependencies (amd64, 13.0) / Metadata	✅ success	view
Load wheel dependencies (arm64, 12.6) / Caching	✅ success	view
Load wheel dependencies (arm64, 12.6) / Finalize	✅ success	view
Load wheel dependencies (arm64, 12.6) / Metadata	✅ success	view
Load wheel dependencies (arm64, 13.0) / Caching	✅ success	view
Load wheel dependencies (arm64, 13.0) / Finalize	✅ success	view
Load wheel dependencies (arm64, 13.0) / Metadata	✅ success	view
Prepare cache clean-up	✅ success	view
Retrieve PR info	✅ success	view

⚠️ Required checks (0/6) — 6 missing — declared in .github/required-checks.yml for push

Required check	Status	Link
Build and test (amd64, llvm, openmpi) / Dev environment (Debug)	❌ failure	view
Build and test (amd64, llvm, openmpi) / Dev environment (Python)	❌ failure	view
Build and test (arm64, llvm, openmpi) / Dev environment (Debug)	❌ failure	view
Build and test (arm64, llvm, openmpi) / Dev environment (Python)	❌ failure	view
Build and test (amd64, gcc12, openmpi) / Dev environment (Debug)	❌ failure	view
Build and test (amd64, gcc12, openmpi) / Dev environment (Python)	❌ failure	view

[schweitzpgi]

Let's land #5113 first to see if that doesn't fix the issue.

[sacpis]

LGTM. Thanks @mitchdz.

[sacpis]

Seems like we have a test failure

#12 199.5 FAIL: CUDAQ :: Transforms/lambda_variable-2.qke (304 of 465)
#12 199.5 ******************** TEST 'CUDAQ :: Transforms/lambda_variable-2.qke' FAILED ********************
#12 199.5 Exit Code: 1
#12 199.5 
#12 199.5 Command Output (stderr):
#12 199.5 --
#12 199.5 RUN: at line 9: cudaq-opt -lambda-lifting=constant-prop=1 -canonicalize /workspaces/cuda-quantum/cudaq/test/Transforms/lambda_variable-2.qke |    /usr/local/llvm/bin/FileCheck /workspaces/cuda-quantum/cudaq/test/Transforms/lambda_variable-2.qke
#12 199.5 + cudaq-opt -lambda-lifting=constant-prop=1 -canonicalize /workspaces/cuda-quantum/cudaq/test/Transforms/lambda_variable-2.qke
#12 199.5 + /usr/local/llvm/bin/FileCheck /workspaces/cuda-quantum/cudaq/test/Transforms/lambda_variable-2.qke
#12 199.5 RUN: at line 11: cudaq-opt -lambda-lifting=constant-prop=1 -canonicalize /workspaces/cuda-quantum/cudaq/test/Transforms/lambda_variable-2.qke |    cudaq-translate --convert-to=qir | /usr/local/llvm/bin/FileCheck --check-prefix=QIR /workspaces/cuda-quantum/cudaq/test/Transforms/lambda_variable-2.qke
#12 199.5 + cudaq-opt -lambda-lifting=constant-prop=1 -canonicalize /workspaces/cuda-quantum/cudaq/test/Transforms/lambda_variable-2.qke
#12 199.5 + /usr/local/llvm/bin/FileCheck --check-prefix=QIR /workspaces/cuda-quantum/cudaq/test/Transforms/lambda_variable-2.qke
#12 199.5 + cudaq-translate --convert-to=qir
#12 199.5 /workspaces/cuda-quantum/cudaq/test/Transforms/lambda_variable-2.qke:102:14: error: QIR-NEXT: is not on the line after the previous match
#12 199.5 // QIR-NEXT: ret void
#12 199.5              ^
#12 199.5 <stdin>:14:2: note: 'next' match was here
#12 199.5  ret void
#12 199.5  ^
#12 199.5 <stdin>:11:42: note: previous match ended here
#12 199.5 define void @__nvqpp__mlirgen__kernel_a() local_unnamed_addr {
#12 199.5                                          ^
#12 199.5 <stdin>:12:1: note: non-matching line after previous match is here
#12 199.5  %1 = tail call ptr @__quantum__rt__qubit_allocate()
#12 199.5 ^
#12 199.5 
#12 199.5 Input file: <stdin>
#12 199.5 Check file: /workspaces/cuda-quantum/cudaq/test/Transforms/lambda_variable-2.qke
#12 199.5 
#12 199.5 -dump-input=help explains the following input dump.
#12 199.5 
#12 199.5 Input was:
#12 199.5 <<<<<<
#12 199.5           .
#12 199.5           .
#12 199.5           .
#12 199.5           9: } 
#12 199.5          10:  
#12 199.5          11: define void @__nvqpp__mlirgen__kernel_a() local_unnamed_addr { 
#12 199.5          12:  %1 = tail call ptr @__quantum__rt__qubit_allocate() 
#12 199.5          13:  tail call void @__quantum__rt__qubit_release(ptr %1) 
#12 199.5          14:  ret void 
#12 199.5 next:102      !~~~~~~~  error: match on wrong line
#12 199.5          15: } 
#12 199.5          16:  
#12 199.5          17: declare ptr @__quantum__rt__qubit_allocate() local_unnamed_addr 
#12 199.5          18:  
#12 199.5          19: declare void @__quantum__rt__qubit_release(ptr) local_unnamed_addr 
#12 199.5           .
#12 199.5           .
#12 199.5           .
#12 199.5 >>>>>>
#12 199.5 
#12 199.5 --
#12 199.5 
#12 199.5 ********************

Seems like the test failure is a correct behavior. kernel_a is an entry point whose only qubit allocation is unused. Before this change, DQE deleted it and the QIR body was just ret void. With the new guard, the allocation is kept, so the QIR now has an allocate/release pair before the return. The test's expectation is written against the old behavior.
Let's update this test.

[mitchdz]

It looks like #5113 fixed this issue.