Skip to content

Research Preview: cudaq-pulse (reopened) - #5244

Open
anthony-santana wants to merge 50 commits into
mainfrom
research_preview/pulse
Open

Research Preview: cudaq-pulse (reopened)#5244
anthony-santana wants to merge 50 commits into
mainfrom
research_preview/pulse

Conversation

@anthony-santana

@anthony-santana anthony-santana commented Aug 24, 2026

Copy link
Copy Markdown
Collaborator

Reopening this PR as a miscommunication let it get removed after merging the first time.

This PR provides a research preview of the CUDA-Q Pulse MLIR dialect, as well as auxillary infrastructure for writing, scheduling, and lowering pulse kernels from Python to GPU simulation via cuDynamics.

What this PR introduces / supports

  • End-to-end GPU pulse simulation: Python pulse-kernel DSL → MLIR (Pulse/QOp/CuDensityMat dialects) → cuDensityMat density-matrix evolution on GPU.
  • Integrator set exposed to the pulse frontend: magnus and crank_nicolson.
  • Advanced cudaq.dynamics integrators (C++/CUDA), validated on GPU: dopri5 — Dormand–Prince RK5(4) adaptive integrator. magnus_cf4 — high-order commutator-free Magnus integrator with propagator cache (uses cuSOLVER/cuBLAS for the matrix exponential).
  • Allocation-driven register sizing: the simulated Hilbert space is defined by the qudits a kernel allocates (its arguments), giving predictable 2^N state shapes; allocated-but-idle qudits stay in |0⟩.
  • Kernel DSL builtins: get_drive_line, square, gaussian, drag, wait, sync, shift_phase, drive.
  • Device model API: Target / Qubit / Coupling topology definition.
  • Test coverage: Bell, Rabi, T1 decay, full-pipeline, multi-qubit selective drive, 8-qubit excitation, plus dopri5/magnus_cf4 C++ unit tests (GPU).
  • Parameterized kernels via runtime args — drive-operator parameters/durations and waveform parameters passed as runtime arguments (in progress; next on the roadmap).

What is not yet supported

  • Custom user-defined simulation target hamiltonians/topologies. Current simulation targets have fixed hamiltonian structures.
  • Observable/expectation evaluation from the pulse evolve path (raises NotImplementedError).
  • Integrators beyond magnus / crank_nicolson at the pulse frontend (the broader cudaq.dynamics set is available directly).
  • Hardware execution / calibration runtime (tracked as higher-order feature work).

Requirements

  • GPU + CUDA toolkit with cuDensityMat (cuQuantum); magnus_cf4 additionally requires the cuSOLVER dev package.

Getting started & docs

  • Roadmap and current status: PULSE_ROADMAP.md
  • Pulse README and examples: pulse/README.md

anthony-santana and others added 30 commits July 17, 2026 14:49
@anthony-santana @amccaskey
hello world
c39cca9
@anthony-santana @amccaskey
Adopt CUDA-Q pulse research-preview integration (from amccaskey/719104e)
89bcf3b 
Introduce the pulse package at the repository root as explicitly opt-in,
non-production research software. Keep it out of CUDAQ_ALL_PROJECTS while
supporting both pulse-only and combined builds through
CUDAQ_ENABLE_PROJECTS. This supersedes the initial experimental/pulse drop
by relocating the package to pulse/ and adopting the hardened integration.

- Add the Python frontend, pulse kernel compiler, MLIR dialects, lowering
  pipelines, nanobind bindings, targets, visualization, and cuDensityMat
  runtime.
- Integrate CUDA and cuDensityMat discovery with automatic GPU-test
  availability detection and numerical validation for driven evolution,
  relaxation, coupling, and the public compile/JIT path.
- Add unit, property, semantic, workload, and MLIR regression suites together
  with examples, benchmarks, API documentation, source-build guidance, and a
  turnkey development container.
- Add scoped CPU, documentation, license, and trusted GPU CI using CUDA-Q's
  prebuilt LLVM environment instead of rebuilding LLVM.
- Generalize CI routing through a central research-preview registry so future
  previews can avoid stable Linux and macOS builds, CodeQL, and spelling while
  retaining formatting and DCO checks.
- Apply CUDA-Q licensing and formatting throughout and document that APIs,
  behavior, and numerical support may evolve without compatibility guarantees.

Co-authored-by: Alex McCaskey <amccaskey@nvidia.com>
Signed-off-by: A.M. Santana <anthonys@nvidia.com>
@anthony-santana
[dynamics] Add Dormand-Prince RK5(4) adaptive integrator (dopri5)
4b577be 
Port the adaptive-timestep integrator from the experimental pulse GPU
runtime into the mainlined cuDensityMat integrator family. The integrator
uses the embedded Dormand-Prince RK5(4) pair on top of the existing
CuDensityMatTimeStepper, with an error-controlled step-size selector suited
to stiff, high-frequency driven (transmon) dynamics.

- Declare cudaq::integrators::dopri5 in cudaq/algorithms/integrator.h and
  implement it in runtime/nvqir/cudensitymat/AdaptiveIntegrator.cpp using the
  shared CuDensityMatIntegratorHelper (setState/getState/ensureStepper).
- Expose it through nanobind (integrators.dopri5) and add a DoPri5Integrator
  Python wrapper alongside RungeKuttaIntegrator.
- Add unittests/dynamics/test_AdaptiveIntegrator.cpp covering Rabi accuracy,
  adaptive-step statistics, parity with runge_kutta(order=4), clone
  reproducibility, and open-system T1 relaxation.

Signed-off-by: A.M. Santana <anthonys@nvidia.com>
@anthony-santana
[dynamics] Add high-order Magnus integrator with propagator cache (CUDA)
2a16d8d 
Port the performance-oriented, commutator-free Magnus path from the
experimental pulse GPU runtime as a new add-on integrator alongside the
existing simple midpoint magnus_expansion.

- Declare cudaq::integrators::magnus_cf4 in cudaq/algorithms/integrator.h
  (PImpl so the header stays CUDA-free) and implement it in
  runtime/nvqir/cudensitymat/MagnusHighOrderIntegrator.cpp. For closed-system
  density-matrix evolution it materializes the dense Hamiltonian at the two
  Gauss-Legendre nodes (CF4, Blanes-Moan), forms exact unitary propagators via
  a GPU matrix exponential, and applies rho <- U rho U^dagger. Propagators are
  cached (LRU) keyed by a quantized Hamiltonian signature so repeated PWC
  slices reuse a single matrix exponential.
- For open systems (collapse operators / super-operator) or state-vector
  evolution it transparently falls back to magnus_expansion for parity.
- Add shared support infra under runtime/nvqir/cudensitymat/support/:
  matrix_exp.{h,cu} (scaling-and-squaring Pade[13/13]), propagator_cache.h,
  hamiltonian_cache.h, and trimmed cuda_check.h / cuda_memory.h. matrix_exp.cu
  compiles as CUDA; the library now links CUDA::cublas and CUDA::cusolver.
- Expose it through nanobind (integrators.magnus_cf4) and add a
  MagnusCF4Integrator Python wrapper.
- Add unittests/dynamics/test_MagnusHighOrderIntegrator.cpp covering Rabi
  accuracy + trace preservation, propagator-cache reuse, parity with
  magnus_expansion, clone reproducibility, and open-system T1 fallback.

Signed-off-by: A.M. Santana <anthonys@nvidia.com>
@anthony-santana
[pulse] Add magnus and crank_nicolson integrators to the GPU runtime
82602b9 
Expose the magnus (Taylor-series midpoint) and crank_nicolson
(predictor-corrector) integrators through the pulse frontend, mirroring
the mainlined cudaq::integrators algorithms of the same name. Both are
driven through the existing cuDensityMat Liouvillian-action primitives in
cudm-runtime, so the RK4 pulse-boundary sampling fix is factored into a
shared boundarySafeSampleTime helper and reused by crank_nicolson.

The adaptive dopri5 and dense-Hamiltonian magnus_cf4 integrators are not
expressible through the operator-action C ABI (no dense H, no adaptive
error control at this layer) and remain available via the cudaq.dynamics
API only; cudm_evolve rejects them explicitly.

- Add IntegratorKind::Magnus (5) and CrankNicolson (6) dialect cases.
- Whitelist "magnus"/"crank_nicolson" in evolve.py with updated docstring.
- Add cudm-to-llvm lit coverage and pytest coverage (MLIR emission +
  GPU Rabi parity across rk4/magnus/crank_nicolson).
@anthony-santana
Port GPU physics regressions as pulse runtime tests
ebc1b91 
Translate the research-preview qpu/physics C++ engine tests into
Python regressions under pulse/tests/runtime, exercising the
dialect-routed cuDensityMat runtime through the pulse frontend:

- decoherence: T1 damping, monotonicity, ground-state fixed point,
  trace preservation, independent multi-qubit decay
- idle_evolution: exponential T1 decay across idle sweeps, split-idle
  equivalence, minimal short-idle perturbation
- 8qubit: ladder connectivity/degree structure + single-excitation
  evolution in the full 256-dim register
- physics_validation: free drift, Rabi pi/2 and pi pulses, selective
  two-qubit drive, XX-coupling excitation transfer
- frame_operations: phase-shift no-ops, opposite-phase cancellation,
  same-phase composition
- iq_modulation: pure-I (X) and pure-Q (Y) rotations, IQ symmetry and
  X/Y orthogonality
- quantum_algorithms: single-qubit X/Y rotations, half rotations,
  gate-sequence identity; Bell/GHZ kept as MLIR-structure checks

Multi-qubit population assertions are written to be independent of the
state-vector qubit-ordering convention. Files are collected
automatically by the existing directory-based pytest runners
(check-pulse for CPU structure checks, check-pulse-gpu for GPU cases).

The offline Google surface-code decoder-analysis script is intentionally
not ported: it needs external experiment data plus stim/pymatching and
is not a pulse-runtime example.
@anthony-santana
Document new pulse GPU integrators and expanded test coverage
2e14d56 
Update the research-preview docs to reflect the magnus and
crank_nicolson integrators now available on the pulse cuDensityMat
runtime path:

- gpu_execution.rst: describe all five frontend integrators
  (rk1/rk2/rk4/magnus/crank_nicolson), note that the adaptive dopri5
  and dense-Hamiltonian magnus_cf4 schemes live in cudaq.dynamics, and
  expand the GPU test-coverage summary to include the new physics
  regressions.
- README.md: update the integrator list in the limitations section.

The api/evolve.rst page is autodoc-driven and already reflects the
updated evolve() docstring. The pulse.yml GPU job runs check-pulse-gpu
(ctest -L gpu -> pytest tests/runtime -m gpu), which collects the new
GPU regressions by directory, so no workflow change is required.
@anthony-santana
[pulse] Size the simulated register by qudit allocation
b541d18 
The GPU register is now sized by the qudits a kernel allocates (its
arguments), not merely the ones it drives. N allocated qudits yield a
predictable 2**N register with idle qudits held in |0>. Adds an 8-qubit
ladder regression exercising a single-qubit excitation in the full
256-dimensional space.
@anthony-santana
[pulse] Clean up GPU port: formatting, dead code, and dynamics test f…
388eb0f 
…ixes

- Fix an operators-API portability bug in the dopri5/magnus_cf4 unit
  tests: matrix_handler::number(0) has no scalar operator*, so use the
  op-returning factory sum_op<matrix_handler>::number(0). The tests now
  compile and pass in a full cudaq.dynamics build (11 GPU tests).
- Remove the unused support/hamiltonian_cache.h (never wired into the
  Magnus path, which uses the propagator cache).
- Drop an unconditional std::cout (folded into the gtest failure
  message) and the now-unused <iostream>.
- De-duplicate set_system in builtin_integrators.py via a shared
  _build_system_and_schedule helper for the ported integrators.
- clang-format (v22) across authored C++; yapf across Python.
@anthony-santana
[pulse] Fail early with a clear message when cuSOLVER is missing
6ef6939 
The high-order Magnus integrator (magnus_cf4) links cuBLAS + cuSOLVER for
its GPU dense matrix-exponential path. Both ship with the CUDA Toolkit, but
minimal/runtime-only toolkit layouts can omit the cuSOLVER dev component,
which otherwise surfaces as a cryptic "target CUDA::cusolver not found" at
CMake generate time. Emit an actionable diagnostic instead.
@anthony-santana
Merge branch 'main' into research_preview/pulse
58c97a3
@anthony-santana
[pulse] Remove benchmarks folder
d2fa250 
The compiler/simulation benchmark scripts are development-only tooling and
are not needed in the merged research preview.
@anthony-santana
Merge remote-tracking branch 'github/main' into research_preview/pulse
42696d1 
# Conflicts:
#	CMakeLists.txt
@anthony-santana
[pulse] Split dynamics integrators out of core; retain non-public copy
c1cb2ab 
Per reviewer feedback (#4834), the dopri5 / magnus_cf4 GPU integrators are
being contributed to CUDA-Q core as their own PR for targeted review. To keep
the experimental pulse stack free of cross-dependencies on core cudaq (so it
can be versioned and, if needed, removed as one atomic unit), this reverts all
core-directory changes back to main and retains a non-public, source-only copy
of the integrator sources under pulse/core/runtime/cudm/integrators/ (not wired
into any build target). The pulse frontend continues to expose only rk1/rk2/rk4/
magnus/crank_nicolson; dopri5/magnus_cf4 remain rejected there. Public docs no
longer advertise them as pulse features.
@anthony-santana
[pulse] Package pulse as an installable cudaq-pulse CMake project
6f902a6 
Mirror the cudaq-realtime packaging so pulse builds either from the source tree
or as a standalone project consumable via find_package(cudaq-pulse CONFIG):

- Root CMake: fold 'pulse' into CUDAQ_ALL_PROJECTS (drop the separate
  CUDAQ_RESEARCH_PROJECTS variable) and add a CUDAQ_PULSE_DIR installed-package
  path mirroring CUDAQ_REALTIME_DIR. Keeps naming clean (no 'experimental'
  token) so promotion to first-class needs no renames.
- pulse/CMakeLists.txt: detect standalone vs integrated build, define
  project()/build-type/install-prefix when standalone, and emit + install a
  cudaq-pulse package config plus the cudaq-pulse-targets export set.
- Export cudaq::cudm_runtime through the package; the config re-discovers
  cuDensityMat/CUDAToolkit via find_dependency.
- Document that CudaqPulseDependencies.cmake is build-time discovery, distinct
  from the installed cudaq-pulse-config.cmake.in.
@anthony-santana
[pulse] Address PR review: CI path source-of-truth, canonicalize CAPI
c91271f 
- research-preview-paths.txt: rewrite the header to describe it accurately as
  an ignore-list consumed by the stable CI workflows via paths-ignore-file, and
  as the single source of truth for a preview's scoped workflow.
- pulse.yml: derive the authoritative changed-path filter in the 'changes' job
  from research-preview-paths.txt (plus the shared build files) instead of
  duplicating the prefix list; note the trigger-level paths are a coarse
  pre-gate only.
- PulseCanonicalize.cpp: the greedy driver ran with an always-empty pattern
  set (a no-op). Populate it with the canonicalization patterns every loaded op
  registers (the same collection the built-in canonicalizer performs) so pulse
  ops with hasCanonicalizer actually fold.
- bindings.cpp: document the PulseModule/PassManager simplification as a tracked
  follow-up (deep change threading through the Python frontend).
@anthony-santana
[pulse] Fix cudaq-pulse export: use build/install interface include dirs
9e96b35 
install(EXPORT cudaq-pulse-targets) rejects a target whose
INTERFACE_INCLUDE_DIRECTORIES points at a raw source path. Wrap the
cudm_runtime include dir in $<BUILD_INTERFACE:>/$<INSTALL_INTERFACE:>
generator expressions and install the public C ABI header so the
installed package resolves it.
@anthony-santana
[pulse] Rename relocated integrator support namespace to cudaq::detail
b74788d 
Addresses review feedback on the (non-public, source-only) relocated
integrator copy: use a single cudaq::detail namespace for the CUDA /
cuBLAS / cuSOLVER helpers instead of cudaq::pulse and the internal
__dynamics_support name. These files are not built or linked by pulse;
the built integrators live in the separate cudaq.dynamics core PR.
@anthony-santana
[pulse] Fix CI content checks: license ignores, yapf, spelling
d968946 
- Ignore py.typed (PEP 561 marker) and pyrightconfig.json (JSON config)
  in .licenserc.yaml; neither can carry a comment license header, and a
  header on the JSON would break the check-json hook.
- Drop the stale `benchmarks` execution environment from
  pulse/pyrightconfig.json (that folder was removed).
- Reformat pulse runtime tests with yapf (Google style).
- Add new pulse domain terms to the spelling allowlist (LC_ALL=C sorted).

Signed-off-by: A.M. Santana <anthonys@nvidia.com>
@anthony-santana
Merge remote-tracking branch 'github/main' into research_preview/pulse
3f37697 
Signed-off-by: A.M. Santana <anthonys@nvidia.com>
@anthony-santana
DCO remediation commit for previously unsigned pulse commits
05ba52b 
I, A.M. Santana <anthonys@nvidia.com>, hereby add my Signed-off-by to this commit: b74788d
I, A.M. Santana <anthonys@nvidia.com>, hereby add my Signed-off-by to this commit: 9e96b35
I, A.M. Santana <anthonys@nvidia.com>, hereby add my Signed-off-by to this commit: c91271f
I, A.M. Santana <anthonys@nvidia.com>, hereby add my Signed-off-by to this commit: 6f902a6
I, A.M. Santana <anthonys@nvidia.com>, hereby add my Signed-off-by to this commit: c1cb2ab
I, A.M. Santana <anthonys@nvidia.com>, hereby add my Signed-off-by to this commit: d2fa250
I, A.M. Santana <anthonys@nvidia.com>, hereby add my Signed-off-by to this commit: 6ef6939
I, A.M. Santana <anthonys@nvidia.com>, hereby add my Signed-off-by to this commit: 388eb0f
I, A.M. Santana <anthonys@nvidia.com>, hereby add my Signed-off-by to this commit: b541d18
I, A.M. Santana <anthonys@nvidia.com>, hereby add my Signed-off-by to this commit: 2e14d56
I, A.M. Santana <anthonys@nvidia.com>, hereby add my Signed-off-by to this commit: ebc1b91
I, A.M. Santana <anthonys@nvidia.com>, hereby add my Signed-off-by to this commit: 82602b9
I, A.M. Santana <anthonys@nvidia.com>, hereby add my Signed-off-by to this commit: c39cca9

Signed-off-by: A.M. Santana <anthonys@nvidia.com>
@anthony-santana
[pulse] CI: authenticate to ghcr.io for the devcontainer image
92e9f09 
The test-and-docs and gpu-validation jobs pull the private NVIDIA
cuda-quantum-devcontainer image but never authenticated, so the pull
failed with "manifest unknown". Grant `packages: read` and supply
container credentials (github.actor / github.token), mirroring the
pattern already used by integration_tests.yml and python_metapackages.yml.

Signed-off-by: A.M. Santana <anthonys@nvidia.com>
@anthony-santana
[pulse] CI: allow a read:packages PAT for the devcontainer pull
9e521da 
GitHub auto-authenticates ghcr.io container jobs with GITHUB_TOKEN, which
is not authorized for the cross-repo cuda-quantum-devcontainer package and
returns "manifest unknown" even though the image is anonymously pullable.
Prefer an optional GHCR_PULL_TOKEN secret (read:packages) for the pull and
fall back to GITHUB_TOKEN so the workflow works once the secret is added or
the package grants this repository read access.

Signed-off-by: A.M. Santana <anthonys@nvidia.com>
@anthony-santana
[pulse] CI: use canonical GITHUB_TOKEN ghcr credentials
29d9460 
Match every other CUDA-Q ghcr consumer by authenticating the devcontainer
container pull with github.actor / github.token (plus packages: read),
rather than a bespoke PAT secret. The cuda-quantum-devcontainer package is
built and pushed from this repo by docker_images.yml, so the remaining
requirement is that the package grant NVIDIA/cuda-quantum read access.

Signed-off-by: A.M. Santana <anthonys@nvidia.com>
@anthony-santana
[pulse] CI: pull the public devcontainer image anonymously via docker…
a2ca6e4 
… run

The job-level `container:` made GitHub auto-authenticate to ghcr.io with
GITHUB_TOKEN, which is not authorized for the cuda-quantum-devcontainer
package and failed with "manifest unknown" even though the image is public.
Drop `container:` and instead `docker pull` the image (anonymous, no creds)
and run the build/test/docs inside it with `docker run`. This makes the
pulse build self-contained: no PAT secret and no package-access change.

Signed-off-by: A.M. Santana <anthonys@nvidia.com>
@anthony-santana
[pulse] CI: revert to container: with forward-compatible ghcr credential
a04568f 
The cuda-quantum-devcontainer image is a PRIVATE ghcr.io package, so the
anonymous docker-run approach cannot fetch its blobs (child manifest 404s
without package-read). Restore the canonical job-level container: and pull
with `secrets.GHCR_PULL_TOKEN || github.token` so the build resolves as soon
as either a read:packages PAT secret is added or the package grants this repo
read access under "Manage Actions access".

Signed-off-by: A.M. Santana <anthonys@nvidia.com>
@anthony-santana
[pulse] CI: use the correct public devcontainer tag
8e6dcb6 
The pulse jobs referenced ghcr.io/nvidia/cuda-quantum-devcontainer:
amd64-cu12.6-llvm-main, which failed with "manifest unknown": the *-llvm-*
tags belong to the cuda-quantum-devdeps image, and the amd64-/arm64- prefixed
tags are per-arch build/cache artifacts whose child blobs get cleaned up. Use
the canonical multi-arch devcontainer tag cu12.6-gcc12-main (public, intact)
and pull it with the standard GITHUB_TOKEN credentials.

Signed-off-by: A.M. Santana <anthonys@nvidia.com>
@anthony-santana
[pulse] CI: mark workspace safe before submodule init
d2621e6 
The devcontainer runs as a non-root user while the checked-out workspace is
owned by the runner UID, so the manual `git submodule update` step failed with
"detected dubious ownership". actions/checkout only scopes its safe.directory
exception to its own step, so add a global safe.directory for the workspace
before initializing submodules (persists for later build steps in the job).

Signed-off-by: A.M. Santana <anthonys@nvidia.com>
@anthony-santana
[pulse] Sync relocated integrator snapshot with cleaned core copy
d8572ae 
Match the non-public integrator snapshot to the versions being contributed in
the standalone integrators PR, addressing the review comments in place:
consolidate the CUDA/cuBLAS/cuSOLVER error checks into a single cuda_check.h
(cuda_memory.h now only holds the RAII device-memory wrapper), and collapse
the nested cudaq/integrators namespaces to cudaq::integrators.

Signed-off-by: A.M. Santana <anthonys@nvidia.com>
@anthony-santana
Merge branch 'main' into research_preview/pulse
603ef89
@github-actions github-actions Bot added runtime CUDA quantum runtime python-lang Anything related to the Python CUDA Quantum language implementation python bridge Involves the python bridge to quake driver (nvq++) The nvq++ compiler driver c++ bridge CUDA-Q in C++ bridge to MLIR compiler kernel builder Relating to the dynamic kernel builder in C++ or Python codegen issue with generating any codegen output (QIR, OpenQASM etc.) simulation trouble getting code to run on one of the simulators core compiler dynamics Related to dynamics simulation realtime Related to CUDA-Q realtime labels Aug 24, 2026
anthony-santana and others added 2 commits August 24, 2026 17:57
@anthony-santana
Merge remote-tracking branch 'github/main' into research_preview/pulse
7291c1d 
Signed-off-by: A.M. Santana <anthonys@nvidia.com>
@lmondada @anthony-santana
DCO remediation commit for b1824f4
52d463f 
I, Luca Mondada <luca@mondada.net>, hereby add my Signed-off-by to this commit: b1824f4

Signed-off-by: Luca Mondada <luca@mondada.net>
@anthony-santana

anthony-santana commented Aug 25, 2026 •
edited by github-actions Bot
Loading

Copy link
Copy Markdown
Collaborator Author

/ok to test 52d463f

Command Bot: Processing...

@anthony-santana anthony-santana added runtime CUDA quantum runtime and removed documentation Improvements or additions to documentation runtime CUDA quantum runtime python bridge Involves the python bridge to quake driver (nvq++) The nvq++ compiler driver c++ bridge CUDA-Q in C++ bridge to MLIR compiler kernel builder Relating to the dynamic kernel builder in C++ or Python codegen issue with generating any codegen output (QIR, OpenQASM etc.) core compiler realtime Related to CUDA-Q realtime labels Aug 25, 2026
@github-actions

github-actions Bot commented Aug 25, 2026 •
edited
Loading

Copy link
Copy Markdown

CI Summary (push) — ✅ passed

Run #32856791767 · ✅ 7 · ⏩ 7 · ❌ 0 · ⛔ 0

Top-level jobs (14)
Job Result
binaries ⏩ skipped
build_and_test ✅ success
changes ✅ success
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 (43) — every matrix leg, with links
Job Status Link
Build and test (amd64, gcc12, openmpi) / Dev environment (Debug) ✅ success view
Build and test (amd64, gcc12, openmpi) / Dev environment (Python) ✅ success view
Build and test (amd64, llvm, openmpi) / Dev environment (Debug) ✅ success view
Build and test (amd64, llvm, openmpi) / Dev environment (Python) ✅ success view
Build and test (arm64, llvm, openmpi) / Dev environment (Debug) ✅ success view
Build and test (arm64, llvm, openmpi) / Dev environment (Python) ✅ success view
CI Summary ❔ in_progress view
Check for stable CUDA-Q changes ✅ success 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 (6/6) — declared in .github/required-checks.yml for push
Required check Status Link
Build and test (amd64, llvm, openmpi) / Dev environment (Debug) ✅ success view
Build and test (amd64, llvm, openmpi) / Dev environment (Python) ✅ success view
Build and test (arm64, llvm, openmpi) / Dev environment (Debug) ✅ success view
Build and test (arm64, llvm, openmpi) / Dev environment (Python) ✅ success view
Build and test (amd64, gcc12, openmpi) / Dev environment (Debug) ✅ success view
Build and test (amd64, gcc12, openmpi) / Dev environment (Python) ✅ success view

@anthony-santana
Merge remote-tracking branch 'github/main' into research_preview/pulse
ead3eb2 
Signed-off-by: A.M. Santana <anthonys@nvidia.com>
@anthony-santana
Merge remote-tracking branch 'github/main' into research_preview/pulse
3b5948a 
Signed-off-by: A.M. Santana <anthonys@nvidia.com>
@anthony-santana
Merge remote-tracking branch 'github/main' into research_preview/pulse
1711ce3 
Signed-off-by: A.M. Santana <anthonys@nvidia.com>
@anthony-santana

anthony-santana commented Aug 25, 2026 •
edited by github-actions Bot
Loading

Copy link
Copy Markdown
Collaborator Author

/ok to test 1711ce3

Command Bot: Processing...

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@bettinaheim bettinaheim Awaiting requested review from bettinaheim bettinaheim is a code owner
@1tnguyen 1tnguyen Awaiting requested review from 1tnguyen 1tnguyen is a code owner
@bmhowe23 bmhowe23 Awaiting requested review from bmhowe23 bmhowe23 is a code owner
@mitchdz mitchdz Awaiting requested review from mitchdz mitchdz is a code owner
@amccaskey amccaskey Awaiting requested review from amccaskey
@boschmitt boschmitt Awaiting requested review from boschmitt
@justinlietz justinlietz Awaiting requested review from justinlietz
@sam-stanwyck sam-stanwyck Awaiting requested review from sam-stanwyck

At least 1 approving review is required to merge this pull request.

Assignees

No one assigned

Labels

build The issue related to the CUDA Quantum build dynamics Related to dynamics simulation python-lang Anything related to the Python CUDA Quantum language implementation runtime CUDA quantum runtime simulation trouble getting code to run on one of the simulators

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@anthony-santana @lmondada