Review CUDA findings in your C++test workflow
Prepare a Visual Studio companion solution from a CMake CUDA project, run analysis, and review mapped findings in the IDE
Embedded software consultancy, Parasoft and Perforce partner in Israel for over 20 years. ESL builds and supports the Parasoft CUDA Bridge.
linkedin.com/in/liezrowice · sales@eswlab.comConnect CUDA sources to familiar C++test review surfaces
Prepare, analyze, and review without leaving Visual Studio
Generate a companion solution
For CMake projects, configure with Visual Studio 2026 and create companion projects for CUDA sources.
Use your installed Parasoft tools
Run C++test on host-visible CUDA source using the project's build settings and nvcc host flags.
Stay in Visual Studio
Mapped findings appear in the Error List and editor. An HTML report is also generated.
Evaluation prerequisites
Your own licensed tools, plus an ESL activation token
- Visual Studio 2026 with C++ and CUDA Toolkit MSBuild integration.
- NVIDIA CUDA Toolkit 13.x and a supported project configuration.
- Parasoft C/C++test Professional 2026.1 with an active license.
- CMake for CMake-folder projects.
- A signed, machine-bound ESL activation token.
- Stage 1 provides 30 days from first successful activation.
- Stage 2 adds its own 30-day period when activated; stage 3 is permanent.
- Unit tests and host coverage additionally require C/C++test Standard and GoogleTest.
From CMake folder to analyzable solution
Five steps, all on your machine with your toolchain
.cu.cpp companions.*-parasoft.sln in Visual Studio.Generated solution with findings
.cu.cpp mirror.
Use the rule pack as a focused review aid
Examples
CUDA-1_3 checks selected CUDA API result handling. CUDA-1_4 checks whether a direct kernel launch is followed by an error check. Both are heuristic pattern checks.
Interpret carefully
Checks can miss cases or report false positives. Review source and control flow; a clean run is not proof of compliance.
Keep analysis layers separate
Review findings under the configuration that produced them

Findings in the local vectorAdd run: 1 file, 144 lines, 3 Very High and 2 High.
Heuristic findings in a separate vectorAdd run, across CUDA-1_3 and CUDA-1_4.
Choose where to review the findings
After each run, the Bridge asks where the findings should go

HTML report only
Open the run's HTML report.
Error List only
Show findings with their C/C++test severity.
Both
HTML report and Error List together.
None
Show nothing now. The files are still written.
The same findings in the Error List

See activation status inside Visual Studio
Activate in two steps

Machine-bound, signed by ESL
- Send the fingerprint from the About dialog to sales@eswlab.com.
- Paste the
PCB1-token you receive.
Know what the workflow does and does not cover
- Host-visible CUDA code is analyzed through companion projects.
- The workflow is not native CMake engine integration.
- CUDA source checks are heuristic, not full semantic proof.
- CUDA GoogleTest supports native Visual Studio projects with adjacent tests.
- CMake-folder test discovery is not supported.
- Host coverage does not instrument device-kernel code.
Validated on a specific evaluation bench
2026 · 18.6.2
Professional edition with C++ workload.
13.2
RTX 5080 Laptop, compute capability 12.0.
4.2.2
Visual Studio 2026 generator path.
2026.1.3
Installed licensed tools required.
Start with one representative project
Confirm prerequisites
Verify Visual Studio, CUDA Toolkit, CMake, Parasoft licensing, and ESL activation.
Compare source and findings
Check mappings, severity, false positives, missing cases, and setup output with the owning engineer.
Agree the evaluation scope
Record project/build details, supported configurations, test needs, and acceptance criteria.
Engineering Software Lab Ltd. · sales@eswlab.com · eswlab.com · Daniel Liezrowice on LinkedIn

