[saqibkh]

build.yml only compiles the code, because the hosted runners have no GPU, so the ctest suites (roundtrip, hostptr, fortran) are never executed in CI. This adds a separate, optional workflow .github/workflows/gpu-sim.yml that builds the library and runs them on a simulated NVIDIA H100 using PantheonSim (no GPU needed). I am the author of PantheonSim, so please weigh this accordingly.

It does not touch any source file, leaves build.yml as it is, and is not meant to be a required check. A simulator cannot reproduce hardware-specific behavior or timing, so it complements real-GPU testing and does not replace it.

What it does:

• installs gfortran, the CUDA 13.0 toolkit and the simulated H100
• configures with -DCMAKE_CUDA_ARCHITECTURES=90-virtual (the simulator runs the embedded PTX, so the default 80 90 is not used) and -DCMAKE_CUDA_RUNTIME_LIBRARY=Shared (the simulator needs a shared CUDA runtime)
• builds, then runs ctest --test-dir build --output-on-failure under vgpu run
• runs on manual dispatch and on pull requests to main that touch src/, include/, tests/, tools/, examples/, fortran/ or CMakeLists.txt, with a concurrency group that cancels superseded runs

Measured:

• Run on my fork: https://github.com/saqibkh/SZ-GPU/actions/runs/42745621146 . All 3 ctest tests passed (roundtrip 7.2 s, hostptr 0.7 s, fortran 0.5 s). The roundtrip cases include the 32M-element float and double fields, all within the error bound.
• The job took about 16 minutes, almost all of it the simulator and toolkit setup; configure and build took about 10 seconds and the tests about 9 seconds.
• The Python tests are not part of this workflow.