Evolve in Digital Design Studio
Digital Design Studio 2.0.2: Evolve in Digital Design Studio. Check the selected host, account permissions and current support before use.
Virtual-FPGA candidate search and parallel candidate evaluation use the Linux desktop native host. Save a baseline, choose the supported board/profile and inspect candidate diagnostics and metrics before accepting a change. This Evolve path is unavailable on Windows and Cloud. Managed Cloud optimization is a separate operation whose availability, estimate and permission must be confirmed by that project.
Overview
Evolve explores design alternatives inside Digital Design Studio. Start from a saved RTL baseline and review each candidate with the same correctness checks and objective. Use the current host’s runtime status to determine which search or analysis operations are available.
Prepare an experiment
For the managed combinational search, save one DUT file (.sv or .v) and choose its top module; DDS captures only the .svh/.vh headers that file includes.
Testbenches, defines and other modules are not inputs.
Choose the objective and record its units and method. The objectives are area and timing.
Choose the runtime and review execution limits, external tools, destination and any usage estimate.
Use a new run or experiment location so previous results remain available.
Run and inspect candidates
Open the Evolve view in the selected project. Read the runtime message before starting. If the operation is available, start it explicitly and follow the run’s status, diagnostics and stored outputs. Opening the view or obtaining a preflight result does not start a search.
Compare candidates against the baseline using the same inputs and checks. Review the source diff, failing tests and missing metrics before accepting a candidate into your normal source-control workflow. Keep the original baseline and experiment artifacts.
CLI and CUDA paths
The CLI has several distinct Evolve contracts. The legacy customer alti evolve action is plan-only. Native Evolve operation requests and compute rtl-evolve use their own schemas and capability checks. Use Engine commands and Job/Cloud commands for the exact installed interface. The alti bundled with the Windows app is for project initialization only and includes no Evolve engine, so alti evolve … and alti compute rtl-evolve cannot run there.
In the bounded CUDA workflow, GPU work calculates combinational activity diagnostics. Candidate generation, correctness checks and objective evaluation remain CPU work, and Icarus is required. Read the recorded backend and limits rather than treating a detected GPU as a completed search or a speedup result.
Resume or cancel carefully
After a disconnect, inspect the existing run before starting another. A cancel request can remain pending while a runtime cleans up. Preserve useful candidate outputs and the last known job state. If source or configuration changed, create a new experiment rather than comparing it under the old identity.
Files and related guides
Compatibility identifiers rtl_evolve, /api/evolve/*, .alEvolve and the workspace evolve/ directory remain unchanged. See Analysis views, Source control and Projects and runners for the surrounding workflow.
Using Digital Design Studio 2.0.2
Virtual-FPGA candidate search and parallel candidate evaluation use the Linux desktop native host. Save a baseline, choose the supported board/profile and inspect candidate diagnostics and metrics before accepting a change. This Evolve path is unavailable on Windows and Cloud. Managed Cloud optimization is a separate operation whose availability, estimate and permission must be confirmed by that project.