Next steps
You have a result bundle and a configuration file. The rest of the site is organised by task, and this page maps the questions people ask after a first profile onto the page that answers each one.
Prerequisites
Section titled “Prerequisites”First profile and Configure a run done.
By what you want to do
Section titled “By what you want to do”| I want to | Go to |
|---|---|
| Understand every file hpx wrote and the units in each column | Output files |
| Compare two runs and know when a comparison is not fair | Analysis and comparison |
| See per-layer cost on the model graph | Model Explorer overlays |
| Try another engine: heliaAOT, TFLM or ExecuTorch | Engines |
Build with armclang or ATfE instead of GCC |
Toolchains |
| Move the arena or the weights to a different memory | Memory placement |
| Capture more counters, or counters from the memory and MVE domains | PMU counters |
| Measure power with a Joulescope or an INA228 | Power measurement |
| Capture over USB CDC or UART instead of RTT | Capture transports |
| Run hpx from Python instead of the shell | Python API |
| Script a batch of runs or wire hpx into CI | Batch runs from YAML |
| Understand a failed run or an issue code | Issue codes |
By section
Section titled “By section”Task pages in four groups: set up a run, measure, read results, and the concepts behind the pipeline.
Recipes for engines, toolchains, boards, power capture and workflows. Each page identifies its configuration and validation status; some show illustrative output rather than a captured result.
The command line, every configuration key, the issue codes and the Python API, all generated from the source, plus the compatibility baseline and the firmware wire protocol.
For agents
Section titled “For agents”Every page on this site has a Markdown rendition at its route plus
index.md, the whole site is in llms-full.txt, and the generated
reference ships as JSON: reference/api/reference.json,
reference/cli/cli.json, reference/configuration/schema.json and
reference/issue-codes/issues.json. Start from
llms.txt for the index.
Expected output
Section titled “Expected output”Nothing runs on this page. If none of the rows above matches your question, the troubleshooting section of First profile and the issue codes are the two places a failed run sends you.
What next
Section titled “What next”Whichever row you picked. The section landing pages carry the same map in more detail.