PowerPro fails in 2026 Roadside Recognition Hackathon VM: unable to generate power.rpt

Hi Matt,

I am participating in the 2026 Roadside Recognition HLS Hackathon and I am consistently unable to complete the PowerPro stage in the provided VM.

The Python model, fixed-point C++ model, Catapult HLS, RTL generation, SCVerify, and switching activity all complete successfully. My baseline model reaches 97.89% accuracy. However, the flow fails when PowerPro is launched, so power.rpt is never generated and I cannot obtain Energy Per Inference for submission.

The main errors are:

Could not find $PPRO_QUESTA_HOME/bin
Could not find $PPRO_VISUALIZER_HOME/bin

[PPRO-BNFCE] ‘vlog’ binary not found
[PPRO-BNFCE] ‘vopt’ binary not found
[PPRO-BNFCE] ‘visualizer’ binary not found

ERROR: Failed to connect 19800
powerpro session open: timeout waiting for IPC connect

I checked the VM and the required binaries actually exist:

/opt/questasim/questasim/bin/vlog
/opt/questasim/questasim/bin/vopt
/opt/visualizer/visualizer/bin/visualizer

The environment variables in my shell are also:

PPRO_QUESTA_HOME=/opt/questasim/questasim
PPRO_VISUALIZER_HOME=/opt/visualizer/visualizer

PowerPro itself is installed and starts successfully through:

/opt/hackathon_2026/catapult_home/pkgs/cds_ppro/powerpro/bin/powerpro

and its required PowerPro license features are checked out successfully.

However, PowerPro still reports that it is searching under:

/opt/questasim/bin
/opt/visualizer/bin

rather than the actual installation directories one level below them.

I have reproduced the same PowerPro failure on more than one complete baseline run, each taking roughly an hour, so it does not appear to be a transient connection issue.

Environment:

Catapult Ultra 2026.2_1
PowerPro 2026.1
RHEL 9.8 hackathon VM

Could you please confirm whether this is a known issue with the current hackathon VM image and advise how to correct it?

Ideally, is there also a supported way to rerun only the PowerPro/Power RTL stage using the already generated Catapult project, rather than repeating the full HLS flow?

Since Energy Per Inference is required on the submission form, I currently cannot produce a valid competition entry despite the accelerator flow completing successfully up to power analysis.

I can provide screenshots, powerpro.log, powerpro.err, and the Catapult log if useful.

Thanks.

A post was merged into an existing topic: Hackathon questions and issues