FPGA Development
Validate before committing a mask set
Running real RTL against real firmware on FPGA finds awkward boundary behaviour while it is still cheap to fix. FPGA validation is the next milestone and is not complete.
Design targets · not measured silicon results
- Trust boundary
- Supporting logic
- Signed data in flight
Simulation covers what the testbench thought to ask
Boundary cases, sagging supply, restore under changed contracts, mesh events during signing, are where the design matters. Finding them after tape-out is not patchable.
- Measured epoch
- Flagged for re-measure
- Light = the digest being extended
Real design, real firmware, at speed
RTL is synthesised for FPGA and brought up with production firmware. Function and protocol transfer to ASIC; timing, area and absolute power do not.
Sequence
Order is the mechanism, and nothing here is optional.
- Ordered step
- Only exit
What follows from the diagram
-
Fabric synthesis
Portability checked by mapping to FPGA resources
-
Firmware bring-up
Integration faults appear against real software
-
Relative energy ranking
Compare variants; do not predict silicon watts
-
Explicit limits
No ASIC claim is built on an FPGA absolute
Reviewing this mechanism?
The specification can still change. That stops being true after tape-out.