mirror of https://github.com/YosysHQ/nextpnr.git
515eacf55a
nextpnr-mistral defaulted to UNCOMPRESSED RBFs (--compress-rbf opted in to compression). An uncompressed bitstream configures the device (the socfpga FPGA manager reports 'operating'), but on the FPGA-manager load path it leaves user IO non-functional — verified on a DE10-Nano/MiSTer: an uncompressed single-LED design stays dark, the byte-identical compressed encoding lights it. Quartus also emits compressed. OPT_B already tracked the compression choice in-tree (the compressed branch set the same OPT_B value Quartus uses), so this is purely a default flip: read a new 'uncompressed_rbf' opt-out instead of a 'compress_rbf' opt-in. --compress-rbf is kept as a deprecated no-op; --uncompressed-rbf restores the old behaviour for debugging. Verified: regenerating a design with no flags now produces bytes identical to a bitstream confirmed working on silicon. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> |
||
|---|---|---|
| .. | ||
| CMakeLists.txt | ||
| arch.cc | ||
| arch.h | ||
| arch_pybindings.cc | ||
| arch_pybindings.h | ||
| archdefs.cc | ||
| archdefs.h | ||
| bitstream.cc | ||
| constids.inc | ||
| delay.cc | ||
| globals.cc | ||
| io.cc | ||
| lab.cc | ||
| m10k.cc | ||
| main.cc | ||
| pack.cc | ||
| pins.cc | ||
| qsf.cc | ||