文件列表:

comp_cli.py (6670, 2022-08-30)
src/ (0, 2022-08-30)
src/configuration.py (22532, 2022-08-30)
src/device_data.py (2250, 2022-08-30)
src/fpga_board.py (5749, 2022-08-30)

# compact_bitstream Reduces the size of bitstreams for iCE40 HX8K FPGAs. There are two subcommands: - `rewrite`: compact bitstream - `benchmark`: measure reconfiguration time of compacted bitstream ## Compacting a Bitstream `./comp_cli.py -i uncompressed.bin -b 0 1 -s -l 4 rewrite -o compressed.bin`\ `-i`: input bitstream\ `-b`: BRAM banks to include in the compacted bitstream, banks 0 and 1 in the example (BRAM block skipping)\ `-s`: remove comment (comment skipping)\ `-l`: compaction level, 5 is highest\ `rewrite`: subcommand to rewrite a bitstream\ `-o`: output bitstream ## Benchmarking a Bitstream `./comp_cli.py --hdf5 results.h5 --input-path "/bitstream/in/hdf5" --result-path "/result/in/hdf5" -b -s -l 3 -d "comment added to HDF5" benchmark -r 10000` `--hdf5`: HDF5 file\ `--input-path`: path in the HDF5 file to the original bitstream; it will be compacted before benchmarking\ `--result-path`: path in the HDF5 file where the time measurements will be stored\ `-b`: BRAM banks to include in the compacted bitstream, no bank in the example (BRAM block skipping)\ `-s`: remove comment (comment skipping)\ `-l`: compaction level, 4 means no chunk sorting\ `-d`: descriptive comment that will be added to the time measurements in the HDF5 file\ `benchmark`: subcommand to benchmark a bitstream\ `-r`: number of measurements taken ## Compaction levels - 0: no optimization, write whole CRAM, one chunk per bank - 1: set bank width, height, number and offset only if differs from currently set value (value persistence) - 2: in addition to 1, skip CRAM banks that only contain zeros - 3: in addition to 2, skip CRAM lines that only contain zeros (zero row skipping) - 4: in addition to 3, order writes by chunk size to reduce number of set height commands (chunk sorting)

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：