Bitstream compression

I see you are following my lead:

My latest baby: XC2VP50, about half FFs (& 2/3 slices) used.

gzip -9 reduces the file size from 2.322M to 498k. I think that amount of compression is typical for most designs. I haven't seen anything much worse than that.

I recommend avoiding Windows compression utilities (such as WinZip) because sometimes the decompression will need to take place on a small microcontroller. Something like gzip has the added benefit of being open source.

Regards, Allan

Agreed. I just started looking into this and using WinZIP is a lazy first step to get a sense of proportion. My gut feeling is that it would be perfectly safe to allocate space equivalent to 50% of the uncompressed bitstream. If less is used, fine. The problem I have is that the design is likely to change after the product is released...so, I can't nail down the configuration file at this stage of the game.

I'm tempted to start with RLE. It would be easy to implement a decompressor with a small uC and limited register/ram resources.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Martin Euredjian eCinema Systems, Inc.

To send private email: x@y where x = "martineu" y = "pacbell.net"

Did bitstream compression in my last project. Used LZW. RLE didn't improve so much vs. "internal" xilinx compression.

For many reasons we had to reconfigure FPGA very quickly; with some care to code, was blazingly fast.

Here are some compression ratios on that design: CAMERA_COMP.BIN 97944 uncompressed

After LZW compression:

BIT SIZE

Google Groups around with "compressing Xilinx bitstreams some test data"

or retrieve the thread clicking here:

Thank you for the various links and recommendations on this subject.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Martin Euredjian eCinema Systems, Inc.

To send private email: x@y where x = "martineu" y = "pacbell.net"