- 新着としてマーク
- ブックマーク
- 購読
- ミュート
- RSS フィードを購読する
- ハイライト
- 印刷
- 不適切なコンテンツを報告
Hello!
My question is very simple: Do I have to do a full compilation run if only one pin changes in the pin assignment? Or do I just have to run assembler or something else, again? My design takes about 18 minutes for a complete run. That is a long time, if I only changed one pin in the assignment editor . . . . Regards, Maikコピーされたリンク
5 返答(返信)
- 新着としてマーク
- ブックマーク
- 購読
- ミュート
- RSS フィードを購読する
- ハイライト
- 印刷
- 不適切なコンテンツを報告
The short answer is no. (You can probably skip analysis and synthesis, and just rerun the fitter, timing analysis and assembler, but the fitter is probably the bulk of your time anyway.)
You could manually reroute this signal in the chip-editor, but I think that would take close to 18 minutes to do anyway once you figured out how to use the tool. If you were doing incremental compilation, you could probably shorten it, but a design that takes 18 minutes to compile probably won't benefit much from that flow, just because the compilation overhead is probably a good bulk of your compile time. (Although it may feel long, 18 minutes isn't bad at all. Wait until you have a 24 hour compile, and then you really need to plan for quicker compiles.) (And you may ask why the fitter doesn't just recognize this change and change that one thing. For this particular scenario, it might actually be possible with a specialized algorithm, but trying to "recognize changes" and only refit those changes has been an area of work by all FPGA companies and I don't think anyone has a perfect flow. In reality, it's a much more complex problem then it appears to be on first glance. The incremental compilation flow works, but requires the user to set up their design into partitions which gives Quartus information as to what has changed and what hasn't.)- 新着としてマーク
- ブックマーク
- 購読
- ミュート
- RSS フィードを購読する
- ハイライト
- 印刷
- 不適切なコンテンツを報告
Rysc,
thanks for the "short" answer ;) . As always, your additional informations are very valuable! Maik- 新着としてマーク
- ブックマーク
- 購読
- ミュート
- RSS フィードを購読する
- ハイライト
- 印刷
- 不適切なコンテンツを報告
Short answer.
Yes. More old timer stuff ----- ------ and as a point of reference, some of us go back to the day of punch cards with a 24 hour turn just for a typo like a missing comma. ------ early Xilinx compiles on an IBM XT (dual floppy) could take up to 36 hours for a 10 x 10 Grid of CLB's (LAB"s) ((200 F/F designs).- 新着としてマーク
- ブックマーク
- 購読
- ミュート
- RSS フィードを購読する
- ハイライト
- 印刷
- 不適切なコンテンツを報告
How do I get a 18 minute compile time? I'm happy with my 2 hour compile with PCIe, 32 FIR filters and serial lite. And I usually have to run multiple fitters with different seeds to make timing.
