I don't know where else to post this, so I hope ya'll won't mind a little complaining:
First of all, the NGDBuild; it's way too picky on constraint names. I don't care if there is nothing in blah blah timing group, don't stop the compile! So the inst/net doesn't exist in my project? Whoopdy freakin' dupe. Don't stop the compile.
edif2ngd: "2 0 0" Need I say more? Talk about minimal support. You'd think it was 1988.
While we're on the topic, XST does a heck of a lot of optimization that could be done in the mapper or the edif2ngd. Those of us not using XST or some other $50k synthesizer are pretty much SOL on automatic optimizations. I'm talking about automatic use of local signals, muxf and carry chain inferrence, etc.
Mapper: those messages about multiple references to padlocs are confusing at best. Not only that, it takes ages for it to realize there's a problem with hardset location constraints.
PAR: This tool has improved a lot in the past five years. Nevertheless, I still lose at least an hour a day to it. I'd pay $1500+ for a board with a few Spartans (or something) and RAM chips to accelerate the tool, if it supported some kind of accelerator board. Is anybody working on this? I need some feature to autotrigger the "user requested termination" after x minutes after 0 untrouted signals, if that makes sense.
Impact: $500 for the USB cable? Good thing it's only five times the price of the parallel cable. Eeesh. I expect Vasaline with my next shipment. The DMA stuff for the Parallel 4 is so picky, I've only got it to work on one of my ten boards. Does it have to be that exact? How about some timeouts on the verify for when it gets stuck? The batch support on the tool is nice but I have no way to pass in parameters. You know, like %1 %2 etc. -- general features to make the batch stuff usable on a real project?
I suppose that's enough ranting for one day.....