Sadly, you're quite mistaken.
I've been in contact with the local XILINX sales office and their FAE agrees that v7.1 was too slow to be of any use, and v8.1 is so badly messed up that one can't really do much with it at all. I've gone back to using v6.3.03, which seems, aside from long-standing issues that I've reported, some dating back as far as release 4.2 yet have never been addressed, such as errors in netlisting, etc. The most annoying problems I've encountered have been associated with ECS, but there have been others, e.g. intermittent failure of the software to adhere to Windows conventions with respect to cut and paste, etc, (I could go on, but what's the use).
I've had over a dozen cases, probably closer to three dozen, open over the last year, and certainly quite a few more than that in years past. Only one or two have been resolved in any sense, the remainder having been escalated via CR's, but still remaining unresolved. They're always scheduled for repair in "the next release" but that seldom happens. I have observed that if I don't complain, nothing gets fixed. Of course, if I do complain, there's no guarantee, but I have to go on record.
In 8.1, several major things that immediately impair progress have been introduced. I've complained about that, but I haven't time to do XILINX' work for them. They need to put a few dozen, or perhaps a few hundred, people who've used software to do useful work, and not just to create more software, on the task testing this software against written documents which provide specific criteria consistent with how it is supposed to work.
In release 5.0 (2002) I reported that ECS fails to change line-draw mode when the user selects "manual mode." The workaround is still the only way of dealing with that, and the workaround is to leave line-draw mode and then re-enter it. That sets manual mode. Going back to autorouting mode requires the same set-exit-reenter sequence. Likewise, I complained in 2002 about the fact that bus taps, which won't always align with the line-draw grid. That's still a problem. The workaround for that is to exit ECS and then re-enter, whereupon the bus-tap to line registration is resolved. There are numerous others that work similarly. ECS occasionally reports errors that it can't show on the display. When one exits ECS and re-enters it, they're gone. The auto-increment/decrement of signal names has, since v4.2, been randomly fouled up. Sometimes, when you want it to increment, it decrements, and vice-versa. It's not entirely predictable, but it goes the wrong way more often than not. Since this is their "try it before you buy it" package, I've not bought XILINX software since v3.1 of Foundation, which wasn't a real value either.
Only today, I finally got a reply from the tech support people regarding the v8.1 sheet size management, which I reported on the weekend. When you change sheet size, v8.1 doesn't fully recognize the change. Consequently, as you place symbols, it repeatedly complains that the symbol is placed outside the sheet boundaries, which certainly isn't visible. In order to get the display to reflect the sheet size, you have to exit ECS and then re-enter.
Of course, one could live with that one, but, combined with the fact that some genius decided it would be "cool" to make the various process menus moveable, which doesn't help anything, yet make them so small, vertically, relative to the entire window, in the default that you can't even read title, then hide the boundary, which, BTW, is not located in the conventional place, so you can't reasonably be expected to expand it, and then make the window overlap rather than border the companion menu to which you concurrently need access. Now, I use a
1600x1200 display because I'm getting too old to use finer resolution, but I think I should be able to see what's in front of me when I'm at the default configuration.
Clearly, nobody even cursorily tested the ECS module when this tool suite was released.
There are problems with the associated ModelSim, as well, but I'm not going to describe 'em here, and I'm not going to mess with that complete mistake that XILINX released until I hear rumors that it's working, which may be a while. In the meantime, I'll use the older stuff that seems to work O.K.
Bitching about XILINX software has done so little good that I don't even get any relief from the frustration of having no other option than to switch to Altera. Their stuff isn't perfect either.
Richard