> I assumed that the latest gcc uses SHT_GROUP, but > don't actually > know which version introduces it. So let's find out > what your new > gcc is doing. Take an object built with your new gcc > compiler, and do: > > % elfdump -c | grep linkonce > > If you get any output, then you're not using a > compiler with the SHT_GROUP > stuff, and the fact that you're getting the same link > times is not surprising.
Bingo! 4.3.2 is still using linkonce. > Alternatively, I do know that the Sun compilers use > SHT_GROUP. > I don't know if that's an option for you. I'm afraid not. We already go through a fair amount of hassle dealing with two compilers; we'd need a really good reason to add a third, and faster link times on my machine (the only one we have running Nevada) isn't it. I'm sure I couldn't talk people into it. > Another data point: When I run your big link on my > Nevada system > (Ultra 24, 4GM memory), it links in 10sec. (My test > yesterday, with > profiling turned on took 18 sec). I assume you mean "minutes" rather than "seconds". That said -- wow! This must be at 3.33GHz, yes? I'm getting like 17 minutes on an Opteron 290 (2.8GHz). I knew the Core 2 was faster than the Opteron, but I didn't know it was that much faster. Anyway ... I do have 24GB of DRAM, so there's no question of thrashing :-) -- Scott -- This message posted from opensolaris.org
