[SM-Discuss] License Choices
cerise at littlegreenmen.armory.com
Sun Jul 21 18:54:56 EDT 2002
Glenn Shannon insolently stated:
> I believe the choice to use tmpfs was because compiling in memory is
Ahh yes, I've heard this myth. It sounds nice, doesn't it? Almost
believable. I think there's good reason to doubt it though.
In theory, compiling memory is faster because memory access is
faster than disk access.
In practice, where you've just unloaded a rather large tarball on
memory, it seems destined to end up in swap which will incur disk
access. Two of them at minimum in fact. One to write to virtual
memory, the other to read it back out.
I'm ignoring lots of other factors considering that disk access is
an order of magnitude slower than CPU time. It seems like disk access
is the thing to measure here. With that in mind, it seems like using
tempfs (which isn't exactly the most sophisticated fs in the world...)
over your native filesystem (which will almost certainly be more
sophisticated than tempfs and should have faster algorithmic run times)
is going to hopelessly slow things down.
More information about the SM-Discuss