Re: NetBSD/vax compiler bounty?

[Mouse <mouse%Rodents-Montreal.ORG@localhost>]

>>> An educated guess for the peak RAM usage of a native build would be
>>> 512MB RAM, this only matters if you can't add swap because you hit
>>> a virtual address space limit.
>> It also matters if you want your build to complete in days instead
>> of months.  Swap is not a panacea.
> Most of the time, it's not an I/O bound problem to start with, so
> paging isn't really that much of an issue.  gcc is just plain slow.
> It takes forever even if you have no paging at all.

Well, I certainly am not going to argue that gcc isn't slow in general.

But I will argue that making gcc run in less RSS than it wants will
make it slower yet, converting a CPU- and memory-bound task into an
I/O-bound task by way of constant paging.  I ran into this myself on
the i386 port back in...hm, probably in the latter half of 200x, maybe
around 2008: a self-hosted i386 build took a _lot_ longer if the
machine had slightly too little RAM.  (I don't remember how much was
"too little", but I do remember thoughts along the lines of "_that_ is
too little?  That should be _plenty_!".)

/~\ The ASCII				  Mouse
\ / Ribbon Campaign
 X  Against HTML		mouse%rodents-montreal.org@localhost
/ \ Email!	     7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B