RAM usage...

[Anoq of the Sun]

Stephen Weeks wrote:
> 
> Hi Anoq.  I wanted to let you know that the new release of MLton fixes the
> problem with the compiler trying to use too much heap space.  Although it still
> tries to grab 85% of RAM.  Please let us know if you still see the problem.

I have tried to install the new binary, and from that compile my
current half-ported version of the source to native Linux.
And it compiles fine now - with 512 MB RAM and no swap and without
setting the 430m heap-flag.

So I think it works just fine! :)

I have also been working some more on the port, but there are
a lot of Unix-specific stuff in the basis library that I have
to put inside #if defined(_WIN32) #else ... #endif
So it's taking some time - but at least I think I'm close now.
And the Makefiles can even do most of the weird stuff
automatically now :)

Also, when compiling the compiler itself, it must of course
run on Linux - because it is a cross-compiler. So I'm now
in doubt about whether I should really use the constants
file from Win32... I guess it's really important to make
sure that only those constants that relate to the generation
of the code should be modifed (as you also mentioned earlier).

I think that implementing the sanity-check you talked about
might actually be very useful here! I have compared the
Linux and the Windows constants files, and there are a
few differences. In particular the first constant is
64 on Linux and 32 on Windows, and I wouldn't be surprised
if this is the size of a pointer or something - which might
give lots of problems... Seeing the names of the
constants might enable us to "handcook" a constants
file that will work for the crosscompilation :)

Also, I hope there will not be too much hazzle in merging
your latest release with my ported version, since I'm still
using the 20010806 release to port from...


Cheers
-- 
http://www.HardcoreProcessing.com