Hi,

On Sat, Oct 12, 2019 at 8:15 AM TK Chia <[email protected]> wrote:
>
> Thanks!  This looks like ... a rather major revamp of the makefiles. :-)

For TC++, I tried to minimize loading the compiler too many times.
It's able to compile multiple sources together in one invocation,
which I guess is possibly faster (dunno ... for something like DJGPP
it's not, AFAIK). Part of the problem is simple CFLAGS switches (which
waste precious cmdline space, but using a response file is too messy
here), so I (half) disabled USINGXMS just so it wouldn't unnecessarily
reload the compiler. (I have no idea if that helps either. It's not
default, but EMS will be used if found.) Then I did a dopey kludge
just to make sure it would compile outbin.c and outobj.c both at once
and not erroneously try to do it again. (Some makes let you do this
easily, others not. It can get complicated fast.)

> It seems that the precompiled nsm09839.zip binaries in iBiblio were
> built with Microsoft C, most probably using Makefile.ms7.  And the
> makefile has this...:
>
>    #   Compile for a 286, ain't nobody using an 8086 anymore

Maybe it was for smaller size, and thus saved RAM? Otherwise it makes
little sense. Sure, 8086 clones aren't as popular, but just from an
emulator standpoint (and the fact that it's possible at all), there
should've been separate builds. (Besides, like I said, 186+ ENTER is
tons slower on semi-modern machines!)

>> Using the above command line, the resulting size of tinyasm.exe is 27,200 
>> bytes.

With the DJGPP-hosted tools, I'm getting 45 kb. Is that correct? Or do
you have newer libs? Sure, after UPX, it's about 27 kb. Either way,
quite nice!

>>> I still want to get PSR Invaders working with it, I haven't quite done it 
>>> yet.
>>> (I may or may not be able to mentally wrangle around the quirks with Sed
>>> scripting. I'm only half successful so far.

I've done it!  :-)  Sure, it took some workarounds for those issues I
mentioned before, but I'd already done similarly for Octasm. (Even
with functionally-identical instructions, if the size is different, it
makes verifying everything much more tedious.) So yeah, now it's 100%
identical to NASM output (all with simple sed scripts, the whole .BAT
is less than 3 kb). I doubt anyone cares besides me, but I like it!
(No, I haven't updated the FD Package yet.)

Now, one of these days, I need to test a bunch more of this stuff
under 8086tinyplus (again).


_______________________________________________
Freedos-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freedos-user

Reply via email to