Date:	Sun, 22 Mar 1998 21:54:43 -0800 (PST)
Message-Id:	<199803230554.VAA24629@adit.ap.net>
Mime-Version:	1.0
To:	Angelo Pesce <us0082 AT uniserv DOT uniplan DOT it>, djgpp AT delorie DOT com
From:	Nate Eldredge <eldredge AT ap DOT net>
Subject:	Re: many questions

At 04:22  3/22/1998 +0100, Angelo Pesce wrote:
[expunged]
>2) What's new in gcc 2.8.0? I currently use PGCC 1.01 that's based on EGCS
>1.01 that's based on gcc 2.8.0. Is there a djgcc port of gcc 2.8.x? 

Yes. It (so I hear) has better template support than previous versions. I
believe it also knows more about instruction timings on x86's, so it can
optimize better. A major feature is exception support in C++, but a
DJGPP-specific bug keeps that from working for the time being. :(

>And
>what's new in binutils 2.8.1? I only know the bad things (fixed size stub
>port), not the good ones... as/gas supports mmx now? 

Yes, although I think one of the MMX instructions generates the wrong
opcode. A patch was posted here recently.

>BTW the bu 2.8.1
>readme says that djp shouldn't work with .exe files, but it works for me...
>Why? (It doesn't work only if I strip the exe: gcc foo.c strip a.exe djp
>a.exe, but if I do: gcc -s foo.c djp a.exe it works well)

Fixed in the latest DJP version.

>
>3) Is the float type 32bit in djgpp? Is the following table correct in djgpp?
>
>float 32bit
>double 64bit
>long double 80bit

Yes, except that `long double' is padded to 96 bits (12 bytes) for
performance reasons. (Btw, you could have found this out with some simple
`sizeof' tests.)

>How can I address 64bits or 80bits with gas (mov QWORD PRT mem,st(0) == ???)?

Isn't it `fldl' for double, and `fldt' for long double?

Nate Eldredge
eldredge AT ap DOT net

- Raw text -