Re: GCC/G++ will not compile i686 binaries when asked.
[EMAIL PROTECTED] wrote:
That may indeed be it. Is there any known way short of looking a the
asm dump of a binary for telling what CPU instructions are used? When
asked I was directed to readelf and file and they work great for
SPARC based development,
I wouldn't bet on that either:-)
Seems to me your observations may have confirmed your expectations.
However, I don't feel like firing up my {,ultra}sparcs to do any testing
atm.
--
Cheers
John
-- spambait
[EMAIL PROTECTED] [EMAIL PROTECTED]
Tourist pics http://portgeographe.environmentaldisasters.cds.merseine.nu/
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: GCC/G++ will not compile i686 binaries when asked.
On Thu, Aug 26, 2004 at 01:32:21PM -0700, [EMAIL PROTECTED] wrote:
> That may indeed be it. Is there any known way short of looking a the
> asm dump of a binary for telling what CPU instructions are used? When
> asked I was directed to readelf and file and they work great for
> SPARC based development, so I had assumed them to be correct for pc
> development as well.
>
> Thanks in advance for the enlightenment.
Nope. Just wrote a tiny unoptimizable program (true clone), compile for
386 and PII, and they cmp the same...
I'd likely go for a pipeline, this works on my system:
objdump --disassemble -j.text testC | grep '^ ' | perl -pe 's/.{32}//; s/ .*//' | sort
-u
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: GCC/G++ will not compile i686 binaries when asked.
That may indeed be it. Is there any known way short of looking a the
asm dump of a binary for telling what CPU instructions are used? When
asked I was directed to readelf and file and they work great for
SPARC based development, so I had assumed them to be correct for pc
development as well.
Thanks in advance for the enlightenment.
On 25 Aug 2004 at 18:23, Stefan O'Rear wrote:
> On Wed, Aug 25, 2004 at 05:48:45PM -0700, [EMAIL PROTECTED] wrote:
> > For the past few weeks I have been co-working on a private project. I
> > have done allot of work on it and I would hate to have to start over.
> >
> > The makefile I created for this project explicitly tells the compiler
> > (g++) to compile the binary using i686 (-march=pentiumpro) assembly
> > instructions and internal timeings. Yet for some [explitive]
> > [explitive] reason it refuses to do so and both ?file? amd ?readelf?
> > tell me it always compiles for a bare i386 despite my explicit
> > commands otherwise.
> >
> > Two questions:
> > Why the [explitive] does this [explitive] thing not compile the end
> > result into a i686 only binary as I asked it to?
> > How the [explitive] do I fix it?
> >
> > I have read the gcc/g++ manual pages. I have googled. I have spent
> > the past hour dedicating myself to this problem, and nothing seems to
> > work.
> >
> > This is [b]really[/b] getting to me. Please help.
>
> C file:
>
> int main(int argc) {
> int sam = 0;
> if (argc > 1) sam = argc;
> return sam;
> }
>
> Compile:
>
> ~ %% gcc -o testC test.c -mcpu=pentiumpro -march=pentiumpro -O3
>
> Dissassembly Excerpt:
>
> 08048320 :
> 8048320: 55 push %ebp
> 8048321: 31 c0 xor%eax,%eax
> 8048323: 89 e5 mov%esp,%ebp
> 8048325: 83 ec 08sub$0x8,%esp
> 8048328: 8b 4d 08mov0x8(%ebp),%ecx
> 804832b: 83 e4 f0and$0xfff0,%esp
> 804832e: 83 f9 02cmp$0x2,%ecx
> 8048331: 0f 4d c1cmovge %ecx,%eax
> 8048334: 89 ec mov%ebp,%esp
> 8048336: 5d pop%ebp
> 8048337: c3 ret
>
> See that cmovge - that's a Pentium ONLY insn.
>
> And readelf says:
> Machine: Intel 80386
>
> Is that ^^^ why you thing gcc won't make 686 programs?
>
>
> 1. A PPro exec is the same as a 386 exec except for the insns. ReadELF
>will tell you it's a 386 exec even if you have a PPro exec.
>
> Please comment. If it doesn't have 686 asm and could benefit from it, we
> (atleast I) need more info.
>
> P.S. I use woody with a GCC-3.3 from unstable and the ness. libs.
>
>
> --
> To UNSUBSCRIBE, email to [EMAIL PROTECTED]
> with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
>
>
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: GCC/G++ will not compile i686 binaries when asked.
On Wed, Aug 25, 2004 at 05:48:45PM -0700, [EMAIL PROTECTED] wrote:
> For the past few weeks I have been co-working on a private project. I
> have done allot of work on it and I would hate to have to start over.
>
> The makefile I created for this project explicitly tells the compiler
> (g++) to compile the binary using i686 (-march=pentiumpro) assembly
> instructions and internal timeings. Yet for some [explitive]
> [explitive] reason it refuses to do so and both ?file? amd ?readelf?
> tell me it always compiles for a bare i386 despite my explicit
> commands otherwise.
>
> Two questions:
> Why the [explitive] does this [explitive] thing not compile the end
> result into a i686 only binary as I asked it to?
> How the [explitive] do I fix it?
>
> I have read the gcc/g++ manual pages. I have googled. I have spent
> the past hour dedicating myself to this problem, and nothing seems to
> work.
>
> This is [b]really[/b] getting to me. Please help.
C file:
int main(int argc) {
int sam = 0;
if (argc > 1) sam = argc;
return sam;
}
Compile:
~ %% gcc -o testC test.c -mcpu=pentiumpro -march=pentiumpro -O3
Dissassembly Excerpt:
08048320 :
8048320: 55 push %ebp
8048321: 31 c0 xor%eax,%eax
8048323: 89 e5 mov%esp,%ebp
8048325: 83 ec 08sub$0x8,%esp
8048328: 8b 4d 08mov0x8(%ebp),%ecx
804832b: 83 e4 f0and$0xfff0,%esp
804832e: 83 f9 02cmp$0x2,%ecx
8048331: 0f 4d c1cmovge %ecx,%eax
8048334: 89 ec mov%ebp,%esp
8048336: 5d pop%ebp
8048337: c3 ret
See that cmovge - that's a Pentium ONLY insn.
And readelf says:
Machine: Intel 80386
Is that ^^^ why you thing gcc won't make 686 programs?
1. A PPro exec is the same as a 386 exec except for the insns. ReadELF
will tell you it's a 386 exec even if you have a PPro exec.
Please comment. If it doesn't have 686 asm and could benefit from it, we
(atleast I) need more info.
P.S. I use woody with a GCC-3.3 from unstable and the ness. libs.
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: GCC/G++ will not compile i686 binaries when asked.
[EMAIL PROTECTED] wrote: For the past few weeks I have been co-working on a private project. I have done allot of work on it and I would hate to have to start over. The makefile I created for this project explicitly tells the compiler (g++) to compile the binary using i686 (-march=pentiumpro) assembly instructions and internal timeings. Yet for some [explitive] [explitive] reason it refuses to do so and both ?file? amd ?readelf? tell me it always compiles for a bare i386 despite my explicit commands otherwise. Two questions: Why the [explitive] does this [explitive] thing not compile the end result into a i686 only binary as I asked it to? How the [explitive] do I fix it? I have read the gcc/g++ manual pages. I have googled. I have spent the past hour dedicating myself to this problem, and nothing seems to work. This is [b]really[/b] getting to me. Please help. I'm curious about why it's particularly important to you? For sure gcc will do what you want: I have been caught booting a kernel that trapped on a Pentium because it was built for 686, and when I redeployed a Pentium II (it was running RHL 7.3) I replaced the kernel and (I think glibc) but got caught on the ssl libraries - it's a mazing how many programs won't run with broken openssl!. So the Q is, how do you know it's not working for you? Have you tried building for an Athlon and running on your target? Do you have a small test program that illustrates your problem? If you come up with (say) ten lines of C, probablly people here will check it out for you on a variety of software. -- Cheers John -- spambait [EMAIL PROTECTED] [EMAIL PROTECTED] Tourist pics http://portgeographe.environmentaldisasters.cds.merseine.nu/ -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
re: gcc / g++
Michael -- Thanks for your reply. I had not tried that. I must admit I've been a bit confused about the differences between gcc, g++, egcs, libc5 and libc6. The program I was trying to compile worked on Slackware 3.5 (libc5, I suspect) with gcc, and I seem to recall compiling it on Red Hat 5.1 (glibc/libc6). I have a much better feeling about things, now that I've gotten C and C++ to compile on my debian system. Thanks again! CSS On Mon, 28 Dec 1998, Michael Procario wrote: > > Have you tried > > g++ -o hello hello.cc > > g++ is the egcs c++ compiler It know waht libraries it needs. > -- Christopher S. Swingley International Arctic Research Center University of Alaska Fairbanks [EMAIL PROTECTED] mercury.bio.uaf.edu/~cswingle.grad -- An evil mind is a great comfort.
Re: gcc & g++
Hi, > Since g++ is now seperate from gcc, I noticed after installing both of > these that I have two directories under /usr/lib/gcc-lib/i486-linux. > One is 2.7.2.3 from gcc and the other is egcs-2.90.26. Other than a few > differences they seem to have the same files each taking almost 3M. I'm > running on a 486 laptop with a 120M hardrive and was wondering if I can > combine the files in these directories to use less space. No, you cannot. gcc and g++ are are not the same version. g++ is from the egcs-1.0.x distribution (the same for g77). gcc was not updated to the egcs version since there are still some problems with egcs gcc, which is why you have 2.7.2.3 and egcs-2.90.26 directories. Both gcc and g++ expect those certain subdirectories that depend on the version of the compiler. For this reason, you cannot combine the 2.7.2.3 and egcs-2.90.26 subdirectory. -Ossama -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
