| |
| |
| |
|
Comments:
<0> gotta run...thanks for the tips velco
<1> np
<2> hi
<2> is there a command-line option to enable thread-local storage?
<3> does anyone know where in the gcc sources cc1 is called?
<1> Christoph_vW: it's exec'd by gcc
<1> maybe in toplev.c
<1> err
<1> toplev.c is cc1
<1> gcc.c then
<3> I will have a look at it
<3> I am still searching for "cannot exec `cc1'"...
<3> I think I know why gcc isn't working here
<3> it is using "/" instead of "\" on Windows
<3> which doesn't seem to work on Windows Vista anymore
<1> oh?!
<1> ****ing M$
<1> I have many many Makefiles with / in them ...
<3> C:\RosBE\3.4.5\bin>../libexec/gcc/mingw32/3.4.5/cc1 --> bad command or filename
<3> C:\RosBE\3.4.5\bin>..\libexec\gcc\mingw32\3.4.5\cc1 --> works
<1> hmm
<1> that's the cmd.exe I guess
<3> yes
<1> I'd be more concerned with CreateFile and such ...
<3> gcc can't find cc1.exe...
<3> "/" works in ShellExecute
<3> the question is: what is gcc using to call cc1?
<1> functions in libiberty
<1> pexecute
<3> nice lib
<3> with this lib it should be possible to port distcc to windows as well
<1> pex_run
<1> gcc.c:execute()
<4> darn this is hard to figure out why doing the cross compile .When its making the final gcc to cross--tools then to /tools it makes the gcc with c,c++,ada,java,objc,obj-c++,fortran and mulltilib. But no0w from tools to root / its failing at java :(
<5> Hi I have a doubt about gdb on Ubuntu. I am trying to debug amarok, but I receive this error. This GDB was configured as "i486-linux-gnu"..."/home/leandro/amarok-1.4.4/amarok/src/collectionscanner/amarokcollectionscanner": not in executable format: I suppose I have to recompile gdb or amarok to solve this. Right?
<5> sorry if this is notr the right place to ask. But I didn find a gdb irc channel
<4> why is it with disabling multilib and still compiling for x86_64 you can compile it with java . bit confused on why it will but not for the multilib O.o
<4> darn looks like if you want the java enabled with gcc you need to drop multilib :(
<6> hi guys, is there a way to have gcc complain, for a C source file, about the "static" storage specifier not at at beginning of a declaration or definition, like for "inline static void foo ();" ?
<7> ping
<7> ok no nickserv needed :) Are there any good websites that state cflags that often help in particular cases ?
<7> yes .... but 50% of them ... i don't even understand :D
<7> if i add "all" of them ... the code runs twice as fast ... but certain pieces of the code run 10X as slow :)
<7> (the main time of the program goes to resampeling and to the operation %)
<7> and tolerates fast-math
<8> Qantourisc: that's the problem with over-optimisation. I hardly use anything but -O2 for fast code and -O0 for debugable code...
<8> for any other optimisation one has to understand what they do. Like fast-math, which could also be called -fbreak-all-but-some-math-which-is-all-I-need-here
<8> but that would be a bit too long...
<7> problem is programs like acovea .. depend on your architecture AND take way to long :)
<7> i also have a lot of loops ... especially enable things such as fun-loop-switch and "move constants out of loops" ?
<8> if they would make all loops faster, they would be in -O2. Everything that is not in -O2 has a sigificant chance to make things slower or to break code.
<7> hmmm
<7> i don't care if they make the code of my program that doesn' get used ... aslong as the 80% of the time that is spend in the 20% of code is fast :)
<8> especially loop-unrolling has most of the times a significant effect. Though it is hard to predict in which direction.
<7> but easily tested no ? :)
<7> what kind of flags are extermely arc senstivice ?
<8> Qantourisc: arc=?
<7> (so i don't include those as default in my makefile, but can be specified by the user)
<7> architecutre
<7> pcc, amd, intel, powerpc, ...
<8> no idea. I hardly use anything but i486 or supersparc.
<7> ok thanks
<7> noshadow: also ... how independent are the cflags of eachother ?
<7> (as in term of execusion speed)
<8> the main effect nowadays on speed is the cache usage. Sometimes a single -Os can make something faster than all other wild combinations of things.
<8> if one flag makes something a little bit bigger, and another flag makes it a little bit better. Both together may lead to it no longer fitting in the cache and making it m***ive slower.
<8> I just stoped caring someday. For almost all programs the time needed to compile it multiple times with some combinations will never got back in reduced run-time...
<7> noshadow: well i was able to find a 50% decreased runtime
<7> but this is rather a bug in gcc i tihnk
<7> Acovea found an option set that was an amazing 57% faster than the default options. This is a remarkable achievement. Being the suspicious sort, I investigated, and found that -funsafe-math-optimizations (a component of -ffast-math) is only effective when it is coupled with -mfpmath=387.
<7> from the acovea site
<8> well, things like -funsafe-math-optimizations might improve runtime a lot. But I'm usually to lasy to find out if it would still calculate the correct results.
<7> in my case evry output is exported to the listener :)
<7> (sound program realtime)
<7> so any minor (important) glitch is notised imediantly
<7> as i said; main cpu power goes to resampeling :)
<9> does gcc have an attribute of something to mark function 'obsolete' ? something which would say "hey you are using an obsolete function HERE" during the compilation ?
<7> zorglu_: yes but it's not even gcc specgic
<7> zorglu_: #warn iirc
<9> #warn seems similar to #error
<9> aka it will display a message, only when compiled
<7> zorglu_: isn't that the point then ?
<8> zorglu_: __attribute__ ((depreceated)) or simular
<9> not when using a function marked as obsolete....
<9> depreciated is the word ?
<9> i will look for it :)
<7> ow right sorry :)
<9> http://www.psc.edu/general/software/packages/gcc/manual/gcc_78.html#SEC81 <- give nothing about it :)
<7> warn is limited by #functions
<8> deprecated it is
<9> http://gcc.gnu.org/onlinedocs/gcc/Function-Attributes.html <- a more authoritative version do gives "deprecated"
<9> thanks for your help
<9> warn_unused_result <- wow i have been looking for such thing for a long time :)
<8> you know about -Wmissing-prototypes and -Wstrict-prototypes. They are also very nice for C if you want a strict compiler.
<8> sadly __attribute__((nonnull) was mostly a noop last I looked.
<7> noshadow: are these attibutes also valid on other compilers ?
<8> Qantourisc: likely not. Some are not supported in older gccs, though.
<9> i code in c++ so i dont handle missing prototypes :)
<9> warning: get_remote_addr is deprecated (declared at ../.include/ <- and it works :)
<8> best make some macro for it, so people using other compilers can disable it globally. (Or put it into a #if __GNUC__ > 3 block and defined to be empty otherwise)
<8> well, most should already work with 3, > 3 should only be needed for sentinell and things like that.
<9> yep that what i will do, for compatibility and for readability too :)
<9> warn_unused_result <- this one seems to fails or i miss something
<9> f() attr(warn_unused_result)
<9> and later, return f()
<9> then the compile mark the result as unused and put a warning
<9> but the result is used, as it is returned
<9> or i miss something there
<8> is it also used by the function calling the function? perhaps inlining is tricking you here.
<8> gcc also has a lot of "might be used uninitialized" false positives due to problems with those warnings and inlining...
<9> maybe
<9> i just wrote a standalone case, and it seems to work ok
<9> http://pastebin.com/852743 <- here is the case
<9> well this is the working case :)
<10> hey all, can someone tell me what does the following error message means: "/usr/bin/ld: can't locate file for: -lglib-2.0"
<9> this mean "you should install glib-2.0"
<10> i did
<9> then this means "you should give me the proper location of glib-2.0" :)
<10> how can i do that?
<9> well very specific to the code you compile and the distribution you are on
<9> you should try the channel specific to your distribution
<10> i'm on osx
<9> ah
<9> i dunno sorry, im a linux guy, i dont know the channel stuff for osx
<9> you can look at gcc doc for -L
<10> does gcc needs the includs of glib-2.0 or what?
<10> ah, ok
<9> i guess that in the -lglib-2.0, the char after the '-' is a L and not a i
<9> correct ?
<9> wait im sure it is a L :)
<9> and no it is not a problem of include
<10> jes it is L
<10> lowercase
<9> it is a problem of library binary
<10> could that be a problem with the install location of glib
<9> dunno, the issue in it self is simple, gcc tries to find glib library and doesnt succeed
<9> it is the solution which is more funky :)
<10> hmm, that means installing glib2 from source could solve thet problem?
<10> how can i customice the library search path for gcc on osx?
<9> see the doc for -L option
<10> zorglu_: ok :)
<11> anyone fluent with the xmmintrin.h functions ?
<12> hi all
<12> I have a stripped binary, and gdb give me an address of a function in binary code but in unstripped version, is it possible to know the stripped address considering an unstripped one ?
Return to
#gcc
or
Go to some related
logs:
#egghelp
#php
#fedora
#linux
#gaim
#lgp
y2base slow
wkrellm gkrellmd
write: Broken pipe
transcode syncinfo write error (0) broken pipe
#fedora
|
|