Jump to content
Read the Funtoo Newsletter: Summer 2023 ×

Sandro

Members
  • Posts

    466
  • Joined

  • Last visited

  • Days Won

    20

Posts posted by Sandro

  1. [ot]

    Cardinal is a real great Master of Funtoo :)

    [/ot]

     

    But ... Cardinal, i see thath you use the "new tecnique" used by Gentoo about te /etc/portage/package.*/package_name

     

    Do You think is a better solution than to have only one file for /etc/portage/package.* ?

     

    Thank You :)

  2.  

    I haven't had any performance problems with Clang, particularly using LTO.  On the theory side of things, I think Clang and GCC perform better for different things.  I have noticed that Clang tends to do better constant folding and strength reductions in the generated code that I've looked at.  That may have changed with more recent GCC releases though.

     

    If you're interested in performance, LTO with the same compiler will probably give a better performance increase than switching to another compiler.  I know the binary Clang emits for vlc using LTO is more than 100KiB smaller than without LTO.  It might be worth benchmarking.

     

    It's a few versions behind for Clang and GCC, but one of the better compiler comparisons I read was for compiling Firefox.  GCC with LTO did generate a smaller and faster binary than Clang with LTO, but again it's a few versions behind for both compilers.

     

     

    Grazie... although my Italian is honestly worse than my Spanish :).

     

    Oh this discussion take a relevant interesting (for me :P) ; I'm happy that You're a very kind man :); i'm a "little penguin"; but i want to learn as more as possible :)

     

    viewing this benchs ( http://www.phoronix.com/scan.php?page=article&item=clang-37-gcc52&num=1) it seems that llvm creates better binary code for some applications ......

    I'd like to try with Your method ..... :)

     

    but there is also 

    * dev-lang/icc
         Available versions:  ~13.0.0.079^m ~13.0.1.117^m ~13.1.2.146^m ~13.1.3.163^m ~13.1.5.192^m ~14.0.0.080^m ~14.0.1.106^m ~14.0.2.144^m ~14.0.3.174^m ~15.0.0.090^m ~15.0.1.133^m ~15.0.2.164^m ~15.0.3.187^m {eclipse examples multilib LINGUAS="ja"}
         Homepage:            http://software.intel.com/en-us/articles/intel-composer-xe/
         Description:         Intel C/C++ Compiler
    
    

    hehe i've "stoled" (copied in a text file) your make.conf (i wanna study it)... cause one of this days i want to make experience with llvm; than i will tell to you about some benchmarks that i can make on my system :)

     

    Only one question: about LDFLAGS, i've "LDFLAGS="${LDFLAGS} -Wl,--has-stile=gnu". so to have:

    ci74771ht ~ # emerge --info|grep LDFLAGS
    LDFLAGS="-Wl,-O1 -Wl,--sort-common -Wl,--as-needed -Wl,--hash-style=gnu"
    
    

    I can see that you've declared also the -march ..... it is necessary using llvm ?

     

    Excuse me for my questions .... 

  3. I know that about the time implied to compile an ebuild may be faster with llvm; but i'm interested not for compiling time, but about the binary code generated by a compiler; in Phoronix for example (excuse me but i've not a link at this moment) there are comparisons about gcc vs llvm/clang.

     

    also comared to ICC (Intel C Compiler) and also Open64 (AMD open source compiler, present since few time ago also in portage tree).

     

    The resultstold to me that gcc is the more performant compiler (about binary code generated) (not in all cases but considering the average....).

     

    But ... since clang is in evolution (but also gcc), the answers may be done looking the benchs :)

     

    Try to "take a look" if You want in Phoronix website :)

     

    Hi ... and hope not have bored You :)

     

    With Estimate.

     

    Sandro :)

     

    PS: Excuse me if i'm little "off topic" :|. May be Sympathic to open a new thread about various compiler performancies ... :) 

     

    Ciao :D (from Italy) :)

  4. Hi uxcn :)

     

    So you've choiced clang/llvm as your "default compiler" ?

    excuse me for my question ..... why that choiche ?

     

    (For example i've experimented PC-BSD 11.X and is very slower vs 10.X (that use GCC as compiler))

     

    mumble... mumble...

     

    Thank You for any suggestion :)

     

    I think that if you declare the "-march" , the "-mtuse" is redundant..... (not utile?)

     

    I see a -j16 .... WOW !!! :P :P :P your system is based on a dual Xeon (with 4 cores / 8 threads) both ?

     

    Thanks for suggestions :)

  5. @Cardinal: Thank You Cardinal: I'm trying some experiment: for example with Funtoo "Stable" on my system the 0.6.8 is correctly compiled.

    Now I try with "Stable" if also the 0.7_beta_! compiles or not on Stable.

     

    _________________________________________________________________

     

    Ok ..... with Stable also the last 0.7_? compiled succesfully ...... 

     

    Now I must terminate the compilation of Chromium .... then I make experiemnts also in "current".

     

    I wish my little feedback can be utile :)

  6. Hi; i've compiled gentoo-sources-4.3.0 with "experimental USE flag".

    I had to unmask nvidia-drivers 358.09 cause the 355.11-r2 doesn't run with this kernel.

     

    But i've seen in stdout and in the use one new features: about the use is "kms" (Kernel Mode Setting).

    In stdouth , now,

    1) Install nvidia module

    2) install nvidia-uvm module

    3) install nvidia-modeset module

     

    How to take advantages from this new feature ?

     

    Thank You for any suggestion.

  7. I think that the simplicity of Portage can really visible when you must change "destination" of your computer.

    With FreeBSD you must use "make config-recursive" and have in your mind if one charachteristic (rappresented in *too by the global or local USE flags) if it is enabled or disabled.

    For me Portage is more rational than FreeBSD ports technology.

    And there are a lot of instruments to see if the system is "consistent" (e.g. revdep-rebuild).

     

    For me Portage is the most evoluted package manager in this planet. there is also Paludis for *too.

     

    Another advantage is that with Funtoo (or Gentoo), 98% of the ebuilds supports full multithreading; in FreeBSD you can't estabilished a jobserver .... most programs use the default (-j2).

     

    try in freebsd to install for example htop:

    then

    whereis htop

    so make config-recursive

    then make install clean

    but it will be single-threaded.

     

    Try to type make -j9 install && make clean .... you'll got error.

     

    More time to compile it..... more time to change it and more difficulties.

     

    But This is only my tought.

     

    The flexibility that give portage is veri high .....

     

    Take a look: 

    this "electronic guitar" has Gentoo inside :)

     

    Bye :)

  8. But .... may be a good idea to insert in genkernel the possibility to runs make clean after kernel & initramfs compilation ?

     

    I can Try to use /dev/shm .......

     

    I'll tell to You the results.

     

    For example debian-sources wants 14 GiB.

     

    Now I make some experiment !!!!

     

    Then I reports if i've issues.

     

    Thank You :)

×
×
  • Create New...