Code quality with uint64_t on x86

Code quality with uint64_t on x86

[Sebastian Biallas]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello!

With uint64_t types I encountered some suboptimal code with gcc 4.1:

$ uname -a
Linux meisenmann 2.6.16 #1 PREEMPT Wed Mar 22 20:40:14 CET 2006 i686 AMD
Athlon(tm) Processor AuthenticAMD GNU/Linux

$ gcc -v
Using built-in specs.
Target: i686-pc-linux-gnu
Configured with: /var/tmp/portage/gcc-4.1.0/work/gcc-4.1.0/configure
- --prefix=/usr --bindir=/usr/i686-pc-linux-gnu/gcc-bin/4.1.0
- --includedir=/usr/lib/gcc/i686-pc-linux-gnu/4.1.0/include
- --datadir=/usr/share/gcc-data/i686-pc-linux-gnu/4.1.0
- --mandir=/usr/share/gcc-data/i686-pc-linux-gnu/4.1.0/man
- --infodir=/usr/share/gcc-data/i686-pc-linux-gnu/4.1.0/info
- --with-gxx-include-dir=/usr/lib/gcc/i686-pc-linux-gnu/4.1.0/include/g++-v4
- --host=i686-pc-linux-gnu --build=i686-pc-linux-gnu --disable-altivec
- --enable-nls --without-included-gettext --with-system-zlib
- --disable-checking --disable-werror --disable-libunwind-exceptions
- --disable-multilib --disable-libmudflap --disable-libssp
- --disable-libgcj --enable-languages=c,c++ --enable-shared
- --enable-threads=posix --enable-__cxa_atexit --enable-clocale=gnu
Thread model: posix
gcc version 4.1.0 (Gentoo 4.1.0)

$ cat t.cc
#include <stdint.h>

typedef uint32_t uint32;
typedef uint64_t uint64;

static inline uint32 delinearize(uint32 d)
{
        return d*0x8088405+1;   /* there's magic in here... */
}

uint64 delinearize64(uint64 d)
{
        return (uint64(delinearize(d>>32))<<32) | delinearize(d);
	/* there's less magic in here... */
}

$ g++ -S -O2 -o t.S t.cc

$ cat t.S
        .file   "t.cc"
        .text
        .align 2
        .p2align 4,,15
.globl _Z13delinearize64y
        .type   _Z13delinearize64y, @function
_Z13delinearize64y:
.LFB14:
        pushl   %ebp
.LCFI0:
        movl    %esp, %ebp
.LCFI1:
        pushl   %ebx
.LCFI2:
        movl    12(%ebp), %ebx
        movl    8(%ebp), %ecx
        movl    %ebx, %edx                            [0]
        movl    %ecx, %eax
        movl    %edx, %eax
        imull   $134775813, %eax, %eax
        xorl    %edx, %edx                            [1]
        imull   $134775813, %ecx, %ecx
        xorl    %edx, %edx                            [1]
        popl    %ebx
        incl    %eax
        popl    %ebp
        movl    %eax, %edx
        incl    %ecx
        movl    $0, %eax                              [3]
        orl     %ecx, %eax
        ret
.LFE14:
        .size   _Z13delinearize64y, .-_Z13delinearize64y
.globl __gxx_personality_v0
        .ident  "GCC: (GNU) 4.1.0 (Gentoo 4.1.0)"
        .section        .note.GNU-stack,"",@progbits


Well, this doesn't look good. For example:
[0] What's all this moving for? imull is one of the most versatile x86
instructions :)
Additional points for clobbering %ebx for no real use.
[1] Useless use of xor award
[2] LOL

Are these known bugs or should I file a PR?


Additionally, this gets worse if I remove the "inline" keyword:

[same program with "inline" removed]
$ cat t2.cc
       .file   "t2.cc"
        .text
        .align 2
        .p2align 4,,15
        .type   _Z11delinearizej, @function
_Z11delinearizej:
.LFB13:
        imull   $134775813, %eax, %eax
        pushl   %ebp
.LCFI0:
        movl    %esp, %ebp
.LCFI1:
        popl    %ebp
        incl    %eax
        ret
.LFE13:
        .size   _Z11delinearizej, .-_Z11delinearizej
.globl __gxx_personality_v0
        .align 2
        .p2align 4,,15
.globl _Z13delinearize64y
        .type   _Z13delinearize64y, @function
_Z13delinearize64y:
.LFB14:
        pushl   %ebp
.LCFI2:
        movl    %esp, %ebp
.LCFI3:
        subl    $16, %esp
.LCFI4:
        movl    %esi, -4(%ebp)
.LCFI5:
        movl    12(%ebp), %esi
        movl    %ebx, -8(%ebp)
.LCFI6:
        movl    8(%ebp), %ebx
        movl    %esi, %edx
        movl    %ebx, %eax
        movl    %edx, %eax
        xorl    %edx, %edx
        call    _Z11delinearizej
        movl    $0, -12(%ebp)
        movl    -12(%ebp), %edx
        movl    %eax, -16(%ebp)
        movl    -16(%ebp), %eax
        movl    %eax, %edx
        movl    $0, %eax
        movl    %eax, -16(%ebp)
        movl    %ebx, %eax
        movl    %edx, -12(%ebp)
        call    _Z11delinearizej
        movl    -12(%ebp), %edx
        movl    -8(%ebp), %ebx
        movl    -4(%ebp), %esi
        movl    %eax, %ecx
        movl    -16(%ebp), %eax
        movl    %ebp, %esp
        popl    %ebp
        orl     %ecx, %eax
        ret
.LFE14:
        .size   _Z13delinearize64y, .-_Z13delinearize64y
        .ident  "GCC: (GNU) 4.1.0 (Gentoo 4.1.0)"
        .section        .note.GNU-stack,"",@progbits


Why doesn't gcc automatically inline "static uint32 delinearize(uint32
d)" here?


Sebastian


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2.2 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iQEVAwUBRDhu3v81M8QtvOSJAQJnSAgAl6AQhd7vaIvX3Uzz21D6jcsCXQJ2Q07v
V6hXuJSwXdT90CM0ns06E9TeRgwDFMeqTTtocIK7APOxcG0MyAI9ccU8y0v7aMEv
gwmVvCzsjnj4hXuXqy6GIAwhCxoH/Y4k7F868wFGGopgdFAbwhdyaRohoJyPoA7j
WiVb8sSFuDWI+72XUDZGuHhu3jdGpif7Hg063URnfG1T5kQzE6hvF6aoPQHnlzGA
biGmAyld7U9h9xpQ7l6zlXVSY77btUGXYMXnTwp27g/UIxzZDYd7MBlgKt7Zou5h
A3JQj390tDxVgwxOb2ZKJZ2LcFZrU9nOHhKr6ZEjHQvdwdSvoXPMng==
=KcR4
-----END PGP SIGNATURE-----