This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

optimization/8013: writing all bitfields in a word

>Number:         8013
>Category:       optimization
>Synopsis:       writing all bitfields in a word
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    unassigned
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Mon Sep 23 15:36:00 PDT 2002
>Originator:     John Reiser
>Release:        gcc version 3.2
Reading specs from /usr/local/gcc-3.2/lib/gcc-lib/i686-pc-linux-gnu/3.2/specs
Configured with: ../gcc-3.2/configure --prefix=/usr/local/gcc-3.2
Thread model: posix
gcc version 3.2
There is an opportunity for improved optimization when writing all the bitfields in a word ('int', 'unsigned', or 'long'): construct the entire word result in a register, then perform only one store to memory, instead of writing the fields to memory separately in turn.

The attached bitfield.i generates (with "-O" optimization)
        movl    (%ebx), %ecx
        andl    $16777215, %ecx
        movl    (%edx), %eax
        andl    $-16777216, %eax
        orl     %ecx, %eax
        movl    %eax, (%edx)
        movzbl  3(%ebx), %ecx
        andb    $127, %cl
        movzbl  3(%edx), %eax
        andb    $-128, %al
        orb     %cl, %al
        orb     $-128, %al
        movb    %al, 3(%edx)
which could be improved to
        movl    (%ebx), %eax
        orl     $1<<31, %eax
        movl    %eax, (%edx)
which is much shorter and faster.
gcc -v -save-temps -O -S bitfield.c

Content-Type: application/octet-stream; name="bitfield.i"
Content-Transfer-Encoding: base64
Content-Disposition: attachment; filename="bitfield.i"


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]