Fix constant folding and poor optimization in byte swapping code
authorTrent Piepho <xyzzy@speakeasy.org>
Tue, 8 May 2007 07:24:05 +0000 (00:24 -0700)
committerLinus Torvalds <torvalds@woody.linux-foundation.org>
Tue, 8 May 2007 18:14:59 +0000 (11:14 -0700)
commit8e2c20023f34b652605a5fb7c68bb843d2b100a8
treebd041c1762724dbbc91f4b2da3fc0716165784e5
parent02fb6149f7a64c62934c035e7635321cb9a8cf2e
Fix constant folding and poor optimization in byte swapping code

Constant folding does not work for the swabXX() byte swapping functions,
and the C versions optimize poorly.

Attempting to initialize a global variable to swab16(0x1234) or put
something like "case swab32(42):" in a switch statement will not compile.
It can work, swab.h just isn't doing it correctly.  This patch fixes that.

Contrary to the comment in asm-i386/byteorder.h, gcc does not recognize the
"C" version of swab16 and turn it into efficient code.  gcc can do this,
just not with the current code.  The simple function:

u16 foo(u16 x) { return swab16(x); }

Would compile to:
        movzwl  %ax, %eax
        movl    %eax, %edx
        shrl    $8, %eax
        sall    $8, %edx
        orl     %eax, %edx

With this patch, it will compile to:
        rolw    $8, %ax

I also attempted to document the maze different macros/inline functions
that are used to create the final product.

Signed-off-by: Trent Piepho <xyzzy@speakeasy.org>
Cc: Francois-Rene Rideau <fare@tunes.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
include/linux/byteorder/swab.h