Thanks, Phil.
This will be done a bit later with a proper fix which counts bit more things 
around shift.
I started a thread in gcc list where possible fix will be discussed.
~d

On Sunday 18 July 2004 04:26, Philip Sambles wrote:
> Hi,
>
> Here is the patch, it is for gcc-3.4.1 but it does work with gcc-3.4.0.  I
> have tested it under the Cygwin environment and it appears to produce valid
> output.  The peephole optimisation to replace the mov/and with a bit
> operation isn't working even though (at a quick look) it seems to match the
> generated RTL.
>
> If you're planning on using this for anything important I would recommend
> that you test it more thoroughly.
>
> To stop the compiler doing the conversion to shifts you need to add
> '-fno-fold-single-bit' to the compiler command line.
>
> Regards
>
> Phil.

Reply via email to