On Wed, Nov 30, 2016 at 02:12:35PM +0100, Dominik Vogt wrote:
> On Wed, Nov 23, 2016 at 02:22:07PM +0000, Segher Boessenkool wrote:
> > r242414, for PR77881, introduces some bugs (PR78390, PR78438, PR78477).
> > It all has the same root cause: that patch makes combine convert every
> > lowpart subreg of a logical shift right to a zero_extract. This cannot
> > work at all if it is not a constant shift, and it has to be a bit more
> > careful exactly which bits it extracts.
> >
> > Tested on powerpc64-linux {-m32,-m64} (where it fixes the regression
> > c-c++-common/torture/vector-compare-1.c fails at -O1, and where it also
> > has a bootstrap failure with some other patches). Also tested that the
> > x86_64 compiler still generates the wanted code for the PR77881 testcase.
>
> Is this a side effect of the patch series?
I do not know; I cannot tell just from this, and there is no source
snippet to try. Maybe Michael can tell?
> Trying 7 -> 9:
> ...
> Failed to match this instruction:
> (set (reg:SI 68 [ v_or ])
> (ior:SI (subreg:SI (zero_extract:DI (reg:DI 2 %r2 [ v_x ])
> ^^^^^^^^^^^^^^^^^^^^^^^^^^
> (const_int 16 [0x10])
> (const_int 0 [0])) 4)
> ^^^
> (reg:SI 70 [ v_and1 ])))
>
> Shouldn't this be simply
>
> ...
> (ior:SI (zero_extract:SI (reg:DI) (16) (0)))
> ...
That seems nicer, sure. OTOH that will never match on a target that
does not have zero_extract:SI from :DI. *_extracts are nasty.
Segher
