On 09/02/2015 03:00 PM, Segher Boessenkool wrote:
On Wed, Sep 02, 2015 at 01:59:58PM -0600, Jeff Law wrote:
(set (reg:CC 66 cc)
(compare:CC (and:DI (lshiftrt:DI (subreg:DI (reg/v:SI 76 [ xD.2641 ])
0)
(const_int 1 [0x1]))
(const_int 1 [0x1]))
(const_int 0 [0])))
Yea, this is an alternative form. I don't offhand remember how/why
this form appears, but it certainly does. I don't think any ports
handle this form (but I certainly have done any checks), but I believe
combine creates it primarily for internal purposes.
Combine replaces zero_ext* with equivalent shift/and patterns and tries
again, if things don't match. Targets with more generic masking insns
do not want to describe the very many cases that can be described with
zero_ext* separately.
rs6000 handles this exact pattern, btw. And I'll be very happy if we can
just drop it :-)
If I still cared, I'd probably look into this for the PA which has some
rough similarities with the PPC architecture in its bit
insertion/extraction capabilities. But the PA just isn't worth the
time :-)
cindex @code{zero_extract}, canonicalization of
@cindex @code{sign_extract}, canonicalization of
@item
Equality comparisons of a group of bits (usually a single bit) with zero
will be written using @code{zero_extract} rather than the equivalent
@code{and} or @code{sign_extract} operations.
Oh it's even documented, thanks. I do still think we should think of
changing this.
Do-able, but I suspect the fallout would be significant across the older
ports.
Jeff
