| Message ID | 20100924165441.GB31666@virgil.arch.suse.de |
|---|---|
| State | New |
| Headers | show |
On Fri, 24 Sep 2010, Martin Jambor wrote: > Hi, > > with SRA producing MEM_REFs to access components of aggregates, > zero-sized trailing bit-fields are no longer a problem (PR 43191, > better description: > http://gcc.gnu.org/ml/gcc-patches/2010-03/msg00207.html) and so there > is no need to handle them specially. > > Bootstrapped and tested on x86_64-linux. OK for trunk? Ok. Thanks, Richard. > Thanks, > > Martin > > > 2010-09-24 Martin Jambor <mjambor@suse.cz> > > * tree-sra.c (type_consists_of_records_p): Do not check for trailing > zero sized bit-fields. > > Index: mine/gcc/tree-sra.c > =================================================================== > --- mine.orig/gcc/tree-sra.c > +++ mine/gcc/tree-sra.c > @@ -815,14 +815,12 @@ create_access (tree expr, gimple stmt, b > > /* Return true iff TYPE is a RECORD_TYPE with fields that are either of gimple > register types or (recursively) records with only these two kinds of fields. > - It also returns false if any of these records has a zero-size field as its > - last field or has a bit-field. */ > + It also returns false if any of these records contains a bit-field. */ > > static bool > type_consists_of_records_p (tree type) > { > tree fld; > - bool last_fld_has_zero_size = false; > > if (TREE_CODE (type) != RECORD_TYPE) > return false; > @@ -838,13 +836,8 @@ type_consists_of_records_p (tree type) > if (!is_gimple_reg_type (ft) > && !type_consists_of_records_p (ft)) > return false; > - > - last_fld_has_zero_size = tree_low_cst (DECL_SIZE (fld), 1) == 0; > } > > - if (last_fld_has_zero_size) > - return false; > - > return true; > } > > >
On Fri, Sep 24, 2010 at 6:54 PM, Martin Jambor <mjambor@suse.cz> wrote: > > * tree-sra.c (type_consists_of_records_p): Do not check for trailing > zero sized bit-fields. Test case? :-D Ciao! Steven
Hi, On Fri, Sep 24, 2010 at 07:07:01PM +0200, Steven Bosscher wrote: > On Fri, Sep 24, 2010 at 6:54 PM, Martin Jambor <mjambor@suse.cz> wrote: > > > > * tree-sra.c (type_consists_of_records_p): Do not check for trailing > > zero sized bit-fields. > > > Test case? :-D Two testcases are already in trunk: testsuite/gcc.c-torture/compile/pr43164.c and testsuite/gcc.c-torture/compile/pr43191.c. See the email I linked when posting the patch (http://gcc.gnu.org/ml/gcc-patches/2010-03/msg00207.html) for more details. I'll re-test the patch today with a fresh checkout of trunk and commit it afterwards. Thanks, Martin
Index: mine/gcc/tree-sra.c =================================================================== --- mine.orig/gcc/tree-sra.c +++ mine/gcc/tree-sra.c @@ -815,14 +815,12 @@ create_access (tree expr, gimple stmt, b /* Return true iff TYPE is a RECORD_TYPE with fields that are either of gimple register types or (recursively) records with only these two kinds of fields. - It also returns false if any of these records has a zero-size field as its - last field or has a bit-field. */ + It also returns false if any of these records contains a bit-field. */ static bool type_consists_of_records_p (tree type) { tree fld; - bool last_fld_has_zero_size = false; if (TREE_CODE (type) != RECORD_TYPE) return false; @@ -838,13 +836,8 @@ type_consists_of_records_p (tree type) if (!is_gimple_reg_type (ft) && !type_consists_of_records_p (ft)) return false; - - last_fld_has_zero_size = tree_low_cst (DECL_SIZE (fld), 1) == 0; } - if (last_fld_has_zero_size) - return false; - return true; }