redundant bitmap_bit_p followed by bitmap_clear_bit [was: Re: [COMMITTED] Kill second order relations in the path solver.]

[Bernhard Reutner-Fischer]

On Mon, 1 Nov 2021 15:21:03 +0100
Aldy Hernandez <aldyh@redhat.com> wrote:

> I'm not convinced this makes the code clearer to read, especially if
> it's not on a critical path.  But if you feel strongly, please submit
> a patch ;-).

No i don't feel strongly about it.
Compiling e.g. -O2 ira.o
# Overhead       Samples  Command  Shared Object  Symbol                   
# ........  ............  .......  .............  .........................
#
   100.00%          4197  cc1plus  cc1plus        [.] mark_reachable_blocks
   100.00%         22000  cc1plus  cc1plus        [.] path_oracle::killing_def
and the mark_elimination is reload.
So it's not just a handful of calls saved but some. And it would be
smaller code as it saves a call. Well maybe another day.
thanks,
> 
> Aldy
> 
> On Mon, Nov 1, 2021 at 3:10 PM Bernhard Reutner-Fischer
> <rep.dot.nop@gmail.com> wrote:
> >
> > On Thu, 28 Oct 2021 01:55:30 +0200
> > Bernhard Reutner-Fischer <rep.dot.nop@gmail.com> wrote:
> >  
> > > On Wed, 27 Oct 2021 20:13:21 +0200
> > > Aldy Hernandez via Gcc-patches <gcc-patches@gcc.gnu.org> wrote:  
> >  
> > > > @@ -1306,6 +1307,24 @@ path_oracle::killing_def (tree ssa)
> > > >    ptr->m_next = m_equiv.m_next;
> > > >    m_equiv.m_next = ptr;
> > > >    bitmap_ior_into (m_equiv.m_names, b);
> > > > +
> > > > +  // Walk the relation list an remove SSA from any relations.  
> > >
> > > s/an /and /
> > >  
> > > > +  if (!bitmap_bit_p (m_relations.m_names, v))
> > > > +    return;
> > > > +
> > > > +  bitmap_clear_bit (m_relations.m_names, v);  
> > >
> > > IIRC bitmap_clear_bit returns true if the bit was set, false otherwise,
> > > so should be used as if(!bitmap_clear_bit) above.  
> >  
> > > > +  relation_chain **prev = &(m_relations.m_head);  
> > >
> > > s/[()]//
> > > thanks,  
> >
> > There seems to be two other spots where a redundant bitmap_bit_p checks
> > if we want to bitmap_clear_bit. In dse and ira.
> > Like:
> > $ cat ~/coccinelle/gcc_bitmap_bit_p-0.cocci ; echo EOF
> > // replace redundant bitmap_bit_p() bitmap_clear_bit with the latter
> > @ rule1 @
> > identifier fn;
> > expression bitmap, bit;
> > @@
> >
> > fn(...) {
> > <...
> > (
> > -if (bitmap_bit_p (bitmap, bit))
> > +if (bitmap_clear_bit (bitmap, bit))
> > {
> >   ...
> > -  bitmap_clear_bit (bitmap, bit);
> >   ...
> > }
> > |
> > -if (bitmap_bit_p (bitmap, bit))
> > +if (bitmap_clear_bit (bitmap, bit))
> > {
> >   ...
> > }
> > ...
> > -bitmap_clear_bit (bitmap, bit);
> > )  
> > ...>  
> > }
> > EOF
> > $ find gcc/ -type f -a \( -name "*.c" -o -name "*.cc" \) -a \( ! -path "gcc/testsuite/*" -a ! -path "gcc/contrib/*" \) -exec spatch -sp_file ~/coccinelle/gcc_bitmap_bit_p-0.cocci --show-diff {} \;
> > diff =
> > --- gcc/dse.c
> > +++ /tmp/cocci-output-1104419-443759-dse.c
> > @@ -3238,9 +3238,8 @@ mark_reachable_blocks (sbitmap unreachab
> >    edge e;
> >    edge_iterator ei;
> >
> > -  if (bitmap_bit_p (unreachable_blocks, bb->index))
> > +  if (bitmap_clear_bit(unreachable_blocks, bb->index))
> >      {
> > -      bitmap_clear_bit (unreachable_blocks, bb->index);
> >        FOR_EACH_EDGE (e, ei, bb->preds)
> >         {
> >           mark_reachable_blocks (unreachable_blocks, e->src);
> > diff =
> > --- gcc/ira.c
> > +++ /tmp/cocci-output-1104678-d8679a-ira.c
> > @@ -2944,17 +2944,15 @@ mark_elimination (int from, int to)
> >    FOR_EACH_BB_FN (bb, cfun)
> >      {
> >        r = DF_LR_IN (bb);
> > -      if (bitmap_bit_p (r, from))
> > +      if (bitmap_clear_bit(r, from))
> >         {
> > -         bitmap_clear_bit (r, from);
> >           bitmap_set_bit (r, to);
> >         }
> >        if (! df_live)
> >          continue;
> >        r = DF_LIVE_IN (bb);
> > -      if (bitmap_bit_p (r, from))
> > +      if (bitmap_clear_bit(r, from))
> >         {
> > -         bitmap_clear_bit (r, from);
> >           bitmap_set_bit (r, to);
> >         }
> >      }
> > # in ira.c one would have to fixup the curly braces manually
> > PS: coccinelle seems to ruin the spaces before braces in the '+' even
> > though i have written them correctly according to GNU style..
> >  
>