diff mbox series

[rs6000] Early gimple folding of vec_mergeh and vec_mergel for float

Message ID 1533669898.4452.11.camel@brimstone.rchland.ibm.com
State New
Headers show
Series [rs6000] Early gimple folding of vec_mergeh and vec_mergel for float | expand

Commit Message

will schmidt Aug. 7, 2018, 7:24 p.m. UTC
Hi,
   This adds support for gimple folding of vec_mergeh and vec_mergel
for float and double types.   Support for the integral types is already
in-tree.
    
This change includes updates to the fold_mergehl_helper function to handle
creating the (integral) permute vector when the vector type non integer.

Relevant testcases are already in-tree.

OK for trunk?

Thanks,
-Will
    
[gcc]
    
2018-08-07 Will Schmidt  <will_schmidt@vnet.ibm.com>
    
	* config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
	allow folding of mergeh() and mergel() for the float and double types.
	(fold_mergehl_helper): Rework to handle building a permute tree
	for float vectors.

Comments

Segher Boessenkool Aug. 7, 2018, 10:59 p.m. UTC | #1
Hi!

On Tue, Aug 07, 2018 at 02:24:58PM -0500, Will Schmidt wrote:
>    This adds support for gimple folding of vec_mergeh and vec_mergel
> for float and double types.   Support for the integral types is already
> in-tree.

> +  /* The permute_type will match the lhs for integral types.  For double and
> +     float types, the permute type needs to map to the V2 or V4 type that
> +     matches size.  */
> +  tree permute_type;
> +  if (INTEGRAL_TYPE_P (TREE_TYPE (lhs_type)))
> +    permute_type = lhs_type;
> +  else
> +    if (TREE_TYPE (lhs_type) == TREE_TYPE (V2DF_type_node))
> +      permute_type = V2DI_type_node;
> +    else if (TREE_TYPE (lhs_type) == TREE_TYPE (V4SF_type_node))
> +      permute_type = V4SI_type_node;
> +    else
> +      gcc_unreachable ();

Please write this as

  if (INTEGRAL_TYPE_P (TREE_TYPE (lhs_type)))
    permute_type = lhs_type;
  else if (TREE_TYPE (lhs_type) == TREE_TYPE (V2DF_type_node))
    permute_type = V2DI_type_node;
  else if (TREE_TYPE (lhs_type) == TREE_TYPE (V4SF_type_node))
    permute_type = V4SI_type_node;
  else
    gcc_unreachable ();

or, if you want to emphasize integer vs. float:

  if (INTEGRAL_TYPE_P (TREE_TYPE (lhs_type)))
    permute_type = lhs_type;
  else
    {
      if (TREE_TYPE (lhs_type) == TREE_TYPE (V2DF_type_node))
        permute_type = V2DI_type_node;
      else if (TREE_TYPE (lhs_type) == TREE_TYPE (V4SF_type_node))
        permute_type = V4SI_type_node;
      else
        gcc_unreachable ();
    }

Okay for trunk with that changed.  Thanks!


Segher
Richard Biener Aug. 17, 2018, 2 p.m. UTC | #2
On Wed, Aug 8, 2018 at 12:59 AM Segher Boessenkool
<segher@kernel.crashing.org> wrote:
>
> Hi!
>
> On Tue, Aug 07, 2018 at 02:24:58PM -0500, Will Schmidt wrote:
> >    This adds support for gimple folding of vec_mergeh and vec_mergel
> > for float and double types.   Support for the integral types is already
> > in-tree.
>
> > +  /* The permute_type will match the lhs for integral types.  For double and
> > +     float types, the permute type needs to map to the V2 or V4 type that
> > +     matches size.  */
> > +  tree permute_type;
> > +  if (INTEGRAL_TYPE_P (TREE_TYPE (lhs_type)))
> > +    permute_type = lhs_type;
> > +  else
> > +    if (TREE_TYPE (lhs_type) == TREE_TYPE (V2DF_type_node))
> > +      permute_type = V2DI_type_node;
> > +    else if (TREE_TYPE (lhs_type) == TREE_TYPE (V4SF_type_node))
> > +      permute_type = V4SI_type_node;
> > +    else
> > +      gcc_unreachable ();
>
> Please write this as
>
>   if (INTEGRAL_TYPE_P (TREE_TYPE (lhs_type)))
>     permute_type = lhs_type;
>   else if (TREE_TYPE (lhs_type) == TREE_TYPE (V2DF_type_node))
>     permute_type = V2DI_type_node;
>   else if (TREE_TYPE (lhs_type) == TREE_TYPE (V4SF_type_node))
>     permute_type = V4SI_type_node;
>   else
>     gcc_unreachable ();
>
> or, if you want to emphasize integer vs. float:
>
>   if (INTEGRAL_TYPE_P (TREE_TYPE (lhs_type)))
>     permute_type = lhs_type;
>   else
>     {
>       if (TREE_TYPE (lhs_type) == TREE_TYPE (V2DF_type_node))

Are you sure lhs_type is never qualified?  That is, for a GIMPLE folder
I'd have expected

  if (types_compatible_p (TREE_TYPE (lhs_type), TREE_TYPE (V2DF_type_node)))

for GENERIC

  if (TYPE_MAIN_VARIANT (TREE_TYPE (lhs_type)) == TREE_TYPE (V2DF_type_node))

Richard.

>         permute_type = V2DI_type_node;
>       else if (TREE_TYPE (lhs_type) == TREE_TYPE (V4SF_type_node))
>         permute_type = V4SI_type_node;
>       else
>         gcc_unreachable ();
>     }
>
> Okay for trunk with that changed.  Thanks!
>
>
> Segher
will schmidt Aug. 17, 2018, 3:02 p.m. UTC | #3
On Fri, 2018-08-17 at 16:00 +0200, Richard Biener wrote:
> On Wed, Aug 8, 2018 at 12:59 AM Segher Boessenkool
> <segher@kernel.crashing.org> wrote:
> >
> > Hi!
> >
> > On Tue, Aug 07, 2018 at 02:24:58PM -0500, Will Schmidt wrote:
> > >    This adds support for gimple folding of vec_mergeh and vec_mergel
> > > for float and double types.   Support for the integral types is already
> > > in-tree.
> >
> > > +  /* The permute_type will match the lhs for integral types.  For double and
> > > +     float types, the permute type needs to map to the V2 or V4 type that
> > > +     matches size.  */
> > > +  tree permute_type;
> > > +  if (INTEGRAL_TYPE_P (TREE_TYPE (lhs_type)))
> > > +    permute_type = lhs_type;
> > > +  else
> > > +    if (TREE_TYPE (lhs_type) == TREE_TYPE (V2DF_type_node))
> > > +      permute_type = V2DI_type_node;
> > > +    else if (TREE_TYPE (lhs_type) == TREE_TYPE (V4SF_type_node))
> > > +      permute_type = V4SI_type_node;
> > > +    else
> > > +      gcc_unreachable ();
> >
> > Please write this as
> >
> >   if (INTEGRAL_TYPE_P (TREE_TYPE (lhs_type)))
> >     permute_type = lhs_type;
> >   else if (TREE_TYPE (lhs_type) == TREE_TYPE (V2DF_type_node))
> >     permute_type = V2DI_type_node;
> >   else if (TREE_TYPE (lhs_type) == TREE_TYPE (V4SF_type_node))
> >     permute_type = V4SI_type_node;
> >   else
> >     gcc_unreachable ();
> >
> > or, if you want to emphasize integer vs. float:
> >
> >   if (INTEGRAL_TYPE_P (TREE_TYPE (lhs_type)))
> >     permute_type = lhs_type;
> >   else
> >     {
> >       if (TREE_TYPE (lhs_type) == TREE_TYPE (V2DF_type_node))
> 
> Are you sure lhs_type is never qualified?

For the V2DF and V4SF types, I am mostly sure, but since you bring it up
I will admit I am not positive.  :-)       

>                                     That is, for a GIMPLE folder
> I'd have expected
>
>   if (types_compatible_p (TREE_TYPE (lhs_type), TREE_TYPE (V2DF_type_node)))

> for GENERIC
> 
>   if (TYPE_MAIN_VARIANT (TREE_TYPE (lhs_type)) == TREE_TYPE (V2DF_type_node))

Either/both of those seem more robust than what I had come up with, I'll
plan to make an update here.

Any guidance on whether I should prefer "types_compatible_p" versus the
GENERIC "TYPE_MAIN_VARIANT" ?   I see more references to the latter in
rs6000.c , but the former seems to make better sense to me just by its
name.

Thanks for the review, 
-Will


> Richard.
> 
> >         permute_type = V2DI_type_node;
> >       else if (TREE_TYPE (lhs_type) == TREE_TYPE (V4SF_type_node))
> >         permute_type = V4SI_type_node;
> >       else
> >         gcc_unreachable ();
> >     }
> >
> > Okay for trunk with that changed.  Thanks!
> >
> >
> > Segher
>
Richard Biener Aug. 17, 2018, 3:16 p.m. UTC | #4
On August 17, 2018 5:02:53 PM GMT+02:00, Will Schmidt <will_schmidt@vnet.ibm.com> wrote:
>On Fri, 2018-08-17 at 16:00 +0200, Richard Biener wrote:
>> On Wed, Aug 8, 2018 at 12:59 AM Segher Boessenkool
>> <segher@kernel.crashing.org> wrote:
>> >
>> > Hi!
>> >
>> > On Tue, Aug 07, 2018 at 02:24:58PM -0500, Will Schmidt wrote:
>> > >    This adds support for gimple folding of vec_mergeh and
>vec_mergel
>> > > for float and double types.   Support for the integral types is
>already
>> > > in-tree.
>> >
>> > > +  /* The permute_type will match the lhs for integral types. 
>For double and
>> > > +     float types, the permute type needs to map to the V2 or V4
>type that
>> > > +     matches size.  */
>> > > +  tree permute_type;
>> > > +  if (INTEGRAL_TYPE_P (TREE_TYPE (lhs_type)))
>> > > +    permute_type = lhs_type;
>> > > +  else
>> > > +    if (TREE_TYPE (lhs_type) == TREE_TYPE (V2DF_type_node))
>> > > +      permute_type = V2DI_type_node;
>> > > +    else if (TREE_TYPE (lhs_type) == TREE_TYPE (V4SF_type_node))
>> > > +      permute_type = V4SI_type_node;
>> > > +    else
>> > > +      gcc_unreachable ();
>> >
>> > Please write this as
>> >
>> >   if (INTEGRAL_TYPE_P (TREE_TYPE (lhs_type)))
>> >     permute_type = lhs_type;
>> >   else if (TREE_TYPE (lhs_type) == TREE_TYPE (V2DF_type_node))
>> >     permute_type = V2DI_type_node;
>> >   else if (TREE_TYPE (lhs_type) == TREE_TYPE (V4SF_type_node))
>> >     permute_type = V4SI_type_node;
>> >   else
>> >     gcc_unreachable ();
>> >
>> > or, if you want to emphasize integer vs. float:
>> >
>> >   if (INTEGRAL_TYPE_P (TREE_TYPE (lhs_type)))
>> >     permute_type = lhs_type;
>> >   else
>> >     {
>> >       if (TREE_TYPE (lhs_type) == TREE_TYPE (V2DF_type_node))
>> 
>> Are you sure lhs_type is never qualified?
>
>For the V2DF and V4SF types, I am mostly sure, but since you bring it
>up
>I will admit I am not positive.  :-)       
>
>>                                     That is, for a GIMPLE folder
>> I'd have expected
>>
>>   if (types_compatible_p (TREE_TYPE (lhs_type), TREE_TYPE
>(V2DF_type_node)))
>
>> for GENERIC
>> 
>>   if (TYPE_MAIN_VARIANT (TREE_TYPE (lhs_type)) == TREE_TYPE
>(V2DF_type_node))
>
>Either/both of those seem more robust than what I had come up with,
>I'll
>plan to make an update here.
>
>Any guidance on whether I should prefer "types_compatible_p" versus the
>GENERIC "TYPE_MAIN_VARIANT" ?   I see more references to the latter in
>rs6000.c , but the former seems to make better sense to me just by its
>name.

The former if this is called from the GIMPLE stmts folding hook. 

Richard. 

>Thanks for the review, 
>-Will
>
>
>> Richard.
>> 
>> >         permute_type = V2DI_type_node;
>> >       else if (TREE_TYPE (lhs_type) == TREE_TYPE (V4SF_type_node))
>> >         permute_type = V4SI_type_node;
>> >       else
>> >         gcc_unreachable ();
>> >     }
>> >
>> > Okay for trunk with that changed.  Thanks!
>> >
>> >
>> > Segher
>>
diff mbox series

Patch

diff --git a/gcc/config/rs6000/rs6000.c b/gcc/config/rs6000/rs6000.c
index 6cb5c87..35c32be 100644
--- a/gcc/config/rs6000/rs6000.c
+++ b/gcc/config/rs6000/rs6000.c
@@ -15119,24 +15119,39 @@  fold_mergehl_helper (gimple_stmt_iterator *gsi, gimple *stmt, int use_high)
 {
   tree arg0 = gimple_call_arg (stmt, 0);
   tree arg1 = gimple_call_arg (stmt, 1);
   tree lhs = gimple_call_lhs (stmt);
   tree lhs_type = TREE_TYPE (lhs);
-  tree lhs_type_type = TREE_TYPE (lhs_type);
   int n_elts = TYPE_VECTOR_SUBPARTS (lhs_type);
   int midpoint = n_elts / 2;
   int offset = 0;
 
   if (use_high == 1)
     offset = midpoint;
 
-  tree_vector_builder elts (lhs_type, VECTOR_CST_NELTS (arg0), 1);
+  /* The permute_type will match the lhs for integral types.  For double and
+     float types, the permute type needs to map to the V2 or V4 type that
+     matches size.  */
+  tree permute_type;
+  if (INTEGRAL_TYPE_P (TREE_TYPE (lhs_type)))
+    permute_type = lhs_type;
+  else
+    if (TREE_TYPE (lhs_type) == TREE_TYPE (V2DF_type_node))
+      permute_type = V2DI_type_node;
+    else if (TREE_TYPE (lhs_type) == TREE_TYPE (V4SF_type_node))
+      permute_type = V4SI_type_node;
+    else
+      gcc_unreachable ();
+
+  tree_vector_builder elts (permute_type, VECTOR_CST_NELTS (arg0), 1);
 
   for (int i = 0; i < midpoint; i++)
     {
-      elts.safe_push (build_int_cst (lhs_type_type, offset + i));
-      elts.safe_push (build_int_cst (lhs_type_type, offset + n_elts + i));
+      elts.safe_push (build_int_cst (TREE_TYPE (permute_type),
+				     offset + i));
+      elts.safe_push (build_int_cst (TREE_TYPE (permute_type),
+				     offset + n_elts + i));
     }
 
   tree permute = elts.build ();
 
   gimple *g = gimple_build_assign (lhs, VEC_PERM_EXPR, arg0, arg1, permute);
@@ -15757,18 +15772,22 @@  rs6000_gimple_fold_builtin (gimple_stmt_iterator *gsi)
     case ALTIVEC_BUILTIN_VMRGLH:
     case ALTIVEC_BUILTIN_VMRGLW:
     case VSX_BUILTIN_XXMRGLW_4SI:
     case ALTIVEC_BUILTIN_VMRGLB:
     case VSX_BUILTIN_VEC_MERGEL_V2DI:
+    case VSX_BUILTIN_XXMRGLW_4SF:
+    case VSX_BUILTIN_VEC_MERGEL_V2DF:
 	fold_mergehl_helper (gsi, stmt, 1);
 	return true;
     /* vec_mergeh (integrals).  */
     case ALTIVEC_BUILTIN_VMRGHH:
     case ALTIVEC_BUILTIN_VMRGHW:
     case VSX_BUILTIN_XXMRGHW_4SI:
     case ALTIVEC_BUILTIN_VMRGHB:
     case VSX_BUILTIN_VEC_MERGEH_V2DI:
+    case VSX_BUILTIN_XXMRGHW_4SF:
+    case VSX_BUILTIN_VEC_MERGEH_V2DF:
 	fold_mergehl_helper (gsi, stmt, 0);
 	return true;
 
     /* d = vec_pack (a, b) */
     case P8V_BUILTIN_VPKUDUM: