[V2] SCCVN: Add LEN_MASK_STORE and fix LEN_STORE

Message ID 20230626093846.3006718-1-juzhe.zhong@rivai.ai
State Unresolved
Headers
Series [V2] SCCVN: Add LEN_MASK_STORE and fix LEN_STORE |

Checks

Context Check Description
snail/gcc-patch-check warning Git am fail log

Commit Message

juzhe.zhong@rivai.ai June 26, 2023, 9:38 a.m. UTC
  From: Ju-Zhe Zhong <juzhe.zhong@rivai.ai>

Hi, Richi. It seems that we use nunits which is len + bias to iterate then we can
simplify the codes.

Also, I fixed behavior of len_store,

Before this patch:
   (len - bias) * BITS_PER_UNIT
After this patch:
   (len + bias) * BITS_PER_UNIT
   
gcc/ChangeLog:

        * tree-ssa-sccvn.cc (vn_reference_lookup_3): Add LEN_MAS_STORE and fix LEN_STORE.

---
 gcc/tree-ssa-sccvn.cc | 19 +++++++++++++++++--
 1 file changed, 17 insertions(+), 2 deletions(-)
  

Comments

Richard Biener June 26, 2023, 11:18 a.m. UTC | #1
On Mon, 26 Jun 2023, juzhe.zhong@rivai.ai wrote:

> From: Ju-Zhe Zhong <juzhe.zhong@rivai.ai>
> 
> Hi, Richi. It seems that we use nunits which is len + bias to iterate then we can
> simplify the codes.
> 
> Also, I fixed behavior of len_store,
> 
> Before this patch:
>    (len - bias) * BITS_PER_UNIT
> After this patch:
>    (len + bias) * BITS_PER_UNIT
>    
> gcc/ChangeLog:
> 
>         * tree-ssa-sccvn.cc (vn_reference_lookup_3): Add LEN_MAS_STORE and fix LEN_STORE.
> 
> ---
>  gcc/tree-ssa-sccvn.cc | 19 +++++++++++++++++--
>  1 file changed, 17 insertions(+), 2 deletions(-)
> 
> diff --git a/gcc/tree-ssa-sccvn.cc b/gcc/tree-ssa-sccvn.cc
> index 11061a374a2..228ec117ff3 100644
> --- a/gcc/tree-ssa-sccvn.cc
> +++ b/gcc/tree-ssa-sccvn.cc
> @@ -3304,6 +3304,16 @@ vn_reference_lookup_3 (ao_ref *ref, tree vuse, void *data_,
>  	  if (!tree_fits_uhwi_p (len) || !tree_fits_shwi_p (bias))
>  	    return (void *)-1;
>  	  break;
> +	case IFN_LEN_MASK_STORE:
> +	  len = gimple_call_arg (call, 2);
> +	  bias = gimple_call_arg (call, 5);
> +	  if (!tree_fits_uhwi_p (len) || !tree_fits_shwi_p (bias))
> +	    return (void *)-1;
> +	  mask = gimple_call_arg (call, internal_fn_mask_index (fn));
> +	  mask = vn_valueize (mask);
> +	  if (TREE_CODE (mask) != VECTOR_CST)
> +	    return (void *)-1;
> +	  break;
>  	default:
>  	  return (void *)-1;
>  	}
> @@ -3344,6 +3354,11 @@ vn_reference_lookup_3 (ao_ref *ref, tree vuse, void *data_,
>  	      tree vectype = TREE_TYPE (def_rhs);
>  	      unsigned HOST_WIDE_INT elsz
>  		= tree_to_uhwi (TYPE_SIZE (TREE_TYPE (vectype)));
> +	      poly_uint64 nunits;
> +	      if (len)
> +		nunits = tree_to_uhwi (len) + tree_to_shwi (bias);
> +	      else
> +		nunits = TYPE_VECTOR_SUBPARTS (vectype);

Are the _LEN ifns accessible via intrinsics as well?  If so I think
we should use MIN (nunits, len + bias) here as otherwise we risk
out-of bound accesses.

Otherwise looks good to me.

Thanks,
Richard.

>  	      if (mask)
>  		{
>  		  HOST_WIDE_INT start = 0, length = 0;
> @@ -3373,7 +3388,7 @@ vn_reference_lookup_3 (ao_ref *ref, tree vuse, void *data_,
>  			length += elsz;
>  		      mask_idx++;
>  		    }
> -		  while (known_lt (mask_idx, TYPE_VECTOR_SUBPARTS (vectype)));
> +		  while (known_lt (mask_idx, nunits));
>  		  if (length != 0)
>  		    {
>  		      pd.rhs_off = start;
> @@ -3389,7 +3404,7 @@ vn_reference_lookup_3 (ao_ref *ref, tree vuse, void *data_,
>  		{
>  		  pd.offset = offset2i;
>  		  pd.size = (tree_to_uhwi (len)
> -			     + -tree_to_shwi (bias)) * BITS_PER_UNIT;
> +			     + tree_to_shwi (bias)) * BITS_PER_UNIT;
>  		  if (BYTES_BIG_ENDIAN)
>  		    pd.rhs_off = pd.size - tree_to_uhwi (TYPE_SIZE (vectype));
>  		  else
>
  
juzhe.zhong@rivai.ai June 26, 2023, 11:47 a.m. UTC | #2
Hi, Richi. I am wondering whether it is true that :?

TYPE_VECTOR_SUBPARTS (vectype).to_constant ()

Thanks.


juzhe.zhong@rivai.ai
 
From: Richard Biener
Date: 2023-06-26 19:18
To: Ju-Zhe Zhong
CC: gcc-patches; richard.sandiford
Subject: Re: [PATCH V2] SCCVN: Add LEN_MASK_STORE and fix LEN_STORE
On Mon, 26 Jun 2023, juzhe.zhong@rivai.ai wrote:
 
> From: Ju-Zhe Zhong <juzhe.zhong@rivai.ai>
> 
> Hi, Richi. It seems that we use nunits which is len + bias to iterate then we can
> simplify the codes.
> 
> Also, I fixed behavior of len_store,
> 
> Before this patch:
>    (len - bias) * BITS_PER_UNIT
> After this patch:
>    (len + bias) * BITS_PER_UNIT
>    
> gcc/ChangeLog:
> 
>         * tree-ssa-sccvn.cc (vn_reference_lookup_3): Add LEN_MAS_STORE and fix LEN_STORE.
> 
> ---
>  gcc/tree-ssa-sccvn.cc | 19 +++++++++++++++++--
>  1 file changed, 17 insertions(+), 2 deletions(-)
> 
> diff --git a/gcc/tree-ssa-sccvn.cc b/gcc/tree-ssa-sccvn.cc
> index 11061a374a2..228ec117ff3 100644
> --- a/gcc/tree-ssa-sccvn.cc
> +++ b/gcc/tree-ssa-sccvn.cc
> @@ -3304,6 +3304,16 @@ vn_reference_lookup_3 (ao_ref *ref, tree vuse, void *data_,
>    if (!tree_fits_uhwi_p (len) || !tree_fits_shwi_p (bias))
>      return (void *)-1;
>    break;
> + case IFN_LEN_MASK_STORE:
> +   len = gimple_call_arg (call, 2);
> +   bias = gimple_call_arg (call, 5);
> +   if (!tree_fits_uhwi_p (len) || !tree_fits_shwi_p (bias))
> +     return (void *)-1;
> +   mask = gimple_call_arg (call, internal_fn_mask_index (fn));
> +   mask = vn_valueize (mask);
> +   if (TREE_CODE (mask) != VECTOR_CST)
> +     return (void *)-1;
> +   break;
>  default:
>    return (void *)-1;
>  }
> @@ -3344,6 +3354,11 @@ vn_reference_lookup_3 (ao_ref *ref, tree vuse, void *data_,
>        tree vectype = TREE_TYPE (def_rhs);
>        unsigned HOST_WIDE_INT elsz
>  = tree_to_uhwi (TYPE_SIZE (TREE_TYPE (vectype)));
> +       poly_uint64 nunits;
> +       if (len)
> + nunits = tree_to_uhwi (len) + tree_to_shwi (bias);
> +       else
> + nunits = TYPE_VECTOR_SUBPARTS (vectype);
 
Are the _LEN ifns accessible via intrinsics as well?  If so I think
we should use MIN (nunits, len + bias) here as otherwise we risk
out-of bound accesses.
 
Otherwise looks good to me.
 
Thanks,
Richard.
 
>        if (mask)
>  {
>    HOST_WIDE_INT start = 0, length = 0;
> @@ -3373,7 +3388,7 @@ vn_reference_lookup_3 (ao_ref *ref, tree vuse, void *data_,
>  length += elsz;
>        mask_idx++;
>      }
> -   while (known_lt (mask_idx, TYPE_VECTOR_SUBPARTS (vectype)));
> +   while (known_lt (mask_idx, nunits));
>    if (length != 0)
>      {
>        pd.rhs_off = start;
> @@ -3389,7 +3404,7 @@ vn_reference_lookup_3 (ao_ref *ref, tree vuse, void *data_,
>  {
>    pd.offset = offset2i;
>    pd.size = (tree_to_uhwi (len)
> -      + -tree_to_shwi (bias)) * BITS_PER_UNIT;
> +      + tree_to_shwi (bias)) * BITS_PER_UNIT;
>    if (BYTES_BIG_ENDIAN)
>      pd.rhs_off = pd.size - tree_to_uhwi (TYPE_SIZE (vectype));
>    else
> 
 
-- 
Richard Biener <rguenther@suse.de>
SUSE Software Solutions Germany GmbH, Frankenstrasse 146, 90461 Nuernberg,
Germany; GF: Ivo Totev, Andrew Myers, Andrew McDonald, Boudien Moerman;
HRB 36809 (AG Nuernberg)
  
Richard Biener June 26, 2023, 11:48 a.m. UTC | #3
On Mon, 26 Jun 2023, juzhe.zhong@rivai.ai wrote:

> Hi, Richi. I am wondering whether it is true that :?
> 
> TYPE_VECTOR_SUBPARTS (vectype).to_constant ()

Not necessarily.

> Thanks.
> 
> 
> juzhe.zhong@rivai.ai
>  
> From: Richard Biener
> Date: 2023-06-26 19:18
> To: Ju-Zhe Zhong
> CC: gcc-patches; richard.sandiford
> Subject: Re: [PATCH V2] SCCVN: Add LEN_MASK_STORE and fix LEN_STORE
> On Mon, 26 Jun 2023, juzhe.zhong@rivai.ai wrote:
>  
> > From: Ju-Zhe Zhong <juzhe.zhong@rivai.ai>
> > 
> > Hi, Richi. It seems that we use nunits which is len + bias to iterate then we can
> > simplify the codes.
> > 
> > Also, I fixed behavior of len_store,
> > 
> > Before this patch:
> >    (len - bias) * BITS_PER_UNIT
> > After this patch:
> >    (len + bias) * BITS_PER_UNIT
> >    
> > gcc/ChangeLog:
> > 
> >         * tree-ssa-sccvn.cc (vn_reference_lookup_3): Add LEN_MAS_STORE and fix LEN_STORE.
> > 
> > ---
> >  gcc/tree-ssa-sccvn.cc | 19 +++++++++++++++++--
> >  1 file changed, 17 insertions(+), 2 deletions(-)
> > 
> > diff --git a/gcc/tree-ssa-sccvn.cc b/gcc/tree-ssa-sccvn.cc
> > index 11061a374a2..228ec117ff3 100644
> > --- a/gcc/tree-ssa-sccvn.cc
> > +++ b/gcc/tree-ssa-sccvn.cc
> > @@ -3304,6 +3304,16 @@ vn_reference_lookup_3 (ao_ref *ref, tree vuse, void *data_,
> >    if (!tree_fits_uhwi_p (len) || !tree_fits_shwi_p (bias))
> >      return (void *)-1;
> >    break;
> > + case IFN_LEN_MASK_STORE:
> > +   len = gimple_call_arg (call, 2);
> > +   bias = gimple_call_arg (call, 5);
> > +   if (!tree_fits_uhwi_p (len) || !tree_fits_shwi_p (bias))
> > +     return (void *)-1;
> > +   mask = gimple_call_arg (call, internal_fn_mask_index (fn));
> > +   mask = vn_valueize (mask);
> > +   if (TREE_CODE (mask) != VECTOR_CST)
> > +     return (void *)-1;
> > +   break;
> >  default:
> >    return (void *)-1;
> >  }
> > @@ -3344,6 +3354,11 @@ vn_reference_lookup_3 (ao_ref *ref, tree vuse, void *data_,
> >        tree vectype = TREE_TYPE (def_rhs);
> >        unsigned HOST_WIDE_INT elsz
> >  = tree_to_uhwi (TYPE_SIZE (TREE_TYPE (vectype)));
> > +       poly_uint64 nunits;
> > +       if (len)
> > + nunits = tree_to_uhwi (len) + tree_to_shwi (bias);
> > +       else
> > + nunits = TYPE_VECTOR_SUBPARTS (vectype);
>  
> Are the _LEN ifns accessible via intrinsics as well?  If so I think
> we should use MIN (nunits, len + bias) here as otherwise we risk
> out-of bound accesses.
>  
> Otherwise looks good to me.
>  
> Thanks,
> Richard.
>  
> >        if (mask)
> >  {
> >    HOST_WIDE_INT start = 0, length = 0;
> > @@ -3373,7 +3388,7 @@ vn_reference_lookup_3 (ao_ref *ref, tree vuse, void *data_,
> >  length += elsz;
> >        mask_idx++;
> >      }
> > -   while (known_lt (mask_idx, TYPE_VECTOR_SUBPARTS (vectype)));
> > +   while (known_lt (mask_idx, nunits));
> >    if (length != 0)
> >      {
> >        pd.rhs_off = start;
> > @@ -3389,7 +3404,7 @@ vn_reference_lookup_3 (ao_ref *ref, tree vuse, void *data_,
> >  {
> >    pd.offset = offset2i;
> >    pd.size = (tree_to_uhwi (len)
> > -      + -tree_to_shwi (bias)) * BITS_PER_UNIT;
> > +      + tree_to_shwi (bias)) * BITS_PER_UNIT;
> >    if (BYTES_BIG_ENDIAN)
> >      pd.rhs_off = pd.size - tree_to_uhwi (TYPE_SIZE (vectype));
> >    else
> > 
>  
>
  

Patch

diff --git a/gcc/tree-ssa-sccvn.cc b/gcc/tree-ssa-sccvn.cc
index 11061a374a2..228ec117ff3 100644
--- a/gcc/tree-ssa-sccvn.cc
+++ b/gcc/tree-ssa-sccvn.cc
@@ -3304,6 +3304,16 @@  vn_reference_lookup_3 (ao_ref *ref, tree vuse, void *data_,
 	  if (!tree_fits_uhwi_p (len) || !tree_fits_shwi_p (bias))
 	    return (void *)-1;
 	  break;
+	case IFN_LEN_MASK_STORE:
+	  len = gimple_call_arg (call, 2);
+	  bias = gimple_call_arg (call, 5);
+	  if (!tree_fits_uhwi_p (len) || !tree_fits_shwi_p (bias))
+	    return (void *)-1;
+	  mask = gimple_call_arg (call, internal_fn_mask_index (fn));
+	  mask = vn_valueize (mask);
+	  if (TREE_CODE (mask) != VECTOR_CST)
+	    return (void *)-1;
+	  break;
 	default:
 	  return (void *)-1;
 	}
@@ -3344,6 +3354,11 @@  vn_reference_lookup_3 (ao_ref *ref, tree vuse, void *data_,
 	      tree vectype = TREE_TYPE (def_rhs);
 	      unsigned HOST_WIDE_INT elsz
 		= tree_to_uhwi (TYPE_SIZE (TREE_TYPE (vectype)));
+	      poly_uint64 nunits;
+	      if (len)
+		nunits = tree_to_uhwi (len) + tree_to_shwi (bias);
+	      else
+		nunits = TYPE_VECTOR_SUBPARTS (vectype);
 	      if (mask)
 		{
 		  HOST_WIDE_INT start = 0, length = 0;
@@ -3373,7 +3388,7 @@  vn_reference_lookup_3 (ao_ref *ref, tree vuse, void *data_,
 			length += elsz;
 		      mask_idx++;
 		    }
-		  while (known_lt (mask_idx, TYPE_VECTOR_SUBPARTS (vectype)));
+		  while (known_lt (mask_idx, nunits));
 		  if (length != 0)
 		    {
 		      pd.rhs_off = start;
@@ -3389,7 +3404,7 @@  vn_reference_lookup_3 (ao_ref *ref, tree vuse, void *data_,
 		{
 		  pd.offset = offset2i;
 		  pd.size = (tree_to_uhwi (len)
-			     + -tree_to_shwi (bias)) * BITS_PER_UNIT;
+			     + tree_to_shwi (bias)) * BITS_PER_UNIT;
 		  if (BYTES_BIG_ENDIAN)
 		    pd.rhs_off = pd.size - tree_to_uhwi (TYPE_SIZE (vectype));
 		  else