public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
From: Richard Biener <rguenther@suse.de>
To: Jan Hubicka <hubicka@ucw.cz>
Cc: gcc-patches@gcc.gnu.org
Subject: Re: Add EAF_NOT_RETURNED flag
Date: Mon, 16 Nov 2020 10:14:58 +0100 (CET)	[thread overview]
Message-ID: <nycvar.YFH.7.76.2011161014230.10073@p653.nepu.fhfr.qr> (raw)
In-Reply-To: <20201115200049.GE1667@kam.mff.cuni.cz>

On Sun, 15 Nov 2020, Jan Hubicka wrote:

> Hi,
> one of common cases why we lose track on SSA name in modref propagation
> is the situation where name is passed to funtion call and we then thus
> continue by propagating through return value.
> 
> This patch adds simple logic to track arguments that does not escape to
> return value.  On cc1plus the improvement in PTA disambiguation is about
> 1.8% that is not too great, but it is easy enough to discover and often
> simplifies the propagation.  I suppose this can be also useful for
> tree-ssa-structalias itself since it may also prevent args from escaping
> to return values.
> 
> Bootstrapped/regtested x86_64-linux, OK?

OK.  Can you amend the EAF_NOESCAPE comment to say "besides through
the return value"?

Thanks,
Richard.

> Honza
> 	* ipa-modref.c (dump_eaf_flags): Add EAF_NOT_RETURNED.
> 	(modref_lattice::init): Add EAF_NOT_RETURNED.
> 	(merge_call_lhs_flags): Do nothing for EAF_NOT_RETURNED
> 	args.
> 	(analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
> 	add EAF_NOT_RETURNED to call flags.
> 	(modref_merge_call_site_flags): Add EAF_NOT_RETURNED.
> 	* tree-core.h (EAF_NOT_RETURNED): New flags.
> diff --git a/gcc/ipa-modref.c b/gcc/ipa-modref.c
> index 7f673675d1d..155bf59fc53 100644
> --- a/gcc/ipa-modref.c
> +++ b/gcc/ipa-modref.c
> @@ -153,6 +153,8 @@ dump_eaf_flags (FILE *out, int flags, bool newline = true)
>      fprintf (out, " noescape");
>    if (flags & EAF_UNUSED)
>      fprintf (out, " unused");
> +  if (flags & EAF_NOT_RETURNED)
> +    fprintf (out, " not_returned");
>    if (newline)
>    fprintf (out, "\n");
>  }
> @@ -1361,7 +1363,8 @@ public:
>  void
>  modref_lattice::init ()
>  {
> -  flags = EAF_DIRECT | EAF_NOCLOBBER | EAF_NOESCAPE | EAF_UNUSED;
> +  flags = EAF_DIRECT | EAF_NOCLOBBER | EAF_NOESCAPE | EAF_UNUSED
> +	  | EAF_NOT_RETURNED;
>    open = true;
>    known = false;
>  }
> @@ -1531,6 +1534,9 @@ merge_call_lhs_flags (gcall *call, int arg, int index, bool deref,
>        && (flags & ERF_RETURN_ARG_MASK) != arg)
>      return;
>  
> +  if (gimple_call_arg_flags (call, arg) & EAF_NOT_RETURNED)
> +    return;
> +
>    /* If return value is SSA name determine its flags.  */
>    if (TREE_CODE (gimple_call_lhs (call)) == SSA_NAME)
>      {
> @@ -1601,11 +1607,17 @@ analyze_ssa_name_flags (tree name, vec<modref_lattice> &lattice, int depth, bool
>  	  print_gimple_stmt (dump_file, use_stmt, 0);
>  	}
>  
> -      /* Gimple return may load the return value.  */
> +      /* Gimple return may load the return value.
> +	 and clears EAF_NOT_RETURNED flag.  */
>        if (greturn *ret = dyn_cast <greturn *> (use_stmt))
>  	{
>  	  if (memory_access_to (gimple_return_retval (ret), name))
> -	    lattice [index].merge_direct_load ();
> +	    {
> +	      lattice[index].merge_direct_load ();
> +	      lattice[index].merge (~EAF_NOT_RETURNED);
> +	    }
> +	  else if (gimple_return_retval (ret) == name)
> +	    lattice[index].merge (~EAF_NOT_RETURNED);
>  	}
>        /* Account for LHS store, arg loads and flags from callee function.  */
>        else if (gcall *call = dyn_cast <gcall *> (use_stmt))
> @@ -1645,7 +1657,8 @@ analyze_ssa_name_flags (tree name, vec<modref_lattice> &lattice, int depth, bool
>  		  {
>  		    if (!(ecf_flags & (ECF_CONST | ECF_NOVOPS)))
>  		      {
> -			int call_flags = gimple_call_arg_flags (call, i);
> +			int call_flags = gimple_call_arg_flags (call, i)
> +					 | EAF_NOT_RETURNED;
>  			if (ignore_stores)
>  			  call_flags |= EAF_NOCLOBBER | EAF_NOESCAPE;
>  
> @@ -1666,8 +1679,9 @@ analyze_ssa_name_flags (tree name, vec<modref_lattice> &lattice, int depth, bool
>  		      lattice[index].merge_direct_load ();
>  		    else
>  		      {
> -			int call_flags = deref_flags (gimple_call_arg_flags (call, i),
> -			    			      ignore_stores);
> +			int call_flags = deref_flags
> +				(gimple_call_arg_flags (call, i),
> +				 ignore_stores) | EAF_NOT_RETURNED;
>  			if (!record_ipa)
>  			    lattice[index].merge (call_flags);
>  			if (record_ipa)
> @@ -3612,8 +3626,9 @@ modref_merge_call_site_flags (escape_summary *sum,
>  	  flags |= EAF_NOESCAPE | EAF_NOCLOBBER;
>  	  flags_lto |= EAF_NOESCAPE | EAF_NOCLOBBER;
>  	}
> -      flags |= ee->min_flags;
> -      flags_lto |= ee->min_flags;
> +      /* Returning the value is already accounted to at local propagation.  */
> +      flags |= ee->min_flags | EAF_NOT_RETURNED;
> +      flags_lto |= ee->min_flags | EAF_NOT_RETURNED;
>        if (cur_summary && ee->parm_index < cur_summary->arg_flags.length ())
>  	{
>  	  int f = cur_summary->arg_flags[ee->parm_index];
> diff --git a/gcc/tree-core.h b/gcc/tree-core.h
> index c9280a8d3b1..3e9455a553b 100644
> --- a/gcc/tree-core.h
> +++ b/gcc/tree-core.h
> @@ -110,6 +110,9 @@ struct die_struct;
>  /* Nonzero if the argument is not used by the function.  */
>  #define EAF_UNUSED		(1 << 3)
>  
> +/* Nonzero if the argument does not escape to return value.  */
> +#define EAF_NOT_RETURNED	(1 << 4)
> +
>  /* Call return flags.  */
>  /* Mask for the argument number that is returned.  Lower two bits of
>     the return flags, encodes argument slots zero to three.  */
> 

-- 
Richard Biener <rguenther@suse.de>
SUSE Software Solutions Germany GmbH, Maxfeldstrasse 5, 90409 Nuernberg,
Germany; GF: Felix Imend

  reply	other threads:[~2020-11-16  9:14 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-11-15 20:00 Jan Hubicka
2020-11-16  9:14 ` Richard Biener [this message]
2021-07-16  8:41 Jan Hubicka
2021-07-16 10:38 ` Richard Biener

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=nycvar.YFH.7.76.2011161014230.10073@p653.nepu.fhfr.qr \
    --to=rguenther@suse.de \
    --cc=gcc-patches@gcc.gnu.org \
    --cc=hubicka@ucw.cz \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).