Re: Fix up -fexcess-precision handling in LTO (was Re: [GCC][middle-end] Add rules to strip away unneeded type casts in expressions (2nd patch))

[Richard Biener <rguenther at suse dot de>]

On Tue, 30 Jul 2019, Jakub Jelinek wrote:

> On Tue, Jul 02, 2019 at 04:43:54PM +0000, Tamar Christina wrote:
> > Here's an updated patch with the changes processed from the previous review.
> > 
> > I've bootstrapped and regtested on aarch64-none-linux-gnu and x86_64-pc-linux-gnu and no issues.
> 
> These changes also broke gcc.dg/torture/c99-contract-1.c with -flto
> on i686-linux.
> 
> The problem is that after moving the folding from convert.c to match.pd,
> it is now performed not only during FE folding, but also much later on,
> including post-IPA optimizations in lto1.  The C FE arranges
> flag_excess_precision_cmdline and flag_excess_precision to be
> EXCESS_PRECISION_STANDARD and thus on i686-linux floating point arithmetics
> is performed in long double, but the lto1 FE has both set to
> EXCESS_PRECISION_FAST and undoes that widening.
> 
> There seems to be quite complicated distinction between
> flag_excess_precision_cmdline and flag_excess_precision, but it seems
> that these days it is unnecessary, flag_excess_precision is only ever set
> from flag_excess_precision_cmdline, perhaps in the past targets used to
> modify flag_excess_precision, but they don't do that anymore.
> 
> Furthermore, some comments claimed that the proper EXCESS_PRECISION_STANDARD
> handling requires FE support, but that also doesn't seem to be the case
> these days, some FEs even just use EXCESS_PRECISION_STANDARD by default
> (go, D).
> 
> So, the following patch gets rid of flag_excess_precision and renames
> flag_excess_precision_cmdline to flag_excess_precision, plus adds
> Optimization flag to that command line option, so that we remember it during
> compilation and e.g. during LTO can then have some functions with standard
> excess precision and others with fast excess precision.
> 
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?

Looks OK to me but I'd like Joseph to chime in here.  I think for
FE support it means adding actual _widening_ casts to the effective
compute precision type (thus long double).  But once that's done
the rest should be indeed FE independent (but I doubt D and go get
it correct).

Thanks,
Richard.

> 2019-07-30  Jakub Jelinek  <jakub@redhat.com>
> 
> 	PR middle-end/91283
> 	* common.opt (fexcess-precision=): Add Optimization flag.  Use
> 	flag_excess_precision variable instead of
> 	flag_excess_precision_cmdline.
> 	* flags.h (class target_flag_state): Remove x_flag_excess_precision
> 	member.
> 	(flag_excess_precision): Don't define.
> 	* langhooks.c (lhd_post_options): Set flag_excess_precision instead of
> 	flag_excess_precision_cmdline.  Remove comment.
> 	* opts.c (set_fast_math_flags): Use frontend_set_flag_excess_precision
> 	and x_flag_excess_precision instead of
> 	frontend_set_flag_excess_precision_cmdline and
> 	x_flag_excess_precision_cmdline.
> 	(fast_math_flags_set_p): Use x_flag_excess_precision instead of
> 	x_flag_excess_precision_cmdline.
> 	* toplev.c (init_excess_precision): Remove.
> 	(lang_dependent_init_target): Don't call it.
> ada/
> 	* gcc-interface/misc.c (gnat_post_options): Set flag_excess_precision
> 	instead of flag_excess_precision_cmdline.
> brig/
> 	* brig-lang.c (brig_langhook_post_options): Set flag_excess_precision
> 	instead of flag_excess_precision_cmdline.
> c-family/
> 	* c-common.c (c_ts18661_flt_eval_method): Use flag_excess_precision
> 	instead of flag_excess_precision_cmdline.
> 	* c-cppbuiltin.c (c_cpp_flt_eval_method_iec_559): Likewise.
> 	* c-opts.c (c_common_post_options): Likewise.
> d/
> 	* d-lang.cc (d_post_options): Set flag_excess_precision instead of
> 	flag_excess_precision_cmdline.
> fortran/
> 	* options.c (gfc_post_options): Set flag_excess_precision instead of
> 	flag_excess_precision_cmdline.  Remove comment.
> go/
> 	* go-lang.c (go_langhook_post_options): Set flag_excess_precision
> 	instead of flag_excess_precision_cmdline.
> lto/
> 	* lto-lang.c (lto_post_options): Set flag_excess_precision instead of
> 	flag_excess_precision_cmdline.  Remove comment.
> 
> --- gcc/common.opt.jj	2019-07-29 12:56:38.968248060 +0200
> +++ gcc/common.opt	2019-07-29 13:01:24.067067583 +0200
> @@ -1399,7 +1399,7 @@ Common Report Var(flag_expensive_optimiz
>  Perform a number of minor, expensive optimizations.
>  
>  fexcess-precision=
> -Common Joined RejectNegative Enum(excess_precision) Var(flag_excess_precision_cmdline) Init(EXCESS_PRECISION_DEFAULT) SetByCombined
> +Common Joined RejectNegative Enum(excess_precision) Var(flag_excess_precision) Init(EXCESS_PRECISION_DEFAULT) Optimization SetByCombined
>  -fexcess-precision=[fast|standard]	Specify handling of excess floating-point precision.
>  
>  Enum
> --- gcc/flags.h.jj	2019-07-10 15:52:20.362155642 +0200
> +++ gcc/flags.h	2019-07-29 13:02:05.488460207 +0200
> @@ -51,9 +51,6 @@ public:
>    align_flags x_align_jumps;
>    align_flags x_align_labels;
>    align_flags x_align_functions;
> -
> -  /* The excess precision currently in effect.  */
> -  enum excess_precision x_flag_excess_precision;
>  };
>  
>  extern class target_flag_state default_target_flag_state;
> @@ -68,12 +65,6 @@ extern class target_flag_state *this_tar
>  #define align_labels	 (this_target_flag_state->x_align_labels)
>  #define align_functions	 (this_target_flag_state->x_align_functions)
>  
> -/* String representaions of the above options are available in
> -   const char *str_align_foo.  NULL if not set.  */
> -
> -#define flag_excess_precision \
> -  (this_target_flag_state->x_flag_excess_precision)
> -
>  /* Returns TRUE if generated code should match ABI version N or
>     greater is in use.  */
>  
> --- gcc/langhooks.c.jj	2019-01-01 12:37:19.531936001 +0100
> +++ gcc/langhooks.c	2019-07-29 13:10:35.053988315 +0200
> @@ -95,9 +95,7 @@ lhd_return_null_const_tree (const_tree A
>  bool
>  lhd_post_options (const char ** ARG_UNUSED (pfilename))
>  {
> -  /* Excess precision other than "fast" requires front-end
> -     support.  */
> -  flag_excess_precision_cmdline = EXCESS_PRECISION_FAST;
> +  flag_excess_precision = EXCESS_PRECISION_FAST;
>    return false;
>  }
>  
> --- gcc/opts.c.jj	2019-07-29 12:56:38.000000000 +0200
> +++ gcc/opts.c	2019-07-29 13:02:40.522946490 +0200
> @@ -2962,9 +2962,8 @@ set_fast_math_flags (struct gcc_options
>      opts->x_flag_errno_math = !set;
>    if (set)
>      {
> -      if (opts->frontend_set_flag_excess_precision_cmdline
> -	  == EXCESS_PRECISION_DEFAULT)
> -	opts->x_flag_excess_precision_cmdline
> +      if (opts->frontend_set_flag_excess_precision == EXCESS_PRECISION_DEFAULT)
> +	opts->x_flag_excess_precision
>  	  = set ? EXCESS_PRECISION_FAST : EXCESS_PRECISION_DEFAULT;
>        if (!opts->frontend_set_flag_signaling_nans)
>  	opts->x_flag_signaling_nans = 0;
> @@ -2999,8 +2998,7 @@ fast_math_flags_set_p (const struct gcc_
>  	  && opts->x_flag_finite_math_only
>  	  && !opts->x_flag_signed_zeros
>  	  && !opts->x_flag_errno_math
> -	  && opts->x_flag_excess_precision_cmdline
> -	     == EXCESS_PRECISION_FAST);
> +	  && opts->x_flag_excess_precision == EXCESS_PRECISION_FAST);
>  }
>  
>  /* Return true iff flags are set as if -ffast-math but using the flags stored
> --- gcc/toplev.c.jj	2019-07-28 17:29:31.246291937 +0200
> +++ gcc/toplev.c	2019-07-29 13:08:49.601534597 +0200
> @@ -1849,27 +1849,11 @@ backend_init (void)
>    init_regs ();
>  }
>  
> -/* Initialize excess precision settings.
> -
> -   We have no need to modify anything here, just keep track of what the
> -   user requested.  We'll figure out any appropriate relaxations
> -   later.  */
> -
> -static void
> -init_excess_precision (void)
> -{
> -  gcc_assert (flag_excess_precision_cmdline != EXCESS_PRECISION_DEFAULT);
> -  flag_excess_precision = flag_excess_precision_cmdline;
> -}
> -
>  /* Initialize things that are both lang-dependent and target-dependent.
>     This function can be called more than once if target parameters change.  */
>  static void
>  lang_dependent_init_target (void)
>  {
> -  /* This determines excess precision settings.  */
> -  init_excess_precision ();
> -
>    /* This creates various _DECL nodes, so needs to be called after the
>       front end is initialized.  It also depends on the HAVE_xxx macros
>       generated from the target machine description.  */
> --- gcc/c-family/c-common.c.jj	2019-07-04 00:18:32.052090626 +0200
> +++ gcc/c-family/c-common.c	2019-07-29 13:12:05.762659524 +0200
> @@ -8342,7 +8342,7 @@ c_ts18661_flt_eval_method (void)
>      = targetm.c.excess_precision (EXCESS_PRECISION_TYPE_IMPLICIT);
>  
>    enum excess_precision_type flag_type
> -    = (flag_excess_precision_cmdline == EXCESS_PRECISION_STANDARD
> +    = (flag_excess_precision == EXCESS_PRECISION_STANDARD
>         ? EXCESS_PRECISION_TYPE_STANDARD
>         : EXCESS_PRECISION_TYPE_FAST);
>  
> --- gcc/c-family/c-cppbuiltin.c.jj	2019-01-16 09:35:04.563323106 +0100
> +++ gcc/c-family/c-cppbuiltin.c	2019-07-29 13:11:06.124532722 +0200
> @@ -746,7 +746,7 @@ static bool
>  c_cpp_flt_eval_method_iec_559 (void)
>  {
>    enum excess_precision_type front_end_ept
> -    = (flag_excess_precision_cmdline == EXCESS_PRECISION_STANDARD
> +    = (flag_excess_precision == EXCESS_PRECISION_STANDARD
>         ? EXCESS_PRECISION_TYPE_STANDARD
>         : EXCESS_PRECISION_TYPE_FAST);
>  
> --- gcc/c-family/c-opts.c.jj	2019-07-10 15:52:20.364155611 +0200
> +++ gcc/c-family/c-opts.c	2019-07-29 13:11:37.110078371 +0200
> @@ -800,14 +800,13 @@ c_common_post_options (const char **pfil
>       support.  */
>    if (c_dialect_cxx ())
>      {
> -      if (flag_excess_precision_cmdline == EXCESS_PRECISION_STANDARD)
> +      if (flag_excess_precision == EXCESS_PRECISION_STANDARD)
>  	sorry ("%<-fexcess-precision=standard%> for C++");
> -      flag_excess_precision_cmdline = EXCESS_PRECISION_FAST;
> +      flag_excess_precision = EXCESS_PRECISION_FAST;
>      }
> -  else if (flag_excess_precision_cmdline == EXCESS_PRECISION_DEFAULT)
> -    flag_excess_precision_cmdline = (flag_iso
> -				     ? EXCESS_PRECISION_STANDARD
> -				     : EXCESS_PRECISION_FAST);
> +  else if (flag_excess_precision == EXCESS_PRECISION_DEFAULT)
> +    flag_excess_precision = (flag_iso ? EXCESS_PRECISION_STANDARD
> +				      : EXCESS_PRECISION_FAST);
>  
>    /* ISO C restricts floating-point expression contraction to within
>       source-language expressions (-ffp-contract=on, currently an alias
> --- gcc/ada/gcc-interface/misc.c.jj	2019-03-11 22:56:52.475722730 +0100
> +++ gcc/ada/gcc-interface/misc.c	2019-07-29 13:08:05.768177333 +0200
> @@ -255,9 +255,9 @@ static bool
>  gnat_post_options (const char **pfilename ATTRIBUTE_UNUSED)
>  {
>    /* Excess precision other than "fast" requires front-end support.  */
> -  if (flag_excess_precision_cmdline == EXCESS_PRECISION_STANDARD)
> +  if (flag_excess_precision == EXCESS_PRECISION_STANDARD)
>      sorry ("%<-fexcess-precision=standard%> for Ada");
> -  flag_excess_precision_cmdline = EXCESS_PRECISION_FAST;
> +  flag_excess_precision = EXCESS_PRECISION_FAST;
>  
>    /* No psABI change warnings for Ada.  */
>    warn_psabi = 0;
> --- gcc/brig/brig-lang.c.jj	2019-06-25 16:03:28.221358420 +0200
> +++ gcc/brig/brig-lang.c	2019-07-29 13:09:20.341083849 +0200
> @@ -166,8 +166,8 @@ brig_langhook_handle_option
>  static bool
>  brig_langhook_post_options (const char **pfilename ATTRIBUTE_UNUSED)
>  {
> -  if (flag_excess_precision_cmdline == EXCESS_PRECISION_DEFAULT)
> -    flag_excess_precision_cmdline = EXCESS_PRECISION_STANDARD;
> +  if (flag_excess_precision == EXCESS_PRECISION_DEFAULT)
> +    flag_excess_precision = EXCESS_PRECISION_STANDARD;
>  
>    /* gccbrig casts pointers around like crazy, TBAA might produce broken
>       code if not disabling it by default.  Some PRM conformance tests such
> --- gcc/d/d-lang.cc.jj	2019-05-20 11:39:15.581117453 +0200
> +++ gcc/d/d-lang.cc	2019-07-29 13:10:05.205425991 +0200
> @@ -772,8 +772,8 @@ d_post_options (const char ** fn)
>    if (global_options_set.x_flag_max_errors)
>      global.errorLimit = flag_max_errors;
>  
> -  if (flag_excess_precision_cmdline == EXCESS_PRECISION_DEFAULT)
> -    flag_excess_precision_cmdline = EXCESS_PRECISION_STANDARD;
> +  if (flag_excess_precision == EXCESS_PRECISION_DEFAULT)
> +    flag_excess_precision = EXCESS_PRECISION_STANDARD;
>  
>    if (global.params.useUnitTests)
>      global.params.useAssert = true;
> --- gcc/fortran/options.c.jj	2019-07-17 09:02:50.211382394 +0200
> +++ gcc/fortran/options.c	2019-07-29 13:07:34.492635929 +0200
> @@ -262,11 +262,9 @@ gfc_post_options (const char **pfilename
>    /* Finalize DEC flags.  */
>    post_dec_flags (flag_dec);
>  
> -  /* Excess precision other than "fast" requires front-end
> -     support.  */
> -  if (flag_excess_precision_cmdline == EXCESS_PRECISION_STANDARD)
> +  if (flag_excess_precision == EXCESS_PRECISION_STANDARD)
>      sorry ("%<-fexcess-precision=standard%> for Fortran");
> -  flag_excess_precision_cmdline = EXCESS_PRECISION_FAST;
> +  flag_excess_precision = EXCESS_PRECISION_FAST;
>  
>    /* Fortran allows associative math - but we cannot reassociate if
>       we want traps or signed zeros. Cf. also flag_protect_parens.  */
> --- gcc/go/go-lang.c.jj	2019-05-08 09:18:28.516742244 +0200
> +++ gcc/go/go-lang.c	2019-07-29 13:09:41.755769840 +0200
> @@ -293,8 +293,8 @@ go_langhook_post_options (const char **p
>      go_add_search_path (dir);
>    go_search_dirs.release ();
>  
> -  if (flag_excess_precision_cmdline == EXCESS_PRECISION_DEFAULT)
> -    flag_excess_precision_cmdline = EXCESS_PRECISION_STANDARD;
> +  if (flag_excess_precision == EXCESS_PRECISION_DEFAULT)
> +    flag_excess_precision = EXCESS_PRECISION_STANDARD;
>  
>    /* Tail call optimizations can confuse uses of runtime.Callers.  */
>    if (!global_options_set.x_flag_optimize_sibling_calls)
> --- gcc/lto/lto-lang.c.jj	2019-06-25 16:03:29.794334463 +0200
> +++ gcc/lto/lto-lang.c	2019-07-29 13:06:52.453252365 +0200
> @@ -925,9 +925,8 @@ lto_post_options (const char **pfilename
>        break;
>      }
>  
> -  /* Excess precision other than "fast" requires front-end
> -     support.  */
> -  flag_excess_precision_cmdline = EXCESS_PRECISION_FAST;
> +  if (flag_excess_precision == EXCESS_PRECISION_DEFAULT)
> +    flag_excess_precision = EXCESS_PRECISION_FAST;
>  
>    /* When partitioning, we can tear appart STRING_CSTs uses from the same
>       TU into multiple partitions.  Without constant merging the constants
> 
> 
> 	Jakub
> 

-- 
Richard Biener <rguenther@suse.de>
SUSE Linux GmbH, Maxfeldstrasse 5, 90409 Nuernberg, Germany;
GF: Felix Imendörffer, Mary Higgins, Sri Rasiah; HRB 21284 (AG Nürnberg)