[PATCH 0/3] arm: fix problems when targetting extended FPUs [PR101723]

[PATCH 0/3] arm: fix problems when targetting extended FPUs [PR101723]

[Richard Earnshaw]

This patch series addresses an issue that has come to light due to a
change in the way GAS handles .fpu directives in the assembler.  A fix
to the assembler made in binutils 2.34 to clear out all features
realated to the FPU when .fpu is emitted has started causing problems
for GCC because of the order in which we emit .fpu and .arch_extension
directives.  To fully address this we need to re-organize the way in
which the compiler does this.

I'll hold of pushing the patches for a couple of days.  Although I've
gone through the testsuite quite carefully and run this through
several configurations, it's possible that this may have some impact
on the testsuite that I've missed.  Christophe, is the any chance you
can run this through your test environment before I commit this?

R.

Richard Earnshaw (3):
  arm: ensure the arch_name is always set for the build target
  arm: Don't reconfigure globals in arm_configure_build_target
  arm: reorder assembler architecture directives [PR101723]

 gcc/config/arm/arm-c.c                        |   1 +
 gcc/config/arm/arm-cpus.in                    |   1 +
 gcc/config/arm/arm.c                          | 190 ++++++++----------
 gcc/testsuite/gcc.target/arm/attr-neon.c      |   9 +-
 gcc/testsuite/gcc.target/arm/attr-neon2.c     |  35 +++-
 gcc/testsuite/gcc.target/arm/attr-neon3.c     |  43 +++-
 .../arm/cortex-m55-nofp-flag-hard.c           |   2 +-
 .../arm/cortex-m55-nofp-flag-softfp.c         |   2 +-
 .../arm/cortex-m55-nofp-nomve-flag-softfp.c   |   2 +-
 .../gcc.target/arm/mve/intrinsics/mve_fpu1.c  |   5 +-
 .../gcc.target/arm/mve/intrinsics/mve_fpu2.c  |   5 +-
 gcc/testsuite/gcc.target/arm/pr98636.c        |   3 +-
 12 files changed, 153 insertions(+), 145 deletions(-)

-- 
2.25.1

[PATCH 1/3] arm: ensure the arch_name is always set for the build target

[Richard Earnshaw]

[-- Attachment #1: Type: text/plain, Size: 624 bytes --]


This should never happen now if GCC is invoked by the driver, but in
the unusual case of calling cc1 (or its ilk) directly from the command
line the build target's arch_name string can remain NULL.  This can
complicate later processing meaning that we need to check for this
case explicitly in some circumstances.  Nothing should rely on this
behaviour, so it's simpler to always set the arch_name when
configuring the build target and be done with it.

gcc:

	* config/arm/arm.c (arm_configure_build_target): Ensure the target's
	arch_name is always set.
---
 gcc/config/arm/arm.c | 2 ++
 1 file changed, 2 insertions(+)


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0001-arm-ensure-the-arch_name-is-always-set-for-the-build.patch --]
[-- Type: text/x-patch; name="0001-arm-ensure-the-arch_name-is-always-set-for-the-build.patch", Size: 597 bytes --]

diff --git a/gcc/config/arm/arm.c b/gcc/config/arm/arm.c
index 6d781e23ee9..b2dd58d8751 100644
--- a/gcc/config/arm/arm.c
+++ b/gcc/config/arm/arm.c
@@ -3432,6 +3432,8 @@ arm_configure_build_target (struct arm_build_target *target,
   const cpu_tune *tune_data = &all_tunes[arm_selected_tune - all_cores];
 
   /* Finish initializing the target structure.  */
+  if (!target->arch_name)
+    target->arch_name = arm_selected_arch->common.name;
   target->arch_pp_name = arm_selected_arch->arch;
   target->base_arch = arm_selected_arch->base_arch;
   target->profile = arm_selected_arch->profile;

[PATCH 2/3] arm: Don't reconfigure globals in arm_configure_build_target

[Richard Earnshaw]

[-- Attachment #1: Type: text/plain, Size: 1266 bytes --]


arm_configure_build_target is usually used to reconfigure the
arm_active_target structure, which is then used to reconfigure a
number of other global variables describing the current target.
Occasionally, however, we need to use arm_configure_build_target to
construct a temporary target structure and in that case it is wrong to
try to reconfigure the global variables (although probably harmless,
since arm_option_reconfigure_globals() only looks at
arm_active_target).  At the very least, however, this is wasted work,
so it is best not to do it unless needed.  What's more, several
callers of arm_configure_build target call
arm_option_reconfigure_globals themselves within a few lines, making
the call from within arm_configure_build_target completely redundant.

So this patch moves the responsibility of calling of
arm_configure_build_target to its callers (only two places needed
updating).

gcc:
	* config/arm/arm.c (arm_configure_build_target): Don't call
	arm_option_reconfigure_globals.
	(arm_option_restore): Call arm_option_reconfigure_globals after
	reconfiguring the target.
	* config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
---
 gcc/config/arm/arm-c.c | 1 +
 gcc/config/arm/arm.c   | 2 +-
 2 files changed, 2 insertions(+), 1 deletion(-)


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0002-arm-Don-t-reconfigure-globals-in-arm_configure_build.patch --]
[-- Type: text/x-patch; name="0002-arm-Don-t-reconfigure-globals-in-arm_configure_build.patch", Size: 1263 bytes --]

diff --git a/gcc/config/arm/arm-c.c b/gcc/config/arm/arm-c.c
index ae2139c4bfa..cc7901bca8d 100644
--- a/gcc/config/arm/arm-c.c
+++ b/gcc/config/arm/arm-c.c
@@ -409,6 +409,7 @@ arm_pragma_target_parse (tree args, tree pop_target)
       target_option_current_node = cur_tree;
       arm_configure_build_target (&arm_active_target,
 				  TREE_TARGET_OPTION (cur_tree), false);
+      arm_option_reconfigure_globals ();
     }
 
   /* Update macros if target_node changes. The global state will be restored
diff --git a/gcc/config/arm/arm.c b/gcc/config/arm/arm.c
index b2dd58d8751..273202ac2fd 100644
--- a/gcc/config/arm/arm.c
+++ b/gcc/config/arm/arm.c
@@ -3058,6 +3058,7 @@ arm_option_restore (struct gcc_options */* opts */,
 		    struct cl_target_option *ptr)
 {
   arm_configure_build_target (&arm_active_target, ptr, false);
+  arm_option_reconfigure_globals ();
 }
 
 /* Reset options between modes that the user has specified.  */
@@ -3441,7 +3442,6 @@ arm_configure_build_target (struct arm_build_target *target,
   target->tune_flags = tune_data->tune_flags;
   target->tune = tune_data->tune;
   target->tune_core = tune_data->scheduler;
-  arm_option_reconfigure_globals ();
 }
 
 /* Fix up any incompatible options that the user has specified.  */

[PATCH 3/3] arm: reorder assembler architecture directives [PR101723]

[Richard Earnshaw]

[-- Attachment #1: Type: text/plain, Size: 3876 bytes --]


A change to the way gas interprets the .fpu directive in binutils-2.34
means that issuing .fpu will clear any features set by .arch_extension
that apply to the floating point or simd units.  This unfortunately
causes problems for more recent versions of the architecture because
we currently emit .arch, .arch_extension and .fpu directives at
different times and try to suppress redundant changes.

This change addresses this by firstly unifying all the places where we
emit these directives to a single block of code and secondly
(re)emitting all the directives if any changes have been made to the
target options.  Whilst this is slightly more than the strict minimum
it should be enough to catch all cases where a change could have
happened.  The new code also emits the directives in the order: .arch,
.fpu, .arch_extension.  This ensures that the additional architectural
extensions are not removed by a later .fpu directive.

Whilst writing this patch I also noticed that in the corner case where
the last function to be compiled had a non-standard set of
architecture flags, the assembler would add an incorrect set of
derived attributes for the file as a whole.  Instead of reflecting the
command-line options it would reflect the flags from the last file in
the function.  To address this I've also added a call to re-emit the
flags from the asm_file_end callback so the assembler will be in the
correct state when it finishes processing the intput.

There's some slight churn to the testsuite as a consequence of this,
because previously we had a hack to suppress emitting a .fpu directive
for one specific case, but with the new order this is no-longer
necessary.

gcc/ChangeLog:

	PR target/101723
	* config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
	writing .cpu directive in asm output.
	* config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
	(arm_last_printed_arch_string): Delete.
	(arm_last-printed_fpu_string): Delete.
	(arm_configure_build_target): If use of floating-point/SIMD is
	disabled, remove all fp/simd related features from the target ISA.
	(last_arm_targ_options): New variable.
	(arm_print_asm_arch_directives): Add new parameters.  Change order
	of emitted directives and handle all cases here.
	(arm_file_start): Always call arm_print_asm_arch_directives, move
	all generation of .arch/.arch_extension here.
	(arm_file_end): Call arm_print_asm_arch.
	(arm_declare_function_name): Call arm_print_asm_arch_directives
	instead of printing .arch/.fpu directives directly.

gcc/testsuite/ChangeLog:

	PR target/101723
	* gcc.target/arm/cortex-m55-nofp-flag-hard.c: Update expected output.
	* gcc.target/arm/cortex-m55-nofp-flag-softfp.c: Likewise.
	* gcc.target/arm/cortex-m55-nofp-nomve-flag-softfp.c: Likewise.
	* gcc.target/arm/mve/intrinsics/mve_fpu1.c: Convert to dg-do assemble.
	Add a non-no-op function body.
	* gcc.target/arm/mve/intrinsics/mve_fpu2.c: Likewise.
	* gcc.target/arm/pr98636.c (dg-options): Add -mfloat-abi=softfp.
	* gcc.target/arm/attr-neon.c: Tighten scan-assembler tests.
	* gcc.target/arm/attr-neon2.c: Use -Ofast, convert test to use
	check-function-bodies.
	* gcc.target/arm/attr-neon3.c: Likewise.
---
 gcc/config/arm/arm-cpus.in                    |   1 +
 gcc/config/arm/arm.c                          | 186 ++++++++----------
 gcc/testsuite/gcc.target/arm/attr-neon.c      |   9 +-
 gcc/testsuite/gcc.target/arm/attr-neon2.c     |  35 ++--
 gcc/testsuite/gcc.target/arm/attr-neon3.c     |  43 ++--
 .../arm/cortex-m55-nofp-flag-hard.c           |   2 +-
 .../arm/cortex-m55-nofp-flag-softfp.c         |   2 +-
 .../arm/cortex-m55-nofp-nomve-flag-softfp.c   |   2 +-
 .../gcc.target/arm/mve/intrinsics/mve_fpu1.c  |   5 +-
 .../gcc.target/arm/mve/intrinsics/mve_fpu2.c  |   5 +-
 gcc/testsuite/gcc.target/arm/pr98636.c        |   3 +-
 11 files changed, 149 insertions(+), 144 deletions(-)


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0003-arm-reorder-assembler-architecture-directives-PR1017.patch --]
[-- Type: text/x-patch; name="0003-arm-reorder-assembler-architecture-directives-PR1017.patch", Size: 20872 bytes --]

diff --git a/gcc/config/arm/arm-cpus.in b/gcc/config/arm/arm-cpus.in
index ab4b6acf5ea..249995a6bca 100644
--- a/gcc/config/arm/arm-cpus.in
+++ b/gcc/config/arm/arm-cpus.in
@@ -1080,6 +1080,7 @@ begin cpu generic-armv7-a
  cname genericv7a
  tune flags LDSCHED
  architecture armv7-a+fp
+ isa quirk_no_asmcpu
  option mp add mp
  option sec add sec
  option vfpv3-d16 add VFPv3 FP_DBL
diff --git a/gcc/config/arm/arm.c b/gcc/config/arm/arm.c
index 273202ac2fd..11dafc70067 100644
--- a/gcc/config/arm/arm.c
+++ b/gcc/config/arm/arm.c
@@ -79,10 +79,6 @@
 typedef struct minipool_node    Mnode;
 typedef struct minipool_fixup   Mfix;
 
-/* The last .arch and .fpu assembly strings that we printed.  */
-static std::string arm_last_printed_arch_string;
-static std::string arm_last_printed_fpu_string;
-
 void (*arm_lang_output_object_attributes_hook)(void);
 
 struct four_ints
@@ -334,6 +330,7 @@ static rtx_insn *thumb1_md_asm_adjust (vec<rtx> &, vec<rtx> &,
 				       vec<machine_mode> &,
 				       vec<const char *> &, vec<rtx> &,
 				       HARD_REG_SET &, location_t);
+static const char *arm_identify_fpu_from_isa (sbitmap);
 \f
 /* Table of machine attributes.  */
 static const struct attribute_spec arm_attribute_table[] =
@@ -3411,6 +3408,11 @@ arm_configure_build_target (struct arm_build_target *target,
       bitmap_ior (target->isa, target->isa, fpu_bits);
     }
 
+  /* If we have the soft-float ABI, clear any feature bits relating to use of
+     floating-point operations.  They'll just confuse things later on.  */
+  if (arm_float_abi == ARM_FLOAT_ABI_SOFT)
+    bitmap_and_compl (target->isa, target->isa, isa_all_fpbits);
+
   /* There may be implied bits which we still need to enable. These are
      non-named features which are needed to complete other sets of features,
      but cannot be enabled from arm-cpus.in due to being shared between
@@ -28096,20 +28098,65 @@ arm_print_tune_info (void)
 	       (int) current_tune->sched_autopref);
 }
 
+/* The last set of target options used to emit .arch directives, etc.  This
+   could be a function-local static if it were not required to expose it as a
+   root to the garbage collector.  */
+static GTY(()) cl_target_option *last_asm_targ_options = NULL;
+
 /* Print .arch and .arch_extension directives corresponding to the
    current architecture configuration.  */
 static void
-arm_print_asm_arch_directives ()
+arm_print_asm_arch_directives (FILE *stream, cl_target_option *targ_options)
 {
+  arm_build_target build_target;
+  /* If the target options haven't changed since the last time we were called
+     there is nothing to do.  This should be sufficient to suppress the
+     majority of redundant work.  */
+  if (last_asm_targ_options == targ_options)
+    return;
+
+  last_asm_targ_options = targ_options;
+
+  build_target.isa = sbitmap_alloc (isa_num_bits);
+  arm_configure_build_target (&build_target, targ_options, false);
+
+  if (build_target.core_name
+      && !bitmap_bit_p (build_target.isa, isa_bit_quirk_no_asmcpu))
+    {
+      const char* truncated_name
+	= arm_rewrite_selected_cpu (build_target.core_name);
+      asm_fprintf (stream, "\t.cpu %s\n", truncated_name);
+    }
+
   const arch_option *arch
     = arm_parse_arch_option_name (all_architectures, "-march",
-				  arm_active_target.arch_name);
+				  build_target.arch_name);
   auto_sbitmap opt_bits (isa_num_bits);
 
   gcc_assert (arch);
 
-  asm_fprintf (asm_out_file, "\t.arch %s\n", arm_active_target.arch_name);
-  arm_last_printed_arch_string = arm_active_target.arch_name;
+  if (strcmp (build_target.arch_name, "armv7ve") == 0)
+    {
+      /* Keep backward compatability for assemblers which don't support
+	 armv7ve.  Fortunately, none of the following extensions are reset
+	 by a .fpu directive.  */
+      asm_fprintf (stream, "\t.arch armv7-a\n");
+      asm_fprintf (stream, "\t.arch_extension virt\n");
+      asm_fprintf (stream, "\t.arch_extension idiv\n");
+      asm_fprintf (stream, "\t.arch_extension sec\n");
+      asm_fprintf (stream, "\t.arch_extension mp\n");
+    }
+  else
+    asm_fprintf (stream, "\t.arch %s\n", build_target.arch_name);
+
+  /* The .fpu directive will reset any architecture extensions from the
+     assembler that relate to the fp/vector extensions.  So put this out before
+     any .arch_extension directives.  */
+  const char *fpu_name = (TARGET_SOFT_FLOAT
+			  ? "softvfp"
+			  : arm_identify_fpu_from_isa (build_target.isa));
+  asm_fprintf (stream, "\t.fpu %s\n", fpu_name);
+
   if (!arch->common.extensions)
     return;
 
@@ -28135,13 +28182,12 @@ arm_print_asm_arch_directives ()
 		  && !TARGET_HAVE_MVE_FLOAT))
 	    continue;
 
-	  /* If every feature bit of this option is set in the target
-	     ISA specification, print out the option name.  However,
-	     don't print anything if all the bits are part of the
-	     FPU specification.  */
-	  if (bitmap_subset_p (opt_bits, arm_active_target.isa)
+	  /* If every feature bit of this option is set in the target ISA
+	     specification, print out the option name.  However, don't print
+	     anything if all the bits are part of the FPU specification.  */
+	  if (bitmap_subset_p (opt_bits, build_target.isa)
 	      && !bitmap_subset_p (opt_bits, isa_all_fpubits_internal))
-	    asm_fprintf (asm_out_file, "\t.arch_extension %s\n", opt->name);
+	    asm_fprintf (stream, "\t.arch_extension %s\n", opt->name);
 	}
     }
 }
@@ -28151,46 +28197,23 @@ arm_file_start (void)
 {
   int val;
 
+  arm_print_asm_arch_directives
+    (asm_out_file, TREE_TARGET_OPTION (target_option_default_node));
+
   if (TARGET_BPABI)
     {
-      /* We don't have a specified CPU.  Use the architecture to
-	 generate the tags.
-
-	 Note: it might be better to do this unconditionally, then the
-	 assembler would not need to know about all new CPU names as
-	 they are added.  */
-      if (!arm_active_target.core_name)
-	{
-	  /* armv7ve doesn't support any extensions.  */
-	  if (strcmp (arm_active_target.arch_name, "armv7ve") == 0)
-	    {
-	      /* Keep backward compatability for assemblers
-		 which don't support armv7ve.  */
-	      asm_fprintf (asm_out_file, "\t.arch armv7-a\n");
-	      asm_fprintf (asm_out_file, "\t.arch_extension virt\n");
-	      asm_fprintf (asm_out_file, "\t.arch_extension idiv\n");
-	      asm_fprintf (asm_out_file, "\t.arch_extension sec\n");
-	      asm_fprintf (asm_out_file, "\t.arch_extension mp\n");
-	      arm_last_printed_arch_string = "armv7ve";
-	    }
-	  else
-	    arm_print_asm_arch_directives ();
-	}
-      else if (startswith (arm_active_target.core_name, "generic"))
-	{
-	  asm_fprintf (asm_out_file, "\t.arch %s\n",
-		       arm_active_target.core_name + 8);
-	  arm_last_printed_arch_string = arm_active_target.core_name + 8;
-	}
-      else
+      /* If we have a named cpu, but we the assembler does not support that
+	 name via .cpu, put out a cpu name attribute; but don't do this if the
+	 name starts with the fictitious prefix, 'generic'.  */
+      if (arm_active_target.core_name
+	  && bitmap_bit_p (arm_active_target.isa, isa_bit_quirk_no_asmcpu)
+	  && !startswith (arm_active_target.core_name, "generic"))
 	{
 	  const char* truncated_name
 	    = arm_rewrite_selected_cpu (arm_active_target.core_name);
 	  if (bitmap_bit_p (arm_active_target.isa, isa_bit_quirk_no_asmcpu))
 	    asm_fprintf (asm_out_file, "\t.eabi_attribute 5, \"%s\"\n",
 			 truncated_name);
-	  else
-	    asm_fprintf (asm_out_file, "\t.cpu %s\n", truncated_name);
 	}
 
       if (print_tune_info)
@@ -28255,6 +28278,13 @@ arm_file_end (void)
 {
   int regno;
 
+  /* Just in case the last function output in the assembler had non-default
+     architecture directives, we force the assembler state back to the default
+     set, so that any 'calculated' build attributes are based on the default
+     options rather than the special options for that function.  */
+  arm_print_asm_arch_directives
+    (asm_out_file, TREE_TARGET_OPTION (target_option_default_node));
+
   if (NEED_INDICATE_EXEC_STACK)
     /* Add .note.GNU-stack.  */
     file_end_indicate_exec_stack ();
@@ -33265,58 +33295,7 @@ arm_declare_function_name (FILE *stream, const char *name, tree decl)
     targ_options = TREE_TARGET_OPTION (target_option_current_node);
   gcc_assert (targ_options);
 
-  /* Only update the assembler .arch string if it is distinct from the last
-     such string we printed. arch_to_print is set conditionally in case
-     targ_options->x_arm_arch_string is NULL which can be the case
-     when cc1 is invoked directly without passing -march option.  */
-  std::string arch_to_print;
-  if (targ_options->x_arm_arch_string)
-    arch_to_print = targ_options->x_arm_arch_string;
-
-  if (arch_to_print != arm_last_printed_arch_string)
-    {
-      std::string arch_name
-	= arch_to_print.substr (0, arch_to_print.find ("+"));
-      asm_fprintf (asm_out_file, "\t.arch %s\n", arch_name.c_str ());
-      const arch_option *arch
-	= arm_parse_arch_option_name (all_architectures, "-march",
-				      targ_options->x_arm_arch_string);
-      auto_sbitmap opt_bits (isa_num_bits);
-
-      gcc_assert (arch);
-      if (arch->common.extensions)
-	{
-	  for (const struct cpu_arch_extension *opt = arch->common.extensions;
-	       opt->name != NULL;
-	       opt++)
-	    {
-	      if (!opt->remove)
-		{
-		  arm_initialize_isa (opt_bits, opt->isa_bits);
-		  /* For the cases "-march=armv8.1-m.main+mve -mfloat-abi=soft"
-		     and "-march=armv8.1-m.main+mve.fp -mfloat-abi=soft" MVE and
-		     MVE with floating point instructions is disabled.  So the
-		     following check restricts the printing of ".arch_extension
-		     mve" and ".arch_extension fp" (for mve.fp) in the assembly
-		     file.    MVE needs this special behaviour because the
-		     feature bit "mve" and "mve_float" are not part of
-		     "fpu bits", so they are not cleared when -mfloat-abi=soft
-		     (i.e nofp) but the marco TARGET_HAVE_MVE and
-		     TARGET_HAVE_MVE_FLOAT are disabled.  */
-		  if ((bitmap_bit_p (opt_bits, isa_bit_mve) && !TARGET_HAVE_MVE)
-		      || (bitmap_bit_p (opt_bits, isa_bit_mve_float)
-			  && !TARGET_HAVE_MVE_FLOAT))
-		    continue;
-		  if (bitmap_subset_p (opt_bits, arm_active_target.isa)
-		      && !bitmap_subset_p (opt_bits, isa_all_fpubits_internal))
-		    asm_fprintf (asm_out_file, "\t.arch_extension %s\n",
-				 opt->name);
-		}
-	     }
-	}
-
-      arm_last_printed_arch_string = arch_to_print;
-    }
+  arm_print_asm_arch_directives (stream, targ_options);
 
   fprintf (stream, "\t.syntax unified\n");
 
@@ -33334,17 +33313,6 @@ arm_declare_function_name (FILE *stream, const char *name, tree decl)
   else
     fprintf (stream, "\t.arm\n");
 
-  std::string fpu_to_print
-    = TARGET_SOFT_FLOAT
-	? "softvfp" : arm_identify_fpu_from_isa (arm_active_target.isa);
-
-  if (!(!strcmp (fpu_to_print.c_str (), "softvfp") && TARGET_VFP_BASE)
-      && (fpu_to_print != arm_last_printed_arch_string))
-    {
-      asm_fprintf (asm_out_file, "\t.fpu %s\n", fpu_to_print.c_str ());
-      arm_last_printed_fpu_string = fpu_to_print;
-    }
-
   if (TARGET_POKE_FUNCTION_NAME)
     arm_poke_function_name (stream, (const char *) name);
 }
diff --git a/gcc/testsuite/gcc.target/arm/attr-neon.c b/gcc/testsuite/gcc.target/arm/attr-neon.c
index 225fb8dc3db..e8e3086247d 100644
--- a/gcc/testsuite/gcc.target/arm/attr-neon.c
+++ b/gcc/testsuite/gcc.target/arm/attr-neon.c
@@ -1,7 +1,10 @@
 /* { dg-do compile } */
 /* { dg-require-effective-target arm_neon_ok } */
 /* { dg-options "-O2 -ftree-vectorize" } */
-/* { dg-add-options arm_neon arm_v8_vfp } */ /* The arm_v8_vfp adds -mfpu=fp-armv8 to the command line, overriding any -mfpu= option set by arm_neon, thus ensuring that the attributes below really are checked for correct fpu selection.  */
+/* { dg-add-options arm_neon arm_v8_vfp } */
+/* The arm_v8_vfp adds -mfpu=fp-armv8 to the command line, overriding any
+   -mfpu= option set by arm_neon, thus ensuring that the attributes below
+   really are checked for correct fpu selection.  */
 
 /* Verify that neon instructions are emitted once.  */
 void __attribute__ ((target("fpu=neon")))
@@ -18,6 +21,6 @@ f3(int n, int x[], int y[]) {
     y[i] = x[i] << 3;
 }
 
-/* { dg-final { scan-assembler-times "\.fpu vfp" 1 } } */
-/* { dg-final { scan-assembler-times "\.fpu neon" 1 } } */
+/* { dg-final { scan-assembler-times "\.fpu\\s+vfp\n" 1 } } */
+/* { dg-final { scan-assembler-times "\.fpu\\s+neon\n" 1 } } */
 /* { dg-final { scan-assembler-times "vshl" 1 } } */
diff --git a/gcc/testsuite/gcc.target/arm/attr-neon2.c b/gcc/testsuite/gcc.target/arm/attr-neon2.c
index 29668256cf5..a7a72dac379 100644
--- a/gcc/testsuite/gcc.target/arm/attr-neon2.c
+++ b/gcc/testsuite/gcc.target/arm/attr-neon2.c
@@ -1,7 +1,7 @@
 /* { dg-do compile } */
 /* { dg-require-effective-target arm_neon_ok } */
 /* { dg-require-effective-target arm_fp_ok } */
-/* { dg-options "-O2" } */
+/* { dg-options "-Ofast" } */
 /* { dg-add-options arm_fp } */
 
 /* Reset fpu to a value compatible with the next pragmas.  */
@@ -12,23 +12,36 @@
 #include <arm_neon.h>
 
 /* Check that pragma target is used.  */
-int8x8_t 
-my (int8x8_t __a, int8x8_t __b)
+/*
+**my:
+**	...
+**	vadd.f32	d[0-9]+, d[0-9]+, d[0-9]+
+**	...
+**	bx	lr
+*/
+float32x2_t
+my (float32x2_t __a, float32x2_t __b)
 {
   return __a + __b;
 }
 
 #pragma GCC pop_options
 
-/* Check that command line option is restored.  */
-int8x8_t 
-my1 (int8x8_t __a, int8x8_t __b)
+/* Check that fpu=vfp is restored.  */
+/*
+**my1:
+**	...
+**	vadd.f32	s[0-9]+, s[0-9]+, s[0-9]+
+**	vadd.f32	s[0-9]+, s[0-9]+, s[0-9]+
+**	...
+**	bx	lr
+*/
+float32x2_t
+my1 (float32x2_t __a, float32x2_t __b)
 {
   return __a + __b;
 }
 
-/* { dg-final { scan-assembler-times "\.fpu vfp" 1 } } */
-/* { dg-final { scan-assembler-times "\.fpu neon" 1 } } */
-/* { dg-final { scan-assembler "vadd" } } */
-
-
+/* { dg-final { scan-assembler "\.fpu\\s+vfp\n" } } */
+/* { dg-final { scan-assembler "\.fpu\\s+neon\n" } } */
+/* { dg-final { check-function-bodies "**" "" } } */
diff --git a/gcc/testsuite/gcc.target/arm/attr-neon3.c b/gcc/testsuite/gcc.target/arm/attr-neon3.c
index 17e429ad739..150a4b2da5c 100644
--- a/gcc/testsuite/gcc.target/arm/attr-neon3.c
+++ b/gcc/testsuite/gcc.target/arm/attr-neon3.c
@@ -1,7 +1,7 @@
 /* { dg-do compile } */
 /* { dg-require-effective-target arm_crypto_ok } */
 /* { dg-require-effective-target arm_fp_ok } */
-/* { dg-options "-O2" } */
+/* { dg-options "-Ofast" } */
 /* { dg-add-options arm_fp } */
 
 /* Reset fpu to a value compatible with the next pragmas.  */
@@ -11,28 +11,49 @@
 #include <arm_neon.h>
 
 /* Check that neon is used.  */
-int8x8_t __attribute__ ((target("fpu=neon")))
-my (int8x8_t __a, int8x8_t __b)
+/*
+**my:
+**	...
+**	vadd.f32	d[0-9]+, d[0-9]+, d[0-9]+
+**	...
+**	bx	lr
+*/
+float32x2_t __attribute__ ((target("fpu=neon")))
+my (float32x2_t __a, float32x2_t __b)
 {
   return __a + __b;
 }
 
 /* Check that crypto builtins are recognized.  */
+/*
+**foo:
+**	...
+**	vld1.64	{d[0-9]+-d[0-9]+}, \[r[0-9]+:64\]
+**	...
+**	bx	lr
+*/
+
 poly128_t __attribute__ ((target("fpu=crypto-neon-fp-armv8")))
 foo (poly128_t* ptr)
 {
   return vldrq_p128 (ptr);
 }
 
-/* Check that default mode is restored.  */
-int8x8_t
-my1 (int8x8_t __a, int8x8_t __b)
+/* Check that fpu=vfp is restored.  */
+/*
+**my1:
+**	...
+**	vadd.f32	s[0-9]+, s[0-9]+, s[0-9]+
+**	vadd.f32	s[0-9]+, s[0-9]+, s[0-9]+
+**	...
+**	bx	lr
+*/float32x2_t
+my1 (float32x2_t __a, float32x2_t __b)
 {
   return __a + __b;
 }
 
-/* { dg-final { scan-assembler-times "\.fpu vfp" 1 } } */
-/* { dg-final { scan-assembler-times "\.fpu neon" 1 } } */
-/* { dg-final { scan-assembler-times "\.fpu crypto-neon-fp-armv8" 1 } } */
-/* { dg-final { scan-assembler-times "vld1" 1 } } */
-/* { dg-final { scan-assembler-times "vadd" 1} } */
+/* { dg-final { scan-assembler "\.fpu\\s+vfp\n" } } */
+/* { dg-final { scan-assembler "\.fpu\\s+neon\n" } } */
+/* { dg-final { scan-assembler "\.fpu\\s+crypto-neon-fp-armv8\n" } } */
+/* { dg-final { check-function-bodies "**" "" } } */
diff --git a/gcc/testsuite/gcc.target/arm/cortex-m55-nofp-flag-hard.c b/gcc/testsuite/gcc.target/arm/cortex-m55-nofp-flag-hard.c
index 6a92dedcc59..e0fb307ac30 100644
--- a/gcc/testsuite/gcc.target/arm/cortex-m55-nofp-flag-hard.c
+++ b/gcc/testsuite/gcc.target/arm/cortex-m55-nofp-flag-hard.c
@@ -1,12 +1,12 @@
 /* { dg-do assemble } */
 /* { dg-require-effective-target arm_v8_1m_mve_fp_ok } */
 /* { dg-additional-options "-mcpu=cortex-m55+nofp -mthumb -mfloat-abi=hard -mfpu=auto --save-temps" } */
+/* { dg-final { scan-assembler "\.fpu softvfp" } } */
 /* { dg-final { scan-assembler "\.arch_extension mve" } } */
 /* { dg-final { scan-assembler "\.arch_extension dsp" } } */
 /* { dg-final { scan-assembler-not "\.arch_extension fp" } } */
 /* { dg-final { scan-assembler-not "\.arch_extension fp.dp" } } */
 /* { dg-final { scan-assembler-not "\.arch_extension mve.fp" } } */
-/* { dg-final { scan-assembler-not "\.fpu" } } */
 
 int
 f ()
diff --git a/gcc/testsuite/gcc.target/arm/cortex-m55-nofp-flag-softfp.c b/gcc/testsuite/gcc.target/arm/cortex-m55-nofp-flag-softfp.c
index 25e80e9bd64..50645e8cd0f 100644
--- a/gcc/testsuite/gcc.target/arm/cortex-m55-nofp-flag-softfp.c
+++ b/gcc/testsuite/gcc.target/arm/cortex-m55-nofp-flag-softfp.c
@@ -1,12 +1,12 @@
 /* { dg-do assemble } */
 /* { dg-require-effective-target arm_v8_1m_mve_fp_ok } */
 /* { dg-additional-options "-mcpu=cortex-m55+nofp -mthumb -mfloat-abi=softfp -mfpu=auto --save-temps" } */
+/* { dg-final { scan-assembler "\.fpu softvfp" } } */
 /* { dg-final { scan-assembler "\.arch_extension mve" } } */
 /* { dg-final { scan-assembler "\.arch_extension dsp" } } */
 /* { dg-final { scan-assembler-not "\.arch_extension fp" } } */
 /* { dg-final { scan-assembler-not "\.arch_extension fp.dp" } } */
 /* { dg-final { scan-assembler-not "\.arch_extension mve.fp" } } */
-/* { dg-final { scan-assembler-not "\.fpu" } } */
 
 int
 f ()
diff --git a/gcc/testsuite/gcc.target/arm/cortex-m55-nofp-nomve-flag-softfp.c b/gcc/testsuite/gcc.target/arm/cortex-m55-nofp-nomve-flag-softfp.c
index 38042cc490c..948f622633c 100644
--- a/gcc/testsuite/gcc.target/arm/cortex-m55-nofp-nomve-flag-softfp.c
+++ b/gcc/testsuite/gcc.target/arm/cortex-m55-nofp-nomve-flag-softfp.c
@@ -1,12 +1,12 @@
 /* { dg-do assemble } */
 /* { dg-require-effective-target arm_v8_1m_mve_fp_ok } */
 /* { dg-additional-options "-mcpu=cortex-m55+nomve+nofp -mthumb -mfloat-abi=softfp -mfpu=auto --save-temps" } */
+/* { dg-final { scan-assembler "\.fpu softvfp" } } */
 /* { dg-final { scan-assembler-not "\.arch_extension mve" } } */
 /* { dg-final { scan-assembler-not "\.arch_extension mve.fp" } } */
 /* { dg-final { scan-assembler-not "\.arch_extension fp" } } */
 /* { dg-final { scan-assembler-not "\.arch_extension fp.dp" } } */
 /* { dg-final { scan-assembler "\.arch_extension dsp" } } */
-/* { dg-final { scan-assembler-not "\.fpu" } } */
 
 int
 f ()
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/mve_fpu1.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/mve_fpu1.c
index 611097ec956..c5acdb5e4e5 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/mve_fpu1.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/mve_fpu1.c
@@ -1,3 +1,4 @@
+/* { dg-do assemble } */
 /* { dg-require-effective-target arm_v8_1m_mve_ok } */
 /* { dg-require-effective-target arm_hard_ok } */
 /* { dg-add-options arm_v8_1m_mve } */
@@ -8,8 +9,6 @@
 int8x16_t
 foo1 (int8x16_t value)
 {
-  int8x16_t b = value;
+  int8x16_t b = -value;
   return b;
 }
-
-/* { dg-final { scan-assembler-not "\.fpu softvfp" }  } */
diff --git a/gcc/testsuite/gcc.target/arm/mve/intrinsics/mve_fpu2.c b/gcc/testsuite/gcc.target/arm/mve/intrinsics/mve_fpu2.c
index b8e105111de..907db5e1572 100644
--- a/gcc/testsuite/gcc.target/arm/mve/intrinsics/mve_fpu2.c
+++ b/gcc/testsuite/gcc.target/arm/mve/intrinsics/mve_fpu2.c
@@ -1,3 +1,4 @@
+/* { dg-do assemble } */
 /* { dg-require-effective-target arm_v8_1m_mve_ok } */
 /* { dg-require-effective-target arm_softfp_ok } */
 /* { dg-add-options arm_v8_1m_mve } */
@@ -8,8 +9,6 @@
 int8x16_t
 foo1 (int8x16_t value)
 {
-  int8x16_t b = value;
+  int8x16_t b = -value;
   return b;
 }
-
-/* { dg-final { scan-assembler-not "\.fpu softvfp" }  } */
diff --git a/gcc/testsuite/gcc.target/arm/pr98636.c b/gcc/testsuite/gcc.target/arm/pr98636.c
index c4d235cb43d..559f9a26c1e 100644
--- a/gcc/testsuite/gcc.target/arm/pr98636.c
+++ b/gcc/testsuite/gcc.target/arm/pr98636.c
@@ -1,5 +1,6 @@
 /* { dg-do compile } */
-/* { dg-options "-mfp16-format=alternative" } */
+/* { dg-require-effective-target arm_softfp_ok } */
+/* { dg-options "-mfp16-format=alternative -mfloat-abi=softfp" } */
 
 #pragma GCC push_options
 # pragma GCC target ("arch=armv8.2-a+fp16") /* { dg-error "selected fp16 options are incompatible" } */

Re: [PATCH 0/3] arm: fix problems when targetting extended FPUs [PR101723]

[Christophe Lyon]

On Mon, Aug 2, 2021 at 4:57 PM Richard Earnshaw <rearnsha@arm.com> wrote:

> This patch series addresses an issue that has come to light due to a
> change in the way GAS handles .fpu directives in the assembler.  A fix
> to the assembler made in binutils 2.34 to clear out all features
> realated to the FPU when .fpu is emitted has started causing problems
> for GCC because of the order in which we emit .fpu and .arch_extension
> directives.  To fully address this we need to re-organize the way in
> which the compiler does this.
>
> I'll hold of pushing the patches for a couple of days.  Although I've
> gone through the testsuite quite carefully and run this through
> several configurations, it's possible that this may have some impact
> on the testsuite that I've missed.  Christophe, is the any chance you
> can run this through your test environment before I commit this?
>
>
Sorry for the delay, still unpacking emails after hollidays.

Yes I can run the validation for these patches. I think you mean with all 3
patches combined, not 3 validations (patch 1, patches 1+2, patches 1-3) ?

Thanks,

Christophe


> R.
>
> Richard Earnshaw (3):
>   arm: ensure the arch_name is always set for the build target
>   arm: Don't reconfigure globals in arm_configure_build_target
>   arm: reorder assembler architecture directives [PR101723]
>
>  gcc/config/arm/arm-c.c                        |   1 +
>  gcc/config/arm/arm-cpus.in                    |   1 +
>  gcc/config/arm/arm.c                          | 190 ++++++++----------
>  gcc/testsuite/gcc.target/arm/attr-neon.c      |   9 +-
>  gcc/testsuite/gcc.target/arm/attr-neon2.c     |  35 +++-
>  gcc/testsuite/gcc.target/arm/attr-neon3.c     |  43 +++-
>  .../arm/cortex-m55-nofp-flag-hard.c           |   2 +-
>  .../arm/cortex-m55-nofp-flag-softfp.c         |   2 +-
>  .../arm/cortex-m55-nofp-nomve-flag-softfp.c   |   2 +-
>  .../gcc.target/arm/mve/intrinsics/mve_fpu1.c  |   5 +-
>  .../gcc.target/arm/mve/intrinsics/mve_fpu2.c  |   5 +-
>  gcc/testsuite/gcc.target/arm/pr98636.c        |   3 +-
>  12 files changed, 153 insertions(+), 145 deletions(-)
>
> --
> 2.25.1
>
>

Re: [PATCH 0/3] arm: fix problems when targetting extended FPUs [PR101723]

[Richard Earnshaw]

On 03/08/2021 16:04, Christophe Lyon via Gcc-patches wrote:
> On Mon, Aug 2, 2021 at 4:57 PM Richard Earnshaw <rearnsha@arm.com> wrote:
> 
>> This patch series addresses an issue that has come to light due to a
>> change in the way GAS handles .fpu directives in the assembler.  A fix
>> to the assembler made in binutils 2.34 to clear out all features
>> realated to the FPU when .fpu is emitted has started causing problems
>> for GCC because of the order in which we emit .fpu and .arch_extension
>> directives.  To fully address this we need to re-organize the way in
>> which the compiler does this.
>>
>> I'll hold of pushing the patches for a couple of days.  Although I've
>> gone through the testsuite quite carefully and run this through
>> several configurations, it's possible that this may have some impact
>> on the testsuite that I've missed.  Christophe, is the any chance you
>> can run this through your test environment before I commit this?
>>
>>
> Sorry for the delay, still unpacking emails after hollidays.
> 
> Yes I can run the validation for these patches. I think you mean with all 3
> patches combined, not 3 validations (patch 1, patches 1+2, patches 1-3) ?

Yes, the first two are trivial changes that just support the interesting 
one, which is the final patch.

R.
> 
> Thanks,
> 
> Christophe
> 
> 
>> R.
>>
>> Richard Earnshaw (3):
>>    arm: ensure the arch_name is always set for the build target
>>    arm: Don't reconfigure globals in arm_configure_build_target
>>    arm: reorder assembler architecture directives [PR101723]
>>
>>   gcc/config/arm/arm-c.c                        |   1 +
>>   gcc/config/arm/arm-cpus.in                    |   1 +
>>   gcc/config/arm/arm.c                          | 190 ++++++++----------
>>   gcc/testsuite/gcc.target/arm/attr-neon.c      |   9 +-
>>   gcc/testsuite/gcc.target/arm/attr-neon2.c     |  35 +++-
>>   gcc/testsuite/gcc.target/arm/attr-neon3.c     |  43 +++-
>>   .../arm/cortex-m55-nofp-flag-hard.c           |   2 +-
>>   .../arm/cortex-m55-nofp-flag-softfp.c         |   2 +-
>>   .../arm/cortex-m55-nofp-nomve-flag-softfp.c   |   2 +-
>>   .../gcc.target/arm/mve/intrinsics/mve_fpu1.c  |   5 +-
>>   .../gcc.target/arm/mve/intrinsics/mve_fpu2.c  |   5 +-
>>   gcc/testsuite/gcc.target/arm/pr98636.c        |   3 +-
>>   12 files changed, 153 insertions(+), 145 deletions(-)
>>
>> --
>> 2.25.1
>>
>>

Re: [PATCH 0/3] arm: fix problems when targetting extended FPUs [PR101723]

[Christophe Lyon]

On Tue, Aug 3, 2021 at 5:40 PM Richard Earnshaw <
Richard.Earnshaw@foss.arm.com> wrote:

>
>
> On 03/08/2021 16:04, Christophe Lyon via Gcc-patches wrote:
> > On Mon, Aug 2, 2021 at 4:57 PM Richard Earnshaw <rearnsha@arm.com>
> wrote:
> >
> >> This patch series addresses an issue that has come to light due to a
> >> change in the way GAS handles .fpu directives in the assembler.  A fix
> >> to the assembler made in binutils 2.34 to clear out all features
> >> realated to the FPU when .fpu is emitted has started causing problems
> >> for GCC because of the order in which we emit .fpu and .arch_extension
> >> directives.  To fully address this we need to re-organize the way in
> >> which the compiler does this.
> >>
> >> I'll hold of pushing the patches for a couple of days.  Although I've
> >> gone through the testsuite quite carefully and run this through
> >> several configurations, it's possible that this may have some impact
> >> on the testsuite that I've missed.  Christophe, is the any chance you
> >> can run this through your test environment before I commit this?
> >>
> >>
> > Sorry for the delay, still unpacking emails after hollidays.
> >
> > Yes I can run the validation for these patches. I think you mean with
> all 3
> > patches combined, not 3 validations (patch 1, patches 1+2, patches 1-3) ?
>
> Yes, the first two are trivial changes that just support the interesting
> one, which is the final patch.
>
>
Hi Richard,

There are a few regressions with these 3 patches applied, see:

https://people.linaro.org/~christophe.lyon/cross-validation/gcc-test-patches/r12-2683-g4d17ca1bc74109e5cc4ef34890b6293c4bcb1d6a-PR101723.patch/report-build-info.html

The cortex-m55-nofp-* failures reported in several configs are not actual
regressions, it seems: they were already failing, but with a different
scan-assembler string, so they are considered different tests.
I think I sent patches for these unresolved cortex-m55-nofp-*
testcases several weeks/months ago, but I'd have to check.

There are regressions when configured --with-cpu cortex-a5 --with-fpu
vfpv3-d16-fp16 as well as on armeb.

Thanks,

Christophe

R.
> >
> > Thanks,
> >
> > Christophe
> >
> >
> >> R.
> >>
> >> Richard Earnshaw (3):
> >>    arm: ensure the arch_name is always set for the build target
> >>    arm: Don't reconfigure globals in arm_configure_build_target
> >>    arm: reorder assembler architecture directives [PR101723]
> >>
> >>   gcc/config/arm/arm-c.c                        |   1 +
> >>   gcc/config/arm/arm-cpus.in                    |   1 +
> >>   gcc/config/arm/arm.c                          | 190 ++++++++----------
> >>   gcc/testsuite/gcc.target/arm/attr-neon.c      |   9 +-
> >>   gcc/testsuite/gcc.target/arm/attr-neon2.c     |  35 +++-
> >>   gcc/testsuite/gcc.target/arm/attr-neon3.c     |  43 +++-
> >>   .../arm/cortex-m55-nofp-flag-hard.c           |   2 +-
> >>   .../arm/cortex-m55-nofp-flag-softfp.c         |   2 +-
> >>   .../arm/cortex-m55-nofp-nomve-flag-softfp.c   |   2 +-
> >>   .../gcc.target/arm/mve/intrinsics/mve_fpu1.c  |   5 +-
> >>   .../gcc.target/arm/mve/intrinsics/mve_fpu2.c  |   5 +-
> >>   gcc/testsuite/gcc.target/arm/pr98636.c        |   3 +-
> >>   12 files changed, 153 insertions(+), 145 deletions(-)
> >>
> >> --
> >> 2.25.1
> >>
> >>
>