diff mbox series

[v1] RISC-V: Bugfix function target attribute pollution

Message ID 20240320060718.1765064-1-pan2.li@intel.com
State New
Headers show
Series [v1] RISC-V: Bugfix function target attribute pollution | expand

Commit Message

Li, Pan2 March 20, 2024, 6:07 a.m. UTC
From: Pan Li <pan2.li@intel.com>

This patch depends on below ICE fix.

https://gcc.gnu.org/pipermail/gcc-patches/2024-March/647915.html

The function target attribute should be on a per-function basis.
For example, we have 3 function as below:

void test_1 () {}

void __attribute__((target("arch=+v"))) test_2 () {}

void __attribute__((target("arch=+zfh"))) test_3 () {}

void test_4 () {}

The scope of the target attribute should not extend the function body.
Aka, test_3 cannot have the 'v' extension, as well as the test_4
cannot have both the 'v' and 'zfh' extension.

Unfortunately, for now the test_4 is able to leverage the 'v' and
the 'zfh' extension which is incorrect.  This patch would like to
fix the sticking attribute by introduce the commandline subset_list.
When parse_arch, we always clone from the cmdline_subset_list instead
of the current_subset_list.

Meanwhile, we correct the print information about arch like below.

.option arch, rv64i2p1_m2p0_a2p1_f2p2_d2p2_c2p0_zicsr2p0_zifencei2p0_zbb1p0

The riscv_declare_function_name hook is always after the hook
riscv_process_target_attr.  Thus, we introduce one hash_map to record
the 1:1 mapping from fndel to its' subset_list in advance.  And later
the riscv_declare_function_name is able to get the right information
about the arch.

Below test are passed for this patch
* The riscv fully regression test.

	PR target/114352

gcc/ChangeLog:

	* common/config/riscv/riscv-common.cc (struct riscv_func_target_info):
	New struct for func decl and target name.
	(struct riscv_func_target_hasher): New hasher for hash table mapping
	from the fn_decl to fn_target_name.
	(riscv_func_decl_hash): New func to compute the hash for fn_decl.
	(riscv_func_target_hasher::hash): New func to impl hash interface.
	(riscv_func_target_hasher::equal): New func to impl equal interface.
	(riscv_cmdline_subset_list): New static var for cmdline subset list.
	(riscv_func_target_table_lazy_init): New func to lazy init the func
	target hash table.
	(riscv_func_target_get): New func to get target name from hash table.
	(riscv_func_target_put): New func to put target name into hash table.
	(riscv_func_target_remove_and_destory): New func to remove target
	info from the hash table and destory it.
	(riscv_parse_arch_string): Set the static var cmdline_subset_list.
	* config/riscv/riscv-subset.h (riscv_cmdline_subset_list): New static
	var for cmdline subset list.
	(riscv_func_target_get): New func decl.
	(riscv_func_target_put): Ditto.
	(riscv_func_target_remove_and_destory): Ditto.
	* config/riscv/riscv-target-attr.cc (riscv_target_attr_parser::parse_arch):
	Take cmdline_subset_list instead of current_subset_list when clone.
	(riscv_process_target_attr): Record the func target info to hash table.
	(riscv_option_valid_attribute_p): Add new arg tree fndel.
	* config/riscv/riscv.cc (riscv_declare_function_name): Consume the
	func target info and print the arch message.

gcc/testsuite/ChangeLog:

	* gcc.target/riscv/rvv/base/pr114352-3.c: New test.

Signed-off-by: Pan Li <pan2.li@intel.com>
---
 gcc/common/config/riscv/riscv-common.cc       | 105 +++++++++++++++-
 gcc/config/riscv/riscv-subset.h               |   4 +
 gcc/config/riscv/riscv-target-attr.cc         |  18 ++-
 gcc/config/riscv/riscv.cc                     |   7 +-
 .../gcc.target/riscv/rvv/base/pr114352-3.c    | 113 ++++++++++++++++++
 5 files changed, 240 insertions(+), 7 deletions(-)
 create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/base/pr114352-3.c

Comments

Kito Cheng March 21, 2024, 12:33 p.m. UTC | #1
LGTM, thanks :)

On Wed, Mar 20, 2024 at 2:07 PM <pan2.li@intel.com> wrote:
>
> From: Pan Li <pan2.li@intel.com>
>
> This patch depends on below ICE fix.
>
> https://gcc.gnu.org/pipermail/gcc-patches/2024-March/647915.html
>
> The function target attribute should be on a per-function basis.
> For example, we have 3 function as below:
>
> void test_1 () {}
>
> void __attribute__((target("arch=+v"))) test_2 () {}
>
> void __attribute__((target("arch=+zfh"))) test_3 () {}
>
> void test_4 () {}
>
> The scope of the target attribute should not extend the function body.
> Aka, test_3 cannot have the 'v' extension, as well as the test_4
> cannot have both the 'v' and 'zfh' extension.
>
> Unfortunately, for now the test_4 is able to leverage the 'v' and
> the 'zfh' extension which is incorrect.  This patch would like to
> fix the sticking attribute by introduce the commandline subset_list.
> When parse_arch, we always clone from the cmdline_subset_list instead
> of the current_subset_list.
>
> Meanwhile, we correct the print information about arch like below.
>
> .option arch, rv64i2p1_m2p0_a2p1_f2p2_d2p2_c2p0_zicsr2p0_zifencei2p0_zbb1p0
>
> The riscv_declare_function_name hook is always after the hook
> riscv_process_target_attr.  Thus, we introduce one hash_map to record
> the 1:1 mapping from fndel to its' subset_list in advance.  And later
> the riscv_declare_function_name is able to get the right information
> about the arch.
>
> Below test are passed for this patch
> * The riscv fully regression test.
>
>         PR target/114352
>
> gcc/ChangeLog:
>
>         * common/config/riscv/riscv-common.cc (struct riscv_func_target_info):
>         New struct for func decl and target name.
>         (struct riscv_func_target_hasher): New hasher for hash table mapping
>         from the fn_decl to fn_target_name.
>         (riscv_func_decl_hash): New func to compute the hash for fn_decl.
>         (riscv_func_target_hasher::hash): New func to impl hash interface.
>         (riscv_func_target_hasher::equal): New func to impl equal interface.
>         (riscv_cmdline_subset_list): New static var for cmdline subset list.
>         (riscv_func_target_table_lazy_init): New func to lazy init the func
>         target hash table.
>         (riscv_func_target_get): New func to get target name from hash table.
>         (riscv_func_target_put): New func to put target name into hash table.
>         (riscv_func_target_remove_and_destory): New func to remove target
>         info from the hash table and destory it.
>         (riscv_parse_arch_string): Set the static var cmdline_subset_list.
>         * config/riscv/riscv-subset.h (riscv_cmdline_subset_list): New static
>         var for cmdline subset list.
>         (riscv_func_target_get): New func decl.
>         (riscv_func_target_put): Ditto.
>         (riscv_func_target_remove_and_destory): Ditto.
>         * config/riscv/riscv-target-attr.cc (riscv_target_attr_parser::parse_arch):
>         Take cmdline_subset_list instead of current_subset_list when clone.
>         (riscv_process_target_attr): Record the func target info to hash table.
>         (riscv_option_valid_attribute_p): Add new arg tree fndel.
>         * config/riscv/riscv.cc (riscv_declare_function_name): Consume the
>         func target info and print the arch message.
>
> gcc/testsuite/ChangeLog:
>
>         * gcc.target/riscv/rvv/base/pr114352-3.c: New test.
>
> Signed-off-by: Pan Li <pan2.li@intel.com>
> ---
>  gcc/common/config/riscv/riscv-common.cc       | 105 +++++++++++++++-
>  gcc/config/riscv/riscv-subset.h               |   4 +
>  gcc/config/riscv/riscv-target-attr.cc         |  18 ++-
>  gcc/config/riscv/riscv.cc                     |   7 +-
>  .../gcc.target/riscv/rvv/base/pr114352-3.c    | 113 ++++++++++++++++++
>  5 files changed, 240 insertions(+), 7 deletions(-)
>  create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/base/pr114352-3.c
>
> diff --git a/gcc/common/config/riscv/riscv-common.cc b/gcc/common/config/riscv/riscv-common.cc
> index d32bf147eca..76ec9bf846c 100644
> --- a/gcc/common/config/riscv/riscv-common.cc
> +++ b/gcc/common/config/riscv/riscv-common.cc
> @@ -425,11 +425,108 @@ bool riscv_subset_list::parse_failed = false;
>
>  static riscv_subset_list *current_subset_list = NULL;
>
> +static riscv_subset_list *cmdline_subset_list = NULL;
> +
> +struct riscv_func_target_info
> +{
> +  tree fn_decl;
> +  std::string fn_target_name;
> +
> +  riscv_func_target_info (const tree &decl, const std::string &target_name)
> +    : fn_decl (decl), fn_target_name (target_name)
> +  {
> +  }
> +};
> +
> +struct riscv_func_target_hasher : nofree_ptr_hash<struct riscv_func_target_info>
> +{
> +  typedef tree compare_type;
> +
> +  static hashval_t hash (value_type);
> +  static bool equal (value_type, const compare_type &);
> +};
> +
> +static hash_table<riscv_func_target_hasher> *func_target_table = NULL;
> +
> +static inline hashval_t riscv_func_decl_hash (tree fn_decl)
> +{
> +  inchash::hash h;
> +
> +  h.add_ptr (fn_decl);
> +
> +  return h.end ();
> +}
> +
> +inline hashval_t
> +riscv_func_target_hasher::hash (value_type value)
> +{
> +  return riscv_func_decl_hash (value->fn_decl);
> +}
> +
> +inline bool
> +riscv_func_target_hasher::equal (value_type value, const compare_type &key)
> +{
> +  return value->fn_decl == key;
> +}
> +
>  const riscv_subset_list *riscv_current_subset_list ()
>  {
>    return current_subset_list;
>  }
>
> +const riscv_subset_list * riscv_cmdline_subset_list ()
> +{
> +  return cmdline_subset_list;
> +}
> +
> +static inline void riscv_func_target_table_lazy_init ()
> +{
> +  if (func_target_table != NULL)
> +    return;
> +
> +  func_target_table = new hash_table<riscv_func_target_hasher> (1023);
> +}
> +
> +std::string * riscv_func_target_get (tree fn_decl)
> +{
> +  riscv_func_target_table_lazy_init ();
> +
> +  hashval_t hash = riscv_func_decl_hash (fn_decl);
> +  struct riscv_func_target_info *info
> +    = func_target_table->find_with_hash (fn_decl, hash);
> +
> +  return info == NULL ? NULL : &info->fn_target_name;
> +}
> +
> +void riscv_func_target_put (tree fn_decl, std::string fn_target_name)
> +{
> +  riscv_func_target_table_lazy_init ();
> +
> +  hashval_t hash = riscv_func_decl_hash (fn_decl);
> +  struct riscv_func_target_info **target_info_slot
> +    = func_target_table->find_slot_with_hash (fn_decl, hash, INSERT);
> +
> +  gcc_assert (!*target_info_slot);
> +
> +  struct riscv_func_target_info *info
> +    = new riscv_func_target_info (fn_decl, fn_target_name);
> +
> +  *target_info_slot = info;
> +}
> +
> +void riscv_func_target_remove_and_destory (tree fn_decl)
> +{
> +  hashval_t hash = riscv_func_decl_hash (fn_decl);
> +  struct riscv_func_target_info *info
> +    = func_target_table->find_with_hash (fn_decl, hash);
> +
> +  if (info)
> +    {
> +      func_target_table->remove_elt_with_hash (fn_decl, hash);
> +      delete info;
> +    }
> +}
> +
>  /* struct for recording multi-lib info.  */
>  struct riscv_multi_lib_info_t {
>    std::string path;
> @@ -1737,10 +1834,14 @@ riscv_parse_arch_string (const char *isa,
>         }
>      }
>
> -  if (current_subset_list)
> +  /* Avoid double delete if current_subset_list equals cmdline_subset_list.  */
> +  if (current_subset_list && current_subset_list != cmdline_subset_list)
>      delete current_subset_list;
>
> -  current_subset_list = subset_list;
> +  if (cmdline_subset_list)
> +    delete cmdline_subset_list;
> +
> +  current_subset_list = cmdline_subset_list = subset_list;
>  }
>
>  /* Return the riscv_cpu_info entry for CPU, NULL if not found.  */
> diff --git a/gcc/config/riscv/riscv-subset.h b/gcc/config/riscv/riscv-subset.h
> index ae849e2a302..c4f1eaf4e26 100644
> --- a/gcc/config/riscv/riscv-subset.h
> +++ b/gcc/config/riscv/riscv-subset.h
> @@ -108,6 +108,10 @@ public:
>  };
>
>  extern const riscv_subset_list *riscv_current_subset_list (void);
> +extern const riscv_subset_list *riscv_cmdline_subset_list (void);
> +extern std::string * riscv_func_target_get (tree);
> +extern void riscv_func_target_put (tree, std::string);
> +extern void riscv_func_target_remove_and_destory (tree);
>  extern void
>  riscv_set_arch_by_subset_list (riscv_subset_list *, struct gcc_options *);
>
> diff --git a/gcc/config/riscv/riscv-target-attr.cc b/gcc/config/riscv/riscv-target-attr.cc
> index 9dbb78f28cc..2476250eed6 100644
> --- a/gcc/config/riscv/riscv-target-attr.cc
> +++ b/gcc/config/riscv/riscv-target-attr.cc
> @@ -54,6 +54,10 @@ public:
>      m_loc = loc;
>    }
>
> +  riscv_subset_list* get_riscv_subset_list () {
> +    return m_subset_list;
> +  }
> +
>    void update_settings (struct gcc_options *opts) const;
>  private:
>    const char *m_raw_attr_str;
> @@ -109,7 +113,7 @@ riscv_target_attr_parser::parse_arch (const char *str)
>        char *str_to_check = buf.get ();
>        strcpy (str_to_check, str);
>        const char *token = strtok_r (str_to_check, ",", &str_to_check);
> -      m_subset_list = riscv_current_subset_list ()->clone ();
> +      m_subset_list = riscv_cmdline_subset_list ()->clone ();
>        m_subset_list->set_loc (m_loc);
>        while (token)
>         {
> @@ -295,7 +299,8 @@ num_occurences_in_str (char c, char *str)
>     and update the global target options space.  */
>
>  static bool
> -riscv_process_target_attr (tree args, location_t loc, struct gcc_options *opts)
> +riscv_process_target_attr (tree fndecl, tree args, location_t loc,
> +                          struct gcc_options *opts)
>  {
>    if (TREE_CODE (args) == TREE_LIST)
>      {
> @@ -304,7 +309,7 @@ riscv_process_target_attr (tree args, location_t loc, struct gcc_options *opts)
>           tree head = TREE_VALUE (args);
>           if (head)
>             {
> -             if (!riscv_process_target_attr (head, loc, opts))
> +             if (!riscv_process_target_attr (fndecl, head, loc, opts))
>                 return false;
>             }
>           args = TREE_CHAIN (args);
> @@ -357,6 +362,11 @@ riscv_process_target_attr (tree args, location_t loc, struct gcc_options *opts)
>    /* Apply settings from target attribute.  */
>    attr_parser.update_settings (opts);
>
> +  /* Add the string of the target attribute to the fndecl hash table.  */
> +  riscv_subset_list *subset_list = attr_parser.get_riscv_subset_list ();
> +  if (subset_list)
> +    riscv_func_target_put (fndecl, subset_list->to_string (true));
> +
>    return true;
>  }
>
> @@ -374,7 +384,7 @@ riscv_option_valid_attribute_p (tree fndecl, tree, tree args, int)
>    /* Save the current target options to restore at the end.  */
>    cl_target_option_save (&cur_target, &global_options, &global_options_set);
>
> -  ret = riscv_process_target_attr (args, loc, &global_options);
> +  ret = riscv_process_target_attr (fndecl, args, loc, &global_options);
>
>    if (ret)
>      {
> diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc
> index 89acb94af10..42bc3e6eeb3 100644
> --- a/gcc/config/riscv/riscv.cc
> +++ b/gcc/config/riscv/riscv.cc
> @@ -8673,8 +8673,13 @@ riscv_declare_function_name (FILE *stream, const char *name, tree fndecl)
>    if (DECL_FUNCTION_SPECIFIC_TARGET (fndecl))
>      {
>        fprintf (stream, "\t.option push\n");
> -      std::string isa = riscv_current_subset_list ()->to_string (true);
> +
> +      std::string *target_name = riscv_func_target_get (fndecl);
> +      std::string isa = target_name != NULL
> +       ? *target_name
> +       : riscv_cmdline_subset_list ()->to_string (true);
>        fprintf (stream, "\t.option arch, %s\n", isa.c_str ());
> +      riscv_func_target_remove_and_destory (fndecl);
>
>        struct cl_target_option *local_cl_target =
>         TREE_TARGET_OPTION (DECL_FUNCTION_SPECIFIC_TARGET (fndecl));
> diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pr114352-3.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pr114352-3.c
> new file mode 100644
> index 00000000000..e7af4223d6a
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pr114352-3.c
> @@ -0,0 +1,113 @@
> +/* { dg-do compile } */
> +/* { dg-options "-march=rv64gc -mabi=lp64 -O3 -fno-schedule-insns -fno-schedule-insns2" } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> +
> +/*
> +** test_1:
> +** sext\.w\s+[atx][0-9]+,\s*[atx][0-9]+
> +** ...
> +*/
> +void
> +test_1 (int *a, int *b, int *out, unsigned count)
> +{
> +  unsigned i;
> +
> +  count = count > 128 ? 128 : count;
> +
> +  for (i = 0; i < count; i++)
> +    out[i] = a[i] + b[i];
> +}
> +
> +/*
> +** test_2:
> +** ...
> +** vadd\.vv\s+v[0-9]+,\s*v[0-9]+,\s*v[0-9]+
> +** ...
> +*/
> +void
> +__attribute__((target("arch=+v")))
> +test_2 (int *a, int *b, int *out, unsigned count)
> +{
> +  unsigned i;
> +
> +  count = count > 128 ? 128 : count;
> +
> +  for (i = 0; i < count; i++)
> +    out[i] = a[i] + b[i];
> +}
> +
> +/*
> +** test_3:
> +** ...
> +** minu\s+[atx][0-9]+,\s*[atx][0-9]+,\s*[atx][0-9]+
> +** ...
> +*/
> +void
> +__attribute__((target("arch=+zbb")))
> +test_3 (int *a, int *b, int *out, unsigned count)
> +{
> +  unsigned i;
> +
> +  count = count > 128 ? 128 : count;
> +
> +  for (i = 0; i < count; i++)
> +    out[i] = a[i] + b[i];
> +}
> +
> +/*
> +** test_4:
> +** sext\.w\s+[atx][0-9]+,\s*[atx][0-9]+
> +** ...
> +*/
> +void
> +test_4 (int *a, int *b, int *out, unsigned count)
> +{
> +  unsigned i;
> +
> +  count = count > 128 ? 128 : count;
> +
> +  for (i = 0; i < count; i++)
> +    out[i] = a[i] + b[i];
> +}
> +
> +/*
> +** test_5:
> +** ...
> +** fadd\.h\s+fa[0-9]+,\s*fa[0-9]+,\s*fa[0-9]+
> +** ...
> +*/
> +void
> +__attribute__((target("arch=+zfh")))
> +test_5 (_Float16 *a, _Float16 *b, _Float16 *out, unsigned count)
> +{
> +  unsigned i;
> +
> +  count = count > 128 ? 128 : count;
> +
> +  for (i = 0; i < count; i++)
> +    out[i] = a[i] + b[i];
> +}
> +
> +/*
> +** test_6:
> +** ...
> +** call\s+__extendhfsf2
> +** ...
> +** call\s+__truncsfhf2
> +** ...
> +*/
> +void
> +test_6 (_Float16 *a, _Float16 *b, _Float16 *out, unsigned count)
> +{
> +  unsigned i;
> +
> +  count = count > 128 ? 128 : count;
> +
> +  for (i = 0; i < count; i++)
> +    out[i] = a[i] + b[i];
> +}
> +
> +/* { dg-final { scan-assembler ".attribute arch, \"rv64i2p1_m2p0_a2p1_f2p2_d2p2_c2p0_zicsr2p0_zifencei2p0\"" } } */
> +/* { dg-final { scan-assembler ".option arch, rv64i2p1_m2p0_a2p1_f2p2_d2p2_c2p0_v1p0_zicsr2p0_zifencei2p0_zve32f1p0_zve32x1p0_zve64d1p0_zve64f1p0_zve64x1p0_zvl128b1p0_zvl32b1p0_zvl64b1p0" } } */
> +/* { dg-final { scan-assembler ".option arch, rv64i2p1_m2p0_a2p1_f2p2_d2p2_c2p0_zicsr2p0_zifencei2p0_zbb1p0" } } */
> +/* { dg-final { scan-assembler ".option arch, rv64i2p1_m2p0_a2p1_f2p2_d2p2_c2p0_zicsr2p0_zifencei2p0_zfh1p0_zfhmin1p0" } } */
> --
> 2.34.1
>
Li, Pan2 March 21, 2024, 2:06 p.m. UTC | #2
Thanks Kito, will commit it after the ICE fix.

Pan

-----Original Message-----
From: Kito Cheng <kito.cheng@gmail.com> 
Sent: Thursday, March 21, 2024 8:33 PM
To: Li, Pan2 <pan2.li@intel.com>
Cc: gcc-patches@gcc.gnu.org; juzhe.zhong@rivai.ai; Wang, Yanzhang <yanzhang.wang@intel.com>
Subject: Re: [PATCH v1] RISC-V: Bugfix function target attribute pollution

LGTM, thanks :)

On Wed, Mar 20, 2024 at 2:07 PM <pan2.li@intel.com> wrote:
>
> From: Pan Li <pan2.li@intel.com>
>
> This patch depends on below ICE fix.
>
> https://gcc.gnu.org/pipermail/gcc-patches/2024-March/647915.html
>
> The function target attribute should be on a per-function basis.
> For example, we have 3 function as below:
>
> void test_1 () {}
>
> void __attribute__((target("arch=+v"))) test_2 () {}
>
> void __attribute__((target("arch=+zfh"))) test_3 () {}
>
> void test_4 () {}
>
> The scope of the target attribute should not extend the function body.
> Aka, test_3 cannot have the 'v' extension, as well as the test_4
> cannot have both the 'v' and 'zfh' extension.
>
> Unfortunately, for now the test_4 is able to leverage the 'v' and
> the 'zfh' extension which is incorrect.  This patch would like to
> fix the sticking attribute by introduce the commandline subset_list.
> When parse_arch, we always clone from the cmdline_subset_list instead
> of the current_subset_list.
>
> Meanwhile, we correct the print information about arch like below.
>
> .option arch, rv64i2p1_m2p0_a2p1_f2p2_d2p2_c2p0_zicsr2p0_zifencei2p0_zbb1p0
>
> The riscv_declare_function_name hook is always after the hook
> riscv_process_target_attr.  Thus, we introduce one hash_map to record
> the 1:1 mapping from fndel to its' subset_list in advance.  And later
> the riscv_declare_function_name is able to get the right information
> about the arch.
>
> Below test are passed for this patch
> * The riscv fully regression test.
>
>         PR target/114352
>
> gcc/ChangeLog:
>
>         * common/config/riscv/riscv-common.cc (struct riscv_func_target_info):
>         New struct for func decl and target name.
>         (struct riscv_func_target_hasher): New hasher for hash table mapping
>         from the fn_decl to fn_target_name.
>         (riscv_func_decl_hash): New func to compute the hash for fn_decl.
>         (riscv_func_target_hasher::hash): New func to impl hash interface.
>         (riscv_func_target_hasher::equal): New func to impl equal interface.
>         (riscv_cmdline_subset_list): New static var for cmdline subset list.
>         (riscv_func_target_table_lazy_init): New func to lazy init the func
>         target hash table.
>         (riscv_func_target_get): New func to get target name from hash table.
>         (riscv_func_target_put): New func to put target name into hash table.
>         (riscv_func_target_remove_and_destory): New func to remove target
>         info from the hash table and destory it.
>         (riscv_parse_arch_string): Set the static var cmdline_subset_list.
>         * config/riscv/riscv-subset.h (riscv_cmdline_subset_list): New static
>         var for cmdline subset list.
>         (riscv_func_target_get): New func decl.
>         (riscv_func_target_put): Ditto.
>         (riscv_func_target_remove_and_destory): Ditto.
>         * config/riscv/riscv-target-attr.cc (riscv_target_attr_parser::parse_arch):
>         Take cmdline_subset_list instead of current_subset_list when clone.
>         (riscv_process_target_attr): Record the func target info to hash table.
>         (riscv_option_valid_attribute_p): Add new arg tree fndel.
>         * config/riscv/riscv.cc (riscv_declare_function_name): Consume the
>         func target info and print the arch message.
>
> gcc/testsuite/ChangeLog:
>
>         * gcc.target/riscv/rvv/base/pr114352-3.c: New test.
>
> Signed-off-by: Pan Li <pan2.li@intel.com>
> ---
>  gcc/common/config/riscv/riscv-common.cc       | 105 +++++++++++++++-
>  gcc/config/riscv/riscv-subset.h               |   4 +
>  gcc/config/riscv/riscv-target-attr.cc         |  18 ++-
>  gcc/config/riscv/riscv.cc                     |   7 +-
>  .../gcc.target/riscv/rvv/base/pr114352-3.c    | 113 ++++++++++++++++++
>  5 files changed, 240 insertions(+), 7 deletions(-)
>  create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/base/pr114352-3.c
>
> diff --git a/gcc/common/config/riscv/riscv-common.cc b/gcc/common/config/riscv/riscv-common.cc
> index d32bf147eca..76ec9bf846c 100644
> --- a/gcc/common/config/riscv/riscv-common.cc
> +++ b/gcc/common/config/riscv/riscv-common.cc
> @@ -425,11 +425,108 @@ bool riscv_subset_list::parse_failed = false;
>
>  static riscv_subset_list *current_subset_list = NULL;
>
> +static riscv_subset_list *cmdline_subset_list = NULL;
> +
> +struct riscv_func_target_info
> +{
> +  tree fn_decl;
> +  std::string fn_target_name;
> +
> +  riscv_func_target_info (const tree &decl, const std::string &target_name)
> +    : fn_decl (decl), fn_target_name (target_name)
> +  {
> +  }
> +};
> +
> +struct riscv_func_target_hasher : nofree_ptr_hash<struct riscv_func_target_info>
> +{
> +  typedef tree compare_type;
> +
> +  static hashval_t hash (value_type);
> +  static bool equal (value_type, const compare_type &);
> +};
> +
> +static hash_table<riscv_func_target_hasher> *func_target_table = NULL;
> +
> +static inline hashval_t riscv_func_decl_hash (tree fn_decl)
> +{
> +  inchash::hash h;
> +
> +  h.add_ptr (fn_decl);
> +
> +  return h.end ();
> +}
> +
> +inline hashval_t
> +riscv_func_target_hasher::hash (value_type value)
> +{
> +  return riscv_func_decl_hash (value->fn_decl);
> +}
> +
> +inline bool
> +riscv_func_target_hasher::equal (value_type value, const compare_type &key)
> +{
> +  return value->fn_decl == key;
> +}
> +
>  const riscv_subset_list *riscv_current_subset_list ()
>  {
>    return current_subset_list;
>  }
>
> +const riscv_subset_list * riscv_cmdline_subset_list ()
> +{
> +  return cmdline_subset_list;
> +}
> +
> +static inline void riscv_func_target_table_lazy_init ()
> +{
> +  if (func_target_table != NULL)
> +    return;
> +
> +  func_target_table = new hash_table<riscv_func_target_hasher> (1023);
> +}
> +
> +std::string * riscv_func_target_get (tree fn_decl)
> +{
> +  riscv_func_target_table_lazy_init ();
> +
> +  hashval_t hash = riscv_func_decl_hash (fn_decl);
> +  struct riscv_func_target_info *info
> +    = func_target_table->find_with_hash (fn_decl, hash);
> +
> +  return info == NULL ? NULL : &info->fn_target_name;
> +}
> +
> +void riscv_func_target_put (tree fn_decl, std::string fn_target_name)
> +{
> +  riscv_func_target_table_lazy_init ();
> +
> +  hashval_t hash = riscv_func_decl_hash (fn_decl);
> +  struct riscv_func_target_info **target_info_slot
> +    = func_target_table->find_slot_with_hash (fn_decl, hash, INSERT);
> +
> +  gcc_assert (!*target_info_slot);
> +
> +  struct riscv_func_target_info *info
> +    = new riscv_func_target_info (fn_decl, fn_target_name);
> +
> +  *target_info_slot = info;
> +}
> +
> +void riscv_func_target_remove_and_destory (tree fn_decl)
> +{
> +  hashval_t hash = riscv_func_decl_hash (fn_decl);
> +  struct riscv_func_target_info *info
> +    = func_target_table->find_with_hash (fn_decl, hash);
> +
> +  if (info)
> +    {
> +      func_target_table->remove_elt_with_hash (fn_decl, hash);
> +      delete info;
> +    }
> +}
> +
>  /* struct for recording multi-lib info.  */
>  struct riscv_multi_lib_info_t {
>    std::string path;
> @@ -1737,10 +1834,14 @@ riscv_parse_arch_string (const char *isa,
>         }
>      }
>
> -  if (current_subset_list)
> +  /* Avoid double delete if current_subset_list equals cmdline_subset_list.  */
> +  if (current_subset_list && current_subset_list != cmdline_subset_list)
>      delete current_subset_list;
>
> -  current_subset_list = subset_list;
> +  if (cmdline_subset_list)
> +    delete cmdline_subset_list;
> +
> +  current_subset_list = cmdline_subset_list = subset_list;
>  }
>
>  /* Return the riscv_cpu_info entry for CPU, NULL if not found.  */
> diff --git a/gcc/config/riscv/riscv-subset.h b/gcc/config/riscv/riscv-subset.h
> index ae849e2a302..c4f1eaf4e26 100644
> --- a/gcc/config/riscv/riscv-subset.h
> +++ b/gcc/config/riscv/riscv-subset.h
> @@ -108,6 +108,10 @@ public:
>  };
>
>  extern const riscv_subset_list *riscv_current_subset_list (void);
> +extern const riscv_subset_list *riscv_cmdline_subset_list (void);
> +extern std::string * riscv_func_target_get (tree);
> +extern void riscv_func_target_put (tree, std::string);
> +extern void riscv_func_target_remove_and_destory (tree);
>  extern void
>  riscv_set_arch_by_subset_list (riscv_subset_list *, struct gcc_options *);
>
> diff --git a/gcc/config/riscv/riscv-target-attr.cc b/gcc/config/riscv/riscv-target-attr.cc
> index 9dbb78f28cc..2476250eed6 100644
> --- a/gcc/config/riscv/riscv-target-attr.cc
> +++ b/gcc/config/riscv/riscv-target-attr.cc
> @@ -54,6 +54,10 @@ public:
>      m_loc = loc;
>    }
>
> +  riscv_subset_list* get_riscv_subset_list () {
> +    return m_subset_list;
> +  }
> +
>    void update_settings (struct gcc_options *opts) const;
>  private:
>    const char *m_raw_attr_str;
> @@ -109,7 +113,7 @@ riscv_target_attr_parser::parse_arch (const char *str)
>        char *str_to_check = buf.get ();
>        strcpy (str_to_check, str);
>        const char *token = strtok_r (str_to_check, ",", &str_to_check);
> -      m_subset_list = riscv_current_subset_list ()->clone ();
> +      m_subset_list = riscv_cmdline_subset_list ()->clone ();
>        m_subset_list->set_loc (m_loc);
>        while (token)
>         {
> @@ -295,7 +299,8 @@ num_occurences_in_str (char c, char *str)
>     and update the global target options space.  */
>
>  static bool
> -riscv_process_target_attr (tree args, location_t loc, struct gcc_options *opts)
> +riscv_process_target_attr (tree fndecl, tree args, location_t loc,
> +                          struct gcc_options *opts)
>  {
>    if (TREE_CODE (args) == TREE_LIST)
>      {
> @@ -304,7 +309,7 @@ riscv_process_target_attr (tree args, location_t loc, struct gcc_options *opts)
>           tree head = TREE_VALUE (args);
>           if (head)
>             {
> -             if (!riscv_process_target_attr (head, loc, opts))
> +             if (!riscv_process_target_attr (fndecl, head, loc, opts))
>                 return false;
>             }
>           args = TREE_CHAIN (args);
> @@ -357,6 +362,11 @@ riscv_process_target_attr (tree args, location_t loc, struct gcc_options *opts)
>    /* Apply settings from target attribute.  */
>    attr_parser.update_settings (opts);
>
> +  /* Add the string of the target attribute to the fndecl hash table.  */
> +  riscv_subset_list *subset_list = attr_parser.get_riscv_subset_list ();
> +  if (subset_list)
> +    riscv_func_target_put (fndecl, subset_list->to_string (true));
> +
>    return true;
>  }
>
> @@ -374,7 +384,7 @@ riscv_option_valid_attribute_p (tree fndecl, tree, tree args, int)
>    /* Save the current target options to restore at the end.  */
>    cl_target_option_save (&cur_target, &global_options, &global_options_set);
>
> -  ret = riscv_process_target_attr (args, loc, &global_options);
> +  ret = riscv_process_target_attr (fndecl, args, loc, &global_options);
>
>    if (ret)
>      {
> diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc
> index 89acb94af10..42bc3e6eeb3 100644
> --- a/gcc/config/riscv/riscv.cc
> +++ b/gcc/config/riscv/riscv.cc
> @@ -8673,8 +8673,13 @@ riscv_declare_function_name (FILE *stream, const char *name, tree fndecl)
>    if (DECL_FUNCTION_SPECIFIC_TARGET (fndecl))
>      {
>        fprintf (stream, "\t.option push\n");
> -      std::string isa = riscv_current_subset_list ()->to_string (true);
> +
> +      std::string *target_name = riscv_func_target_get (fndecl);
> +      std::string isa = target_name != NULL
> +       ? *target_name
> +       : riscv_cmdline_subset_list ()->to_string (true);
>        fprintf (stream, "\t.option arch, %s\n", isa.c_str ());
> +      riscv_func_target_remove_and_destory (fndecl);
>
>        struct cl_target_option *local_cl_target =
>         TREE_TARGET_OPTION (DECL_FUNCTION_SPECIFIC_TARGET (fndecl));
> diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pr114352-3.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pr114352-3.c
> new file mode 100644
> index 00000000000..e7af4223d6a
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pr114352-3.c
> @@ -0,0 +1,113 @@
> +/* { dg-do compile } */
> +/* { dg-options "-march=rv64gc -mabi=lp64 -O3 -fno-schedule-insns -fno-schedule-insns2" } */
> +/* { dg-final { check-function-bodies "**" "" } } */
> +
> +/*
> +** test_1:
> +** sext\.w\s+[atx][0-9]+,\s*[atx][0-9]+
> +** ...
> +*/
> +void
> +test_1 (int *a, int *b, int *out, unsigned count)
> +{
> +  unsigned i;
> +
> +  count = count > 128 ? 128 : count;
> +
> +  for (i = 0; i < count; i++)
> +    out[i] = a[i] + b[i];
> +}
> +
> +/*
> +** test_2:
> +** ...
> +** vadd\.vv\s+v[0-9]+,\s*v[0-9]+,\s*v[0-9]+
> +** ...
> +*/
> +void
> +__attribute__((target("arch=+v")))
> +test_2 (int *a, int *b, int *out, unsigned count)
> +{
> +  unsigned i;
> +
> +  count = count > 128 ? 128 : count;
> +
> +  for (i = 0; i < count; i++)
> +    out[i] = a[i] + b[i];
> +}
> +
> +/*
> +** test_3:
> +** ...
> +** minu\s+[atx][0-9]+,\s*[atx][0-9]+,\s*[atx][0-9]+
> +** ...
> +*/
> +void
> +__attribute__((target("arch=+zbb")))
> +test_3 (int *a, int *b, int *out, unsigned count)
> +{
> +  unsigned i;
> +
> +  count = count > 128 ? 128 : count;
> +
> +  for (i = 0; i < count; i++)
> +    out[i] = a[i] + b[i];
> +}
> +
> +/*
> +** test_4:
> +** sext\.w\s+[atx][0-9]+,\s*[atx][0-9]+
> +** ...
> +*/
> +void
> +test_4 (int *a, int *b, int *out, unsigned count)
> +{
> +  unsigned i;
> +
> +  count = count > 128 ? 128 : count;
> +
> +  for (i = 0; i < count; i++)
> +    out[i] = a[i] + b[i];
> +}
> +
> +/*
> +** test_5:
> +** ...
> +** fadd\.h\s+fa[0-9]+,\s*fa[0-9]+,\s*fa[0-9]+
> +** ...
> +*/
> +void
> +__attribute__((target("arch=+zfh")))
> +test_5 (_Float16 *a, _Float16 *b, _Float16 *out, unsigned count)
> +{
> +  unsigned i;
> +
> +  count = count > 128 ? 128 : count;
> +
> +  for (i = 0; i < count; i++)
> +    out[i] = a[i] + b[i];
> +}
> +
> +/*
> +** test_6:
> +** ...
> +** call\s+__extendhfsf2
> +** ...
> +** call\s+__truncsfhf2
> +** ...
> +*/
> +void
> +test_6 (_Float16 *a, _Float16 *b, _Float16 *out, unsigned count)
> +{
> +  unsigned i;
> +
> +  count = count > 128 ? 128 : count;
> +
> +  for (i = 0; i < count; i++)
> +    out[i] = a[i] + b[i];
> +}
> +
> +/* { dg-final { scan-assembler ".attribute arch, \"rv64i2p1_m2p0_a2p1_f2p2_d2p2_c2p0_zicsr2p0_zifencei2p0\"" } } */
> +/* { dg-final { scan-assembler ".option arch, rv64i2p1_m2p0_a2p1_f2p2_d2p2_c2p0_v1p0_zicsr2p0_zifencei2p0_zve32f1p0_zve32x1p0_zve64d1p0_zve64f1p0_zve64x1p0_zvl128b1p0_zvl32b1p0_zvl64b1p0" } } */
> +/* { dg-final { scan-assembler ".option arch, rv64i2p1_m2p0_a2p1_f2p2_d2p2_c2p0_zicsr2p0_zifencei2p0_zbb1p0" } } */
> +/* { dg-final { scan-assembler ".option arch, rv64i2p1_m2p0_a2p1_f2p2_d2p2_c2p0_zicsr2p0_zifencei2p0_zfh1p0_zfhmin1p0" } } */
> --
> 2.34.1
>
diff mbox series

Patch

diff --git a/gcc/common/config/riscv/riscv-common.cc b/gcc/common/config/riscv/riscv-common.cc
index d32bf147eca..76ec9bf846c 100644
--- a/gcc/common/config/riscv/riscv-common.cc
+++ b/gcc/common/config/riscv/riscv-common.cc
@@ -425,11 +425,108 @@  bool riscv_subset_list::parse_failed = false;
 
 static riscv_subset_list *current_subset_list = NULL;
 
+static riscv_subset_list *cmdline_subset_list = NULL;
+
+struct riscv_func_target_info
+{
+  tree fn_decl;
+  std::string fn_target_name;
+
+  riscv_func_target_info (const tree &decl, const std::string &target_name)
+    : fn_decl (decl), fn_target_name (target_name)
+  {
+  }
+};
+
+struct riscv_func_target_hasher : nofree_ptr_hash<struct riscv_func_target_info>
+{
+  typedef tree compare_type;
+
+  static hashval_t hash (value_type);
+  static bool equal (value_type, const compare_type &);
+};
+
+static hash_table<riscv_func_target_hasher> *func_target_table = NULL;
+
+static inline hashval_t riscv_func_decl_hash (tree fn_decl)
+{
+  inchash::hash h;
+
+  h.add_ptr (fn_decl);
+
+  return h.end ();
+}
+
+inline hashval_t
+riscv_func_target_hasher::hash (value_type value)
+{
+  return riscv_func_decl_hash (value->fn_decl);
+}
+
+inline bool
+riscv_func_target_hasher::equal (value_type value, const compare_type &key)
+{
+  return value->fn_decl == key;
+}
+
 const riscv_subset_list *riscv_current_subset_list ()
 {
   return current_subset_list;
 }
 
+const riscv_subset_list * riscv_cmdline_subset_list ()
+{
+  return cmdline_subset_list;
+}
+
+static inline void riscv_func_target_table_lazy_init ()
+{
+  if (func_target_table != NULL)
+    return;
+
+  func_target_table = new hash_table<riscv_func_target_hasher> (1023);
+}
+
+std::string * riscv_func_target_get (tree fn_decl)
+{
+  riscv_func_target_table_lazy_init ();
+
+  hashval_t hash = riscv_func_decl_hash (fn_decl);
+  struct riscv_func_target_info *info
+    = func_target_table->find_with_hash (fn_decl, hash);
+
+  return info == NULL ? NULL : &info->fn_target_name;
+}
+
+void riscv_func_target_put (tree fn_decl, std::string fn_target_name)
+{
+  riscv_func_target_table_lazy_init ();
+
+  hashval_t hash = riscv_func_decl_hash (fn_decl);
+  struct riscv_func_target_info **target_info_slot
+    = func_target_table->find_slot_with_hash (fn_decl, hash, INSERT);
+
+  gcc_assert (!*target_info_slot);
+
+  struct riscv_func_target_info *info
+    = new riscv_func_target_info (fn_decl, fn_target_name);
+
+  *target_info_slot = info;
+}
+
+void riscv_func_target_remove_and_destory (tree fn_decl)
+{
+  hashval_t hash = riscv_func_decl_hash (fn_decl);
+  struct riscv_func_target_info *info
+    = func_target_table->find_with_hash (fn_decl, hash);
+
+  if (info)
+    {
+      func_target_table->remove_elt_with_hash (fn_decl, hash);
+      delete info;
+    }
+}
+
 /* struct for recording multi-lib info.  */
 struct riscv_multi_lib_info_t {
   std::string path;
@@ -1737,10 +1834,14 @@  riscv_parse_arch_string (const char *isa,
 	}
     }
 
-  if (current_subset_list)
+  /* Avoid double delete if current_subset_list equals cmdline_subset_list.  */
+  if (current_subset_list && current_subset_list != cmdline_subset_list)
     delete current_subset_list;
 
-  current_subset_list = subset_list;
+  if (cmdline_subset_list)
+    delete cmdline_subset_list;
+
+  current_subset_list = cmdline_subset_list = subset_list;
 }
 
 /* Return the riscv_cpu_info entry for CPU, NULL if not found.  */
diff --git a/gcc/config/riscv/riscv-subset.h b/gcc/config/riscv/riscv-subset.h
index ae849e2a302..c4f1eaf4e26 100644
--- a/gcc/config/riscv/riscv-subset.h
+++ b/gcc/config/riscv/riscv-subset.h
@@ -108,6 +108,10 @@  public:
 };
 
 extern const riscv_subset_list *riscv_current_subset_list (void);
+extern const riscv_subset_list *riscv_cmdline_subset_list (void);
+extern std::string * riscv_func_target_get (tree);
+extern void riscv_func_target_put (tree, std::string);
+extern void riscv_func_target_remove_and_destory (tree);
 extern void
 riscv_set_arch_by_subset_list (riscv_subset_list *, struct gcc_options *);
 
diff --git a/gcc/config/riscv/riscv-target-attr.cc b/gcc/config/riscv/riscv-target-attr.cc
index 9dbb78f28cc..2476250eed6 100644
--- a/gcc/config/riscv/riscv-target-attr.cc
+++ b/gcc/config/riscv/riscv-target-attr.cc
@@ -54,6 +54,10 @@  public:
     m_loc = loc;
   }
 
+  riscv_subset_list* get_riscv_subset_list () {
+    return m_subset_list;
+  }
+
   void update_settings (struct gcc_options *opts) const;
 private:
   const char *m_raw_attr_str;
@@ -109,7 +113,7 @@  riscv_target_attr_parser::parse_arch (const char *str)
       char *str_to_check = buf.get ();
       strcpy (str_to_check, str);
       const char *token = strtok_r (str_to_check, ",", &str_to_check);
-      m_subset_list = riscv_current_subset_list ()->clone ();
+      m_subset_list = riscv_cmdline_subset_list ()->clone ();
       m_subset_list->set_loc (m_loc);
       while (token)
 	{
@@ -295,7 +299,8 @@  num_occurences_in_str (char c, char *str)
    and update the global target options space.  */
 
 static bool
-riscv_process_target_attr (tree args, location_t loc, struct gcc_options *opts)
+riscv_process_target_attr (tree fndecl, tree args, location_t loc,
+			   struct gcc_options *opts)
 {
   if (TREE_CODE (args) == TREE_LIST)
     {
@@ -304,7 +309,7 @@  riscv_process_target_attr (tree args, location_t loc, struct gcc_options *opts)
 	  tree head = TREE_VALUE (args);
 	  if (head)
 	    {
-	      if (!riscv_process_target_attr (head, loc, opts))
+	      if (!riscv_process_target_attr (fndecl, head, loc, opts))
 		return false;
 	    }
 	  args = TREE_CHAIN (args);
@@ -357,6 +362,11 @@  riscv_process_target_attr (tree args, location_t loc, struct gcc_options *opts)
   /* Apply settings from target attribute.  */
   attr_parser.update_settings (opts);
 
+  /* Add the string of the target attribute to the fndecl hash table.  */
+  riscv_subset_list *subset_list = attr_parser.get_riscv_subset_list ();
+  if (subset_list)
+    riscv_func_target_put (fndecl, subset_list->to_string (true));
+
   return true;
 }
 
@@ -374,7 +384,7 @@  riscv_option_valid_attribute_p (tree fndecl, tree, tree args, int)
   /* Save the current target options to restore at the end.  */
   cl_target_option_save (&cur_target, &global_options, &global_options_set);
 
-  ret = riscv_process_target_attr (args, loc, &global_options);
+  ret = riscv_process_target_attr (fndecl, args, loc, &global_options);
 
   if (ret)
     {
diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc
index 89acb94af10..42bc3e6eeb3 100644
--- a/gcc/config/riscv/riscv.cc
+++ b/gcc/config/riscv/riscv.cc
@@ -8673,8 +8673,13 @@  riscv_declare_function_name (FILE *stream, const char *name, tree fndecl)
   if (DECL_FUNCTION_SPECIFIC_TARGET (fndecl))
     {
       fprintf (stream, "\t.option push\n");
-      std::string isa = riscv_current_subset_list ()->to_string (true);
+
+      std::string *target_name = riscv_func_target_get (fndecl);
+      std::string isa = target_name != NULL
+	? *target_name
+	: riscv_cmdline_subset_list ()->to_string (true);
       fprintf (stream, "\t.option arch, %s\n", isa.c_str ());
+      riscv_func_target_remove_and_destory (fndecl);
 
       struct cl_target_option *local_cl_target =
 	TREE_TARGET_OPTION (DECL_FUNCTION_SPECIFIC_TARGET (fndecl));
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pr114352-3.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pr114352-3.c
new file mode 100644
index 00000000000..e7af4223d6a
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pr114352-3.c
@@ -0,0 +1,113 @@ 
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gc -mabi=lp64 -O3 -fno-schedule-insns -fno-schedule-insns2" } */
+/* { dg-final { check-function-bodies "**" "" } } */
+
+/*
+** test_1:
+** sext\.w\s+[atx][0-9]+,\s*[atx][0-9]+
+** ...
+*/
+void
+test_1 (int *a, int *b, int *out, unsigned count)
+{
+  unsigned i;
+
+  count = count > 128 ? 128 : count;
+
+  for (i = 0; i < count; i++)
+    out[i] = a[i] + b[i];
+}
+
+/*
+** test_2:
+** ...
+** vadd\.vv\s+v[0-9]+,\s*v[0-9]+,\s*v[0-9]+
+** ...
+*/
+void
+__attribute__((target("arch=+v")))
+test_2 (int *a, int *b, int *out, unsigned count)
+{
+  unsigned i;
+
+  count = count > 128 ? 128 : count;
+
+  for (i = 0; i < count; i++)
+    out[i] = a[i] + b[i];
+}
+
+/*
+** test_3:
+** ...
+** minu\s+[atx][0-9]+,\s*[atx][0-9]+,\s*[atx][0-9]+
+** ...
+*/
+void
+__attribute__((target("arch=+zbb")))
+test_3 (int *a, int *b, int *out, unsigned count)
+{
+  unsigned i;
+
+  count = count > 128 ? 128 : count;
+
+  for (i = 0; i < count; i++)
+    out[i] = a[i] + b[i];
+}
+
+/*
+** test_4:
+** sext\.w\s+[atx][0-9]+,\s*[atx][0-9]+
+** ...
+*/
+void
+test_4 (int *a, int *b, int *out, unsigned count)
+{
+  unsigned i;
+
+  count = count > 128 ? 128 : count;
+
+  for (i = 0; i < count; i++)
+    out[i] = a[i] + b[i];
+}
+
+/*
+** test_5:
+** ...
+** fadd\.h\s+fa[0-9]+,\s*fa[0-9]+,\s*fa[0-9]+
+** ...
+*/
+void
+__attribute__((target("arch=+zfh")))
+test_5 (_Float16 *a, _Float16 *b, _Float16 *out, unsigned count)
+{
+  unsigned i;
+
+  count = count > 128 ? 128 : count;
+
+  for (i = 0; i < count; i++)
+    out[i] = a[i] + b[i];
+}
+
+/*
+** test_6:
+** ...
+** call\s+__extendhfsf2
+** ...
+** call\s+__truncsfhf2
+** ...
+*/
+void
+test_6 (_Float16 *a, _Float16 *b, _Float16 *out, unsigned count)
+{
+  unsigned i;
+
+  count = count > 128 ? 128 : count;
+
+  for (i = 0; i < count; i++)
+    out[i] = a[i] + b[i];
+}
+
+/* { dg-final { scan-assembler ".attribute arch, \"rv64i2p1_m2p0_a2p1_f2p2_d2p2_c2p0_zicsr2p0_zifencei2p0\"" } } */
+/* { dg-final { scan-assembler ".option arch, rv64i2p1_m2p0_a2p1_f2p2_d2p2_c2p0_v1p0_zicsr2p0_zifencei2p0_zve32f1p0_zve32x1p0_zve64d1p0_zve64f1p0_zve64x1p0_zvl128b1p0_zvl32b1p0_zvl64b1p0" } } */
+/* { dg-final { scan-assembler ".option arch, rv64i2p1_m2p0_a2p1_f2p2_d2p2_c2p0_zicsr2p0_zifencei2p0_zbb1p0" } } */
+/* { dg-final { scan-assembler ".option arch, rv64i2p1_m2p0_a2p1_f2p2_d2p2_c2p0_zicsr2p0_zifencei2p0_zfh1p0_zfhmin1p0" } } */