vec.h, v2: Make some ops work with non-trivially copy constructible and/or destructible types

  On Wed, Sep 27, 2023 at 07:17:22AM +0000, Richard Biener wrote:
> OK I guess.  Can you summarize the limitations for non-POD types
> in the big comment at the start of vec.h?

Still haven't done that, but will do after we flesh out the details
below.

> (can we put in static_asserts
> in the places that obviously do not work?)

I've tried to do this though, I think the static_asserts will allow
making sure we only use what is supportable and will serve better than
any kind of comment.

But, I've run into quite a few triggered assertion failures with that, and
the question is what we want to do with them.  Applying

	Jakub
  

On Wed, 27 Sep 2023, Jakub Jelinek wrote:

> On Wed, Sep 27, 2023 at 07:17:22AM +0000, Richard Biener wrote:
> > OK I guess.  Can you summarize the limitations for non-POD types
> > in the big comment at the start of vec.h?
> 
> Still haven't done that, but will do after we flesh out the details
> below.
> 
> > (can we put in static_asserts
> > in the places that obviously do not work?)
> 
> I've tried to do this though, I think the static_asserts will allow
> making sure we only use what is supportable and will serve better than
> any kind of comment.
> 
> But, I've run into quite a few triggered assertion failures with that, and
> the question is what we want to do with them.  Applying
> --- gcc/vec.h.jj	2023-09-27 12:11:56.000000000 +0200
> +++ gcc/vec.h	2023-09-27 12:39:50.971613964 +0200
> @@ -1160,7 +1160,7 @@ template<typename T, typename A>
>  inline void
>  vec<T, A, vl_embed>::qsort (int (*cmp) (const void *, const void *))
>  {
> -  static_assert (std::is_trivially_copyable <T>::value, "");
> +//  static_assert (std::is_trivially_copyable <T>::value, "");
>    if (length () > 1)
>      gcc_qsort (address (), length (), sizeof (T), cmp);
>  }
> @@ -1359,7 +1359,7 @@ inline void
>  vec<T, A, vl_embed>::quick_grow (unsigned len)
>  {
>    gcc_checking_assert (length () <= len && len <= m_vecpfx.m_alloc);
> -  static_assert (std::is_trivially_default_constructible <T>::value, "");
> +//  static_assert (std::is_trivially_default_constructible <T>::value, "");
>    m_vecpfx.m_num = len;
>  }
>  
> incremental patch makes stuff work (at lest make in gcc subdir for
> x86_64-linux with --enable-languages=c,c++,fortran,lto succeeds), so it is just
> those 2 asserts.
> Following is full list of failures and discusses details.
> 
> dbgcnt.cc:132  limits[index].qsort (cmp_tuples);
> T = std::pair<unsigned int, unsigned int>
> where std::pair is not trivially copyable.  Our qsort implementation uses
> memcpys/memmoves to reshuffle the array elements (as it isn't inlined and
> so can't use std::swap and the like), so I think we need the types trivially
> copyable (or at least trivially copy assignable from the same type or
> something similar if we close all eyes).  Is there some std::pair
> alternative which is trivially copyable, or do we need to define structures
> to achieve that?  Or just close all eyes and allow qsort/sort/stablesort
> either on trivially copyable types, or on std::pair where both template
> parameters are trivially copyable?
> 
> genrecog.cc:3466      candidates.qsort (subroutine_candidate_cmp);
> T = std::pair<transition*, state_size>
> ditto
> 
> dwarf2asm.cc:1061  temp.qsort (compare_strings);
> T = std::pair<const char*, tree_node*>
> ditto
> 
> tree-ssa-dce.cc:1776      args.qsort (sort_phi_args);
> T = std::pair<edge_def*, unsigned int>
> ditto
> 
> tree-ssa-loop-manip.cc:376  names.qsort (loop_name_cmp);
> T = std::pair<int, int>
> ditto
> 
> omp-oacc-neuter-broadcast.cc:1730  priority.qsort (sort_size_descending);
> T = std::pair<int, tree_node*>
> ditto
> 
> ipa-icf.cc:3087  to_split.qsort (sort_congruence_split);
> T = std::pair<ipa_icf::congruence_class*, bitmap_head*>
> ditto
> 
> ipa-icf.cc:3360  classes.qsort (sort_congruence_class_groups_by_decl_uid);
> T = std::pair<ipa_icf::congruence_class_group*, int>
> ditto
> 
> tree-vect-slp.cc:6991  li_scalar_costs.qsort (li_cost_vec_cmp);
> T = std::pair<unsigned int, stmt_info_for_cost*>
> ditto
> 
> tree-vect-slp.cc:7249              lane_defs.qsort (vld_cmp);
> T = std::pair<unsigned int, tree_node*>
> ditto

make an exception as discussed on IRC

> cfganal.cc:471  control_dependence_map.quick_grow (last_basic_block_for_fn (cfun));
> T = bitmap_head
> This is a different case, bitmap_head has a non-trivial default constructor
> that essentially fills it with zeros, and for quick_grow we have
> quick_grow_cleared which default constructs elements, but quick_grow leaves
> them unconstructed/uninitialized, which is why I wanted to add an assert
> there, e.g. quick_grow on the wide_int/widest_int WIP would break stuff
> terribly.  In the cfganal.cc case, we call bitmap_initialize on it after
> growing it, which overwrites all elements and doesn't depend on values of
> any, so grow_cleared actually is invalid from strict C++ POV, but not
> in reality.  Do we still want the assert in and slow cfganal.cc slightly
> by using quick_grow_cleared?  Or another option would be to make
> quick_grow work transparently the same as quick_grow_cleared for
> non-trivially default constructible types.  Though, I think it might be
> better if people explicitly see that the operation is more expensive.

I've added the CTOR for checking purposes, I suppose we can add
a bitmap_head_pod omitting it?

> tree-ssa-loop-im.cc:2592                      first_edge_seq.safe_grow (fes_length
> 			                                                + extra_refs.length ());
> T = seq_entry
> Here, seq_entry is:
> struct seq_entry
> {
>   seq_entry () {}
>   seq_entry (unsigned f, sm_kind k, tree fr = NULL)
>     : first (f), second (k), from (fr) {}
>   unsigned first;
>   sm_kind second;
>   tree from;
> };
> Wonder if making seq_entry () = default;
> wouldn't cure this.

Possibly?  It's also a pattern that might go into vec:: directly,
it's a splice_at (aka insert other vector at position).  Using
safe_grow_cleared would be OK here.

> 
> tree-ssa-loop-im.cc:3499  memory_accesses.refs_loaded_in_loop.quick_grow (number_of_loops (cfun));
> T = bitmap_head
> like the cfganal.cc case.
> 
> tree-ssa-live.cc:1364  active.quick_grow (last_basic_block_for_fn (fn));
> T = bitmap_head
> ditto
> 
> rtl-ssa/blocks.cc:60  bb_phis.quick_grow (num_bb_indices);
> T = rtl_ssa::function_info::bb_phi_info
> This structure contains bitmap_head, so again isn't default constructible.
> 
> rtl-ssa/blocks.cc:617  frontiers.safe_grow (num_bb_indices);
> T = bitmap_head
> see above
> 
> tree-vect-data-refs.cc:5369          sel.quick_grow (nelt);
> T = poly_int<1, long int>
> Should this use poly_int_pod<1, long int> instead?  Or quick_grow_cleared?

That's vec_perm_builder in the end, maybe yes, poly_int_pod would work
there.

> tree-vect-patterns.cc:2947  unprom.quick_grow (nops);
> T = vect_unpromoted_value
> Go for quick_grow_cleared?  Something else?

The CTOR zero-initializes everything, so maybe it can go.  In theory
.set_op could also be changed to .push_op ...

> tree-ssa-sccvn.cc:5562                accesses.quick_grow (accesses.length () + 1);
> T = ao_ref
> This isn't trivially copyable because of 3 poly_int64 members I believe. 
> Shall those be poly_int64_pod?

I guess so.  I fail to know the difference between both.

> 
> tree-vect-stmts.cc:2811      sel.quick_grow (count);
> T = poly_int<1, long int>
> see above
> 
> ipa-fnsummary.cc:683                            avals->m_known_value_ranges.safe_grow (count, true);
> T = Value_Range
> Go for cleared?

It does

                            avals->m_known_value_ranges.safe_grow (count, 
true);
                            for (int i = 0; i < count; ++i)
                              new (&avals->m_known_value_ranges[i])
                                Value_Range ();

if that's then semantically the same yes.

Richard.

> tree-vect-slp.cc:8344  mask.quick_grow (count);
> T = poly_int<1, long int>
> see above
> 
> tree-vect-loop.cc:9042        sel.quick_grow (6);
> T = poly_int<1, long int>
> see above
> 
> config/i386/sse.md:16172          sel.quick_grow (8);
> T = poly_int<1, long int>
> see above
>
> 2023-09-27  Jakub Jelinek  <jakub@redhat.com>
> 
> 	* vec.h (vec_destruct): New function template.
> 	(release): Use it for non-trivially destructible T.
> 	(truncate): Likewise.
> 	(quick_push): Perform a placement new into slot
> 	instead of assignment.
> 	(pop): For non-trivially destructible T return void
> 	rather than T & and destruct the popped element.
> 	(quick_insert, ordered_remove): Note that they aren't suitable
> 	for non-trivially copyable types.  Add static_asserts for that.
> 	(block_remove, qsort, sort, stablesort): Assert T is trivially
> 	copyable.
> 	(quick_grow): Assert T is trivially default constructible.
> 	(quick_grow_cleared): Don't call quick_grow, instead inline it
> 	by hand except for the new static_assert.
> 	(gt_ggc_mx): Assert T is trivially destructable.
> 	(auto_vec::operator=): Formatting fixes.
> 	(auto_vec::auto_vec): Likewise.
> 	(vec_safe_grow_cleared): Don't call vec_safe_grow, instead inline
> 	it manually and call quick_grow_cleared method rather than quick_grow.
> 	(safe_grow_cleared): Likewise.
> 	* edit-context.cc (class line_event): Move definition earlier.
> 
> --- gcc/vec.h.jj	2023-09-27 10:38:50.635845540 +0200
> +++ gcc/vec.h	2023-09-27 12:11:56.665586490 +0200
> @@ -185,6 +185,16 @@ extern void dump_vec_loc_statistics (voi
>  /* Hashtable mapping vec addresses to descriptors.  */
>  extern htab_t vec_mem_usage_hash;
>  
> +/* Destruct N elements in DST.  */
> +
> +template <typename T>
> +inline void
> +vec_destruct (T *dst, unsigned n)
> +{
> +  for ( ; n; ++dst, --n)
> +    dst->~T ();
> +}
> +
>  /* Control data for vectors.  This contains the number of allocated
>     and used slots inside a vector.  */
>  
> @@ -310,6 +320,9 @@ va_heap::release (vec<T, va_heap, vl_emb
>    if (v == NULL)
>      return;
>  
> +  if (!std::is_trivially_destructible <T>::value)
> +    vec_destruct (v->address (), v->length ());
> +
>    if (GATHER_STATISTICS)
>      v->m_vecpfx.release_overhead (v, elt_size * v->allocated (),
>  				  v->allocated (), true);
> @@ -588,7 +601,10 @@ public:
>    void splice (const vec &);
>    void splice (const vec *src);
>    T *quick_push (const T &);
> -  T &pop (void);
> +  using pop_ret_type
> +    = typename std::conditional <std::is_trivially_destructible <T>::value,
> +				 T &, void>::type;
> +  pop_ret_type pop (void);
>    void truncate (unsigned);
>    void quick_insert (unsigned, const T &);
>    void ordered_remove (unsigned);
> @@ -735,8 +751,9 @@ vec_safe_grow_cleared (vec<T, A, vl_embe
>  		       bool exact = false CXX_MEM_STAT_INFO)
>  {
>    unsigned oldlen = vec_safe_length (v);
> -  vec_safe_grow (v, len, exact PASS_MEM_STAT);
> -  vec_default_construct (v->address () + oldlen, len - oldlen);
> +  gcc_checking_assert (len >= oldlen);
> +  vec_safe_reserve (v, len - oldlen, exact PASS_MEM_STAT);
> +  v->quick_grow_cleared (len);
>  }
>  
>  
> @@ -1005,19 +1022,24 @@ vec<T, A, vl_embed>::quick_push (const T
>  {
>    gcc_checking_assert (space (1));
>    T *slot = &address ()[m_vecpfx.m_num++];
> -  *slot = obj;
> +  ::new (static_cast<void*>(slot)) T (obj);
>    return slot;
>  }
>  
>  
> -/* Pop and return the last element off the end of the vector.  */
> +/* Pop and return a reference to the last element off the end of the
> +   vector.  If T has non-trivial destructor, this method just pops
> +   the element and returns void type.  */
>  
>  template<typename T, typename A>
> -inline T &
> +inline typename vec<T, A, vl_embed>::pop_ret_type
>  vec<T, A, vl_embed>::pop (void)
>  {
>    gcc_checking_assert (length () > 0);
> -  return address ()[--m_vecpfx.m_num];
> +  T &last = address ()[--m_vecpfx.m_num];
> +  if (!std::is_trivially_destructible <T>::value)
> +    last.~T ();
> +  return static_cast <pop_ret_type> (last);
>  }
>  
>  
> @@ -1028,13 +1050,17 @@ template<typename T, typename A>
>  inline void
>  vec<T, A, vl_embed>::truncate (unsigned size)
>  {
> -  gcc_checking_assert (length () >= size);
> +  unsigned l = length ();
> +  gcc_checking_assert (l >= size);
> +  if (!std::is_trivially_destructible <T>::value)
> +    vec_destruct (address () + l, l - size);
>    m_vecpfx.m_num = size;
>  }
>  
>  
>  /* Insert an element, OBJ, at the IXth position of this vector.  There
> -   must be sufficient space.  */
> +   must be sufficient space.  This operation is not suitable for non-trivially
> +   copyable types.  */
>  
>  template<typename T, typename A>
>  inline void
> @@ -1042,6 +1068,7 @@ vec<T, A, vl_embed>::quick_insert (unsig
>  {
>    gcc_checking_assert (length () < allocated ());
>    gcc_checking_assert (ix <= length ());
> +  static_assert (std::is_trivially_copyable <T>::value, "");
>    T *slot = &address ()[ix];
>    memmove (slot + 1, slot, (m_vecpfx.m_num++ - ix) * sizeof (T));
>    *slot = obj;
> @@ -1050,13 +1077,14 @@ vec<T, A, vl_embed>::quick_insert (unsig
>  
>  /* Remove an element from the IXth position of this vector.  Ordering of
>     remaining elements is preserved.  This is an O(N) operation due to
> -   memmove.  */
> +   memmove.  Not suitable for non-trivially copyable types.  */
>  
>  template<typename T, typename A>
>  inline void
>  vec<T, A, vl_embed>::ordered_remove (unsigned ix)
>  {
>    gcc_checking_assert (ix < length ());
> +  static_assert (std::is_trivially_copyable <T>::value, "");
>    T *slot = &address ()[ix];
>    memmove (slot, slot + 1, (--m_vecpfx.m_num - ix) * sizeof (T));
>  }
> @@ -1104,6 +1132,7 @@ inline void
>  vec<T, A, vl_embed>::unordered_remove (unsigned ix)
>  {
>    gcc_checking_assert (ix < length ());
> +  static_assert (std::is_trivially_copyable <T>::value, "");
>    T *p = address ();
>    p[ix] = p[--m_vecpfx.m_num];
>  }
> @@ -1117,6 +1146,7 @@ inline void
>  vec<T, A, vl_embed>::block_remove (unsigned ix, unsigned len)
>  {
>    gcc_checking_assert (ix + len <= length ());
> +  static_assert (std::is_trivially_copyable <T>::value, "");
>    T *slot = &address ()[ix];
>    m_vecpfx.m_num -= len;
>    memmove (slot, slot + len, (m_vecpfx.m_num - ix) * sizeof (T));
> @@ -1130,6 +1160,7 @@ template<typename T, typename A>
>  inline void
>  vec<T, A, vl_embed>::qsort (int (*cmp) (const void *, const void *))
>  {
> +  static_assert (std::is_trivially_copyable <T>::value, "");
>    if (length () > 1)
>      gcc_qsort (address (), length (), sizeof (T), cmp);
>  }
> @@ -1142,6 +1173,7 @@ inline void
>  vec<T, A, vl_embed>::sort (int (*cmp) (const void *, const void *, void *),
>  			   void *data)
>  {
> +  static_assert (std::is_trivially_copyable <T>::value, "");
>    if (length () > 1)
>      gcc_sort_r (address (), length (), sizeof (T), cmp, data);
>  }
> @@ -1154,6 +1186,7 @@ inline void
>  vec<T, A, vl_embed>::stablesort (int (*cmp) (const void *, const void *,
>  					     void *), void *data)
>  {
> +  static_assert (std::is_trivially_copyable <T>::value, "");
>    if (length () > 1)
>      gcc_stablesort_r (address (), length (), sizeof (T), cmp, data);
>  }
> @@ -1326,6 +1359,7 @@ inline void
>  vec<T, A, vl_embed>::quick_grow (unsigned len)
>  {
>    gcc_checking_assert (length () <= len && len <= m_vecpfx.m_alloc);
> +  static_assert (std::is_trivially_default_constructible <T>::value, "");
>    m_vecpfx.m_num = len;
>  }
>  
> @@ -1339,7 +1373,8 @@ vec<T, A, vl_embed>::quick_grow_cleared
>  {
>    unsigned oldlen = length ();
>    size_t growby = len - oldlen;
> -  quick_grow (len);
> +  gcc_checking_assert (length () <= len && len <= m_vecpfx.m_alloc);
> +  m_vecpfx.m_num = len;
>    if (growby != 0)
>      vec_default_construct (address () + oldlen, growby);
>  }
> @@ -1350,6 +1385,7 @@ template<typename T>
>  void
>  gt_ggc_mx (vec<T, va_gc> *v)
>  {
> +  static_assert (std::is_trivially_destructible <T>::value, "");
>    extern void gt_ggc_mx (T &);
>    for (unsigned i = 0; i < v->length (); i++)
>      gt_ggc_mx ((*v)[i]);
> @@ -1359,6 +1395,7 @@ template<typename T>
>  void
>  gt_ggc_mx (vec<T, va_gc_atomic, vl_embed> *v ATTRIBUTE_UNUSED)
>  {
> +  static_assert (std::is_trivially_destructible <T>::value, "");
>    /* Nothing to do.  Vectors of atomic types wrt GC do not need to
>       be traversed.  */
>  }
> @@ -1518,7 +1555,10 @@ public:
>    void safe_splice (const vec & CXX_MEM_STAT_INFO);
>    T *quick_push (const T &);
>    T *safe_push (const T &CXX_MEM_STAT_INFO);
> -  T &pop (void);
> +  using pop_ret_type
> +    = typename std::conditional <std::is_trivially_destructible <T>::value,
> +				 T &, void>::type;
> +  pop_ret_type pop (void);
>    void truncate (unsigned);
>    void safe_grow (unsigned, bool = false CXX_MEM_STAT_INFO);
>    void safe_grow_cleared (unsigned, bool = false CXX_MEM_STAT_INFO);
> @@ -1627,8 +1667,8 @@ public:
>  
>    auto_vec& operator= (vec<T, va_heap>&& r)
>      {
> -	    if (this == &r)
> -		    return *this;
> +      if (this == &r)
> +	return *this;
>  
>        gcc_assert (!r.using_auto_storage ());
>        this->release ();
> @@ -1639,8 +1679,8 @@ public:
>  
>    auto_vec& operator= (auto_vec<T> &&r)
>      {
> -	    if (this == &r)
> -		    return *this;
> +      if (this == &r)
> +	return *this;
>  
>        gcc_assert (!r.using_auto_storage ());
>        this->release ();
> @@ -1660,7 +1700,7 @@ public:
>    // You probably don't want to copy a vector, so these are deleted to prevent
>    // unintentional use.  If you really need a copy of the vectors contents you
>    // can use copy ().
> -  auto_vec(const auto_vec &) = delete;
> +  auto_vec (const auto_vec &) = delete;
>    auto_vec &operator= (const auto_vec &) = delete;
>  };
>  
> @@ -1986,10 +2026,12 @@ vec<T, va_heap, vl_ptr>::safe_push (cons
>  }
>  
>  
> -/* Pop and return the last element off the end of the vector.  */
> +/* Pop and return a reference to the last element off the end of the
> +   vector.  If T has non-trivial destructor, this method just pops
> +   last element and returns void.  */
>  
>  template<typename T>
> -inline T &
> +inline typename vec<T, va_heap, vl_ptr>::pop_ret_type
>  vec<T, va_heap, vl_ptr>::pop (void)
>  {
>    return m_vec->pop ();
> @@ -2038,10 +2080,12 @@ vec<T, va_heap, vl_ptr>::safe_grow_clear
>  					    MEM_STAT_DECL)
>  {
>    unsigned oldlen = length ();
> -  size_t growby = len - oldlen;
> -  safe_grow (len, exact PASS_MEM_STAT);
> -  if (growby != 0)
> -    vec_default_construct (address () + oldlen, growby);
> +  gcc_checking_assert (oldlen <= len);
> +  reserve (len - oldlen, exact PASS_MEM_STAT);
> +  if (m_vec)
> +    m_vec->quick_grow_cleared (len);
> +  else
> +    gcc_checking_assert (len == 0);
>  }
>  
>  
> --- gcc/edit-context.cc.jj	2023-09-27 10:37:38.600848724 +0200
> +++ gcc/edit-context.cc	2023-09-27 10:40:04.834812220 +0200
> @@ -122,6 +122,32 @@ class added_line
>    int m_len;
>  };
>  
> +/* Class for representing edit events that have occurred on one line of
> +   one file: the replacement of some text betweeen some columns
> +   on the line.
> +
> +   Subsequent events will need their columns adjusting if they're
> +   are on this line and their column is >= the start point.  */
> +
> +class line_event
> +{
> + public:
> +  line_event (int start, int next, int len) : m_start (start),
> +    m_delta (len - (next - start)) {}
> +
> +  int get_effective_column (int orig_column) const
> +  {
> +    if (orig_column >= m_start)
> +      return orig_column += m_delta;
> +    else
> +      return orig_column;
> +  }
> +
> + private:
> +  int m_start;
> +  int m_delta;
> +};
> +
>  /* The state of one edited line within an edited_file.
>     As well as the current content of the line, it contains a record of
>     the changes, so that further changes can be applied in the correct
> @@ -172,32 +198,6 @@ class edited_line
>    auto_vec <added_line *> m_predecessors;
>  };
>  
> -/* Class for representing edit events that have occurred on one line of
> -   one file: the replacement of some text betweeen some columns
> -   on the line.
> -
> -   Subsequent events will need their columns adjusting if they're
> -   are on this line and their column is >= the start point.  */
> -
> -class line_event
> -{
> - public:
> -  line_event (int start, int next, int len) : m_start (start),
> -    m_delta (len - (next - start)) {}
> -
> -  int get_effective_column (int orig_column) const
> -  {
> -    if (orig_column >= m_start)
> -      return orig_column += m_delta;
> -    else
> -      return orig_column;
> -  }
> -
> - private:
> -  int m_start;
> -  int m_delta;
> -};
> -
>  /* Forward decls.  */
>  
>  static void
> 
> 	Jakub
> 
>
  

Hello,

Le 27/09/2023 à 12:46, Jakub Jelinek a écrit :
> --- gcc/vec.h.jj	2023-09-27 10:38:50.635845540 +0200
> +++ gcc/vec.h	2023-09-27 12:11:56.665586490 +0200
> @@ -1028,13 +1050,17 @@ template<typename T, typename A>
>   inline void
>   vec<T, A, vl_embed>::truncate (unsigned size)
>   {
> -  gcc_checking_assert (length () >= size);
> +  unsigned l = length ();
> +  gcc_checking_assert (l >= size);
> +  if (!std::is_trivially_destructible <T>::value)
> +    vec_destruct (address () + l, l - size);

Shouldn't this line be:

	vec_destruct (address () + *size*, l - size);

instead?

>     m_vecpfx.m_num = size;
>   }
>   
>