tree-vect-patterns: Pattern recognize ctz or ffs using clz, popcount or ctz [PR109011]

  Hi!

The following patch allows to vectorize __builtin_ffs*/.FFS even if
we just have vector .CTZ support, or __builtin_ffs*/.FFS/__builtin_ctz*/.CTZ
if we just have vector .CLZ or .POPCOUNT support.
It uses various expansions from Hacker's Delight book as well as GCC's
expansion, in particular:
.CTZ (X) = PREC - .CLZ ((X - 1) & ~X)
.CTZ (X) = .POPCOUNT ((X - 1) & ~X)
.CTZ (X) = (PREC - 1) - .CLZ (X & -X)
.FFS (X) = PREC - .CLZ (X & -X)
.CTZ (X) = PREC - .POPCOUNT (X | -X)
.FFS (X) = (PREC + 1) - .POPCOUNT (X | -X)
.FFS (X) = .CTZ (X) + 1
where the first one can be only used if both CTZ and CLZ have value
defined at zero (kind 2) and both have value of PREC there.
If the original has value defined at zero and the latter doesn't
for other forms or if it doesn't have matching value for that case,
a COND_EXPR is added for that afterwards.

The patch also modifies vect_recog_popcount_clz_ctz_ffs_pattern
such that the two can work together.

Bootstrapped/regtested on x86_64-linux and i686-linux, plus tested
on the testcases on powerpc64le-linux and s390x-linux crosses, ok for trunk?

2023-04-20  Jakub Jelinek  <jakub@redhat.com>

	PR tree-optimization/109011
	* tree-vect-patterns.cc (vect_recog_ctz_ffs_pattern): New function.
	(vect_recog_popcount_clz_ctz_ffs_pattern): Move vect_pattern_detected
	call later.  Don't punt for IFN_CTZ or IFN_FFS if it doesn't have
	direct optab support, but has instead IFN_CLZ, IFN_POPCOUNT or
	for IFN_FFS IFN_CTZ support, use vect_recog_ctz_ffs_pattern for that
	case.
	(vect_vect_recog_func_ptrs): Add ctz_ffs entry.

	* gcc.dg/vect/pr109011-1.c: Remove -mpower9-vector from
	dg-additional-options.
	(baz, qux): Remove functions and corresponding dg-final.
	* gcc.dg/vect/pr109011-2.c: New test.
	* gcc.dg/vect/pr109011-3.c: New test.
	* gcc.dg/vect/pr109011-4.c: New test.
	* gcc.dg/vect/pr109011-5.c: New test.


	Jakub
  

On Thu, 20 Apr 2023, Jakub Jelinek wrote:

> Hi!
> 
> The following patch allows to vectorize __builtin_ffs*/.FFS even if
> we just have vector .CTZ support, or __builtin_ffs*/.FFS/__builtin_ctz*/.CTZ
> if we just have vector .CLZ or .POPCOUNT support.
> It uses various expansions from Hacker's Delight book as well as GCC's
> expansion, in particular:
> .CTZ (X) = PREC - .CLZ ((X - 1) & ~X)
> .CTZ (X) = .POPCOUNT ((X - 1) & ~X)
> .CTZ (X) = (PREC - 1) - .CLZ (X & -X)
> .FFS (X) = PREC - .CLZ (X & -X)
> .CTZ (X) = PREC - .POPCOUNT (X | -X)
> .FFS (X) = (PREC + 1) - .POPCOUNT (X | -X)
> .FFS (X) = .CTZ (X) + 1
> where the first one can be only used if both CTZ and CLZ have value
> defined at zero (kind 2) and both have value of PREC there.
> If the original has value defined at zero and the latter doesn't
> for other forms or if it doesn't have matching value for that case,
> a COND_EXPR is added for that afterwards.
> 
> The patch also modifies vect_recog_popcount_clz_ctz_ffs_pattern
> such that the two can work together.
> 
> Bootstrapped/regtested on x86_64-linux and i686-linux, plus tested
> on the testcases on powerpc64le-linux and s390x-linux crosses, ok for trunk?

OK.

Thanks,
Richard.

> 2023-04-20  Jakub Jelinek  <jakub@redhat.com>
> 
> 	PR tree-optimization/109011
> 	* tree-vect-patterns.cc (vect_recog_ctz_ffs_pattern): New function.
> 	(vect_recog_popcount_clz_ctz_ffs_pattern): Move vect_pattern_detected
> 	call later.  Don't punt for IFN_CTZ or IFN_FFS if it doesn't have
> 	direct optab support, but has instead IFN_CLZ, IFN_POPCOUNT or
> 	for IFN_FFS IFN_CTZ support, use vect_recog_ctz_ffs_pattern for that
> 	case.
> 	(vect_vect_recog_func_ptrs): Add ctz_ffs entry.
> 
> 	* gcc.dg/vect/pr109011-1.c: Remove -mpower9-vector from
> 	dg-additional-options.
> 	(baz, qux): Remove functions and corresponding dg-final.
> 	* gcc.dg/vect/pr109011-2.c: New test.
> 	* gcc.dg/vect/pr109011-3.c: New test.
> 	* gcc.dg/vect/pr109011-4.c: New test.
> 	* gcc.dg/vect/pr109011-5.c: New test.
> 
> --- gcc/tree-vect-patterns.cc.jj	2023-04-19 11:14:17.445843870 +0200
> +++ gcc/tree-vect-patterns.cc	2023-04-19 20:49:27.946432713 +0200
> @@ -1501,6 +1501,266 @@ vect_recog_widen_minus_pattern (vec_info
>  				      "vect_recog_widen_minus_pattern");
>  }
>  
> +/* Function vect_recog_ctz_ffs_pattern
> +
> +   Try to find the following pattern:
> +
> +   TYPE1 A;
> +   TYPE1 B;
> +
> +   B = __builtin_ctz{,l,ll} (A);
> +
> +   or
> +
> +   B = __builtin_ffs{,l,ll} (A);
> +
> +   Input:
> +
> +   * STMT_VINFO: The stmt from which the pattern search begins.
> +   here it starts with B = __builtin_* (A);
> +
> +   Output:
> +
> +   * TYPE_OUT: The vector type of the output of this pattern.
> +
> +   * Return value: A new stmt that will be used to replace the sequence of
> +   stmts that constitute the pattern, using clz or popcount builtins.  */
> +
> +static gimple *
> +vect_recog_ctz_ffs_pattern (vec_info *vinfo, stmt_vec_info stmt_vinfo,
> +			    tree *type_out)
> +{
> +  gimple *call_stmt = stmt_vinfo->stmt;
> +  gimple *pattern_stmt;
> +  tree rhs_oprnd, rhs_type, lhs_oprnd, lhs_type, vec_type, vec_rhs_type;
> +  tree new_var;
> +  internal_fn ifn = IFN_LAST, ifnnew = IFN_LAST;
> +  bool defined_at_zero = true, defined_at_zero_new = false;
> +  int val = 0, val_new = 0;
> +  int prec;
> +  int sub = 0, add = 0;
> +  location_t loc;
> +
> +  if (!is_gimple_call (call_stmt))
> +    return NULL;
> +
> +  if (gimple_call_num_args (call_stmt) != 1)
> +    return NULL;
> +
> +  rhs_oprnd = gimple_call_arg (call_stmt, 0);
> +  rhs_type = TREE_TYPE (rhs_oprnd);
> +  lhs_oprnd = gimple_call_lhs (call_stmt);
> +  if (!lhs_oprnd)
> +    return NULL;
> +  lhs_type = TREE_TYPE (lhs_oprnd);
> +  if (!INTEGRAL_TYPE_P (lhs_type)
> +      || !INTEGRAL_TYPE_P (rhs_type)
> +      || !type_has_mode_precision_p (rhs_type)
> +      || TREE_CODE (rhs_oprnd) != SSA_NAME)
> +    return NULL;
> +
> +  switch (gimple_call_combined_fn (call_stmt))
> +    {
> +    CASE_CFN_CTZ:
> +      ifn = IFN_CTZ;
> +      if (!gimple_call_internal_p (call_stmt)
> +	  || CTZ_DEFINED_VALUE_AT_ZERO (SCALAR_INT_TYPE_MODE (rhs_type),
> +					val) != 2)
> +	defined_at_zero = false;
> +      break;
> +    CASE_CFN_FFS:
> +      ifn = IFN_FFS;
> +      break;
> +    default:
> +      return NULL;
> +    }
> +
> +  prec = TYPE_PRECISION (rhs_type);
> +  loc = gimple_location (call_stmt);
> +
> +  vec_type = get_vectype_for_scalar_type (vinfo, lhs_type);
> +  if (!vec_type)
> +    return NULL;
> +
> +  vec_rhs_type = get_vectype_for_scalar_type (vinfo, rhs_type);
> +  if (!vec_rhs_type)
> +    return NULL;
> +
> +  /* Do it only if the backend doesn't have ctz<vector_mode>2 or
> +     ffs<vector_mode>2 pattern but does have clz<vector_mode>2 or
> +     popcount<vector_mode>2.  */
> +  if (!vec_type
> +      || direct_internal_fn_supported_p (ifn, vec_rhs_type,
> +					 OPTIMIZE_FOR_SPEED))
> +    return NULL;
> +
> +  if (ifn == IFN_FFS
> +      && direct_internal_fn_supported_p (IFN_CTZ, vec_rhs_type,
> +					 OPTIMIZE_FOR_SPEED))
> +    {
> +      ifnnew = IFN_CTZ;
> +      defined_at_zero_new
> +	= CTZ_DEFINED_VALUE_AT_ZERO (SCALAR_INT_TYPE_MODE (rhs_type),
> +				     val_new) == 2;
> +    }
> +  else if (direct_internal_fn_supported_p (IFN_CLZ, vec_rhs_type,
> +					   OPTIMIZE_FOR_SPEED))
> +    {
> +      ifnnew = IFN_CLZ;
> +      defined_at_zero_new
> +	= CLZ_DEFINED_VALUE_AT_ZERO (SCALAR_INT_TYPE_MODE (rhs_type),
> +				     val_new) == 2;
> +    }
> +  if ((ifnnew == IFN_LAST
> +       || (defined_at_zero && !defined_at_zero_new))
> +      && direct_internal_fn_supported_p (IFN_POPCOUNT, vec_rhs_type,
> +					 OPTIMIZE_FOR_SPEED))
> +    {
> +      ifnnew = IFN_POPCOUNT;
> +      defined_at_zero_new = true;
> +      val_new = prec;
> +    }
> +  if (ifnnew == IFN_LAST)
> +    return NULL;
> +
> +  vect_pattern_detected ("vec_recog_ctz_ffs_pattern", call_stmt);
> +
> +  if ((ifnnew == IFN_CLZ
> +       && defined_at_zero
> +       && defined_at_zero_new
> +       && val == prec
> +       && val_new == prec)
> +      || (ifnnew == IFN_POPCOUNT && ifn == IFN_CLZ))
> +    {
> +      /* .CTZ (X) = PREC - .CLZ ((X - 1) & ~X)
> +	 .CTZ (X) = .POPCOUNT ((X - 1) & ~X).  */
> +      if (ifnnew == IFN_CLZ)
> +	sub = prec;
> +      val_new = prec;
> +
> +      if (!TYPE_UNSIGNED (rhs_type))
> +	{
> +	  rhs_type = unsigned_type_for (rhs_type);
> +	  vec_rhs_type = get_vectype_for_scalar_type (vinfo, rhs_type);
> +	  new_var = vect_recog_temp_ssa_var (rhs_type, NULL);
> +	  pattern_stmt = gimple_build_assign (new_var, NOP_EXPR, rhs_oprnd);
> +	  append_pattern_def_seq (vinfo, stmt_vinfo, pattern_stmt,
> +				  vec_rhs_type);
> +	  rhs_oprnd = new_var;
> +	}
> +
> +      tree m1 = vect_recog_temp_ssa_var (rhs_type, NULL);
> +      pattern_stmt = gimple_build_assign (m1, PLUS_EXPR, rhs_oprnd,
> +					  build_int_cst (rhs_type, -1));
> +      gimple_set_location (pattern_stmt, loc);
> +      append_pattern_def_seq (vinfo, stmt_vinfo, pattern_stmt, vec_rhs_type);
> +
> +      new_var = vect_recog_temp_ssa_var (rhs_type, NULL);
> +      pattern_stmt = gimple_build_assign (new_var, BIT_NOT_EXPR, rhs_oprnd);
> +      gimple_set_location (pattern_stmt, loc);
> +      append_pattern_def_seq (vinfo, stmt_vinfo, pattern_stmt, vec_rhs_type);
> +      rhs_oprnd = new_var;
> +
> +      new_var = vect_recog_temp_ssa_var (rhs_type, NULL);
> +      pattern_stmt = gimple_build_assign (new_var, BIT_AND_EXPR,
> +					  m1, rhs_oprnd);
> +      gimple_set_location (pattern_stmt, loc);
> +      append_pattern_def_seq (vinfo, stmt_vinfo, pattern_stmt, vec_rhs_type);
> +      rhs_oprnd = new_var;
> +    }
> +  else if (ifnnew == IFN_CLZ)
> +    {
> +      /* .CTZ (X) = (PREC - 1) - .CLZ (X & -X)
> +	 .FFS (X) = PREC - .CLZ (X & -X).  */
> +      sub = prec - (ifn == IFN_CTZ);
> +      val_new = sub - val_new;
> +
> +      tree neg = vect_recog_temp_ssa_var (rhs_type, NULL);
> +      pattern_stmt = gimple_build_assign (neg, NEGATE_EXPR, rhs_oprnd);
> +      gimple_set_location (pattern_stmt, loc);
> +      append_pattern_def_seq (vinfo, stmt_vinfo, pattern_stmt, vec_rhs_type);
> +
> +      new_var = vect_recog_temp_ssa_var (rhs_type, NULL);
> +      pattern_stmt = gimple_build_assign (new_var, BIT_AND_EXPR,
> +					  rhs_oprnd, neg);
> +      gimple_set_location (pattern_stmt, loc);
> +      append_pattern_def_seq (vinfo, stmt_vinfo, pattern_stmt, vec_rhs_type);
> +      rhs_oprnd = new_var;
> +    }
> +  else if (ifnnew == IFN_POPCOUNT)
> +    {
> +      /* .CTZ (X) = PREC - .POPCOUNT (X | -X)
> +	 .FFS (X) = (PREC + 1) - .POPCOUNT (X | -X).  */
> +      sub = prec + (ifn == IFN_FFS);
> +      val_new = sub;
> +
> +      tree neg = vect_recog_temp_ssa_var (rhs_type, NULL);
> +      pattern_stmt = gimple_build_assign (neg, NEGATE_EXPR, rhs_oprnd);
> +      gimple_set_location (pattern_stmt, loc);
> +      append_pattern_def_seq (vinfo, stmt_vinfo, pattern_stmt, vec_rhs_type);
> +
> +      new_var = vect_recog_temp_ssa_var (rhs_type, NULL);
> +      pattern_stmt = gimple_build_assign (new_var, BIT_IOR_EXPR,
> +					  rhs_oprnd, neg);
> +      gimple_set_location (pattern_stmt, loc);
> +      append_pattern_def_seq (vinfo, stmt_vinfo, pattern_stmt, vec_rhs_type);
> +      rhs_oprnd = new_var;
> +    }
> +  else if (ifnnew == IFN_CTZ)
> +    {
> +      /* .FFS (X) = .CTZ (X) + 1.  */
> +      add = 1;
> +      val_new++;
> +    }
> +
> +  /* Create B = .IFNNEW (A).  */
> +  new_var = vect_recog_temp_ssa_var (lhs_type, NULL);
> +  pattern_stmt = gimple_build_call_internal (ifnnew, 1, rhs_oprnd);
> +  gimple_call_set_lhs (pattern_stmt, new_var);
> +  gimple_set_location (pattern_stmt, loc);
> +  *type_out = vec_type;
> +
> +  if (sub)
> +    {
> +      append_pattern_def_seq (vinfo, stmt_vinfo, pattern_stmt, vec_type);
> +      tree ret_var = vect_recog_temp_ssa_var (lhs_type, NULL);
> +      pattern_stmt = gimple_build_assign (ret_var, MINUS_EXPR,
> +					  build_int_cst (lhs_type, sub),
> +					  new_var);
> +      gimple_set_location (pattern_stmt, loc);
> +      new_var = ret_var;
> +    }
> +  else if (add)
> +    {
> +      append_pattern_def_seq (vinfo, stmt_vinfo, pattern_stmt, vec_type);
> +      tree ret_var = vect_recog_temp_ssa_var (lhs_type, NULL);
> +      pattern_stmt = gimple_build_assign (ret_var, PLUS_EXPR, new_var,
> +					  build_int_cst (lhs_type, add));
> +      gimple_set_location (pattern_stmt, loc);
> +      new_var = ret_var;
> +    }
> +
> +  if (defined_at_zero
> +      && (!defined_at_zero_new || val != val_new))
> +    {
> +      append_pattern_def_seq (vinfo, stmt_vinfo, pattern_stmt, vec_type);
> +      tree ret_var = vect_recog_temp_ssa_var (lhs_type, NULL);
> +      rhs_oprnd = gimple_call_arg (call_stmt, 0);
> +      rhs_type = TREE_TYPE (rhs_oprnd);
> +      tree cmp = build2_loc (loc, NE_EXPR, boolean_type_node,
> +			     rhs_oprnd, build_zero_cst (rhs_type));
> +      pattern_stmt = gimple_build_assign (ret_var, COND_EXPR, cmp,
> +					  new_var,
> +					  build_int_cst (lhs_type, val));
> +    }
> +
> +  if (dump_enabled_p ())
> +    dump_printf_loc (MSG_NOTE, vect_location,
> +		     "created pattern stmt: %G", pattern_stmt);
> +
> +  return pattern_stmt;
> +}
> +
>  /* Function vect_recog_popcount_clz_ctz_ffs_pattern
>  
>     Try to find the following pattern:
> @@ -1680,15 +1940,42 @@ vect_recog_popcount_clz_ctz_ffs_pattern
>  	gcc_unreachable ();
>        }
>  
> -  vect_pattern_detected ("vec_recog_popcount_clz_ctz_ffs_pattern",
> -			 call_stmt);
>    vec_type = get_vectype_for_scalar_type (vinfo, lhs_type);
>    /* Do it only if the backend has popcount<vector_mode>2 etc. pattern.  */
> -  if (!vec_type
> -      || !direct_internal_fn_supported_p (ifn, vec_type,
> -					  OPTIMIZE_FOR_SPEED))
> +  if (!vec_type)
>      return NULL;
>  
> +  bool supported
> +    = direct_internal_fn_supported_p (ifn, vec_type, OPTIMIZE_FOR_SPEED);
> +  if (!supported)
> +    switch (ifn)
> +      {
> +      case IFN_POPCOUNT:
> +      case IFN_CLZ:
> +	return NULL;
> +      case IFN_FFS:
> +	/* vect_recog_ctz_ffs_pattern can implement ffs using ctz.  */
> +	if (direct_internal_fn_supported_p (IFN_CTZ, vec_type,
> +					    OPTIMIZE_FOR_SPEED))
> +	  break;
> +	/* FALLTHRU */
> +      case IFN_CTZ:
> +	/* vect_recog_ctz_ffs_pattern can implement ffs or ctz using
> +	   clz or popcount.  */
> +	if (direct_internal_fn_supported_p (IFN_CLZ, vec_type,
> +					    OPTIMIZE_FOR_SPEED))
> +	  break;
> +	if (direct_internal_fn_supported_p (IFN_POPCOUNT, vec_type,
> +					    OPTIMIZE_FOR_SPEED))
> +	  break;
> +	return NULL;
> +      default:
> +	gcc_unreachable ();
> +      }
> +
> +  vect_pattern_detected ("vec_recog_popcount_clz_ctz_ffs_pattern",
> +			 call_stmt);
> +
>    /* Create B = .POPCOUNT (A).  */
>    new_var = vect_recog_temp_ssa_var (lhs_type, NULL);
>    pattern_stmt = gimple_build_call_internal (ifn, 1, unprom_diff.op);
> @@ -1702,11 +1989,26 @@ vect_recog_popcount_clz_ctz_ffs_pattern
>  
>    if (addend)
>      {
> +      gcc_assert (supported);
>        append_pattern_def_seq (vinfo, stmt_vinfo, pattern_stmt, vec_type);
>        tree ret_var = vect_recog_temp_ssa_var (lhs_type, NULL);
>        pattern_stmt = gimple_build_assign (ret_var, PLUS_EXPR, new_var,
>  					  build_int_cst (lhs_type, addend));
>      }
> +  else if (!supported)
> +    {
> +      stmt_vec_info new_stmt_info = vinfo->add_stmt (pattern_stmt);
> +      STMT_VINFO_VECTYPE (new_stmt_info) = vec_type;
> +      pattern_stmt
> +	= vect_recog_ctz_ffs_pattern (vinfo, new_stmt_info, type_out);
> +      if (pattern_stmt == NULL)
> +	return NULL;
> +      if (gimple_seq seq = STMT_VINFO_PATTERN_DEF_SEQ (new_stmt_info))
> +	{
> +	  gimple_seq *pseq = &STMT_VINFO_PATTERN_DEF_SEQ (stmt_vinfo);
> +	  gimple_seq_add_seq_without_update (pseq, seq);
> +	}
> +    }
>    return pattern_stmt;
>  }
>  
> @@ -6150,6 +6452,7 @@ static vect_recog_func vect_vect_recog_f
>    { vect_recog_widen_sum_pattern, "widen_sum" },
>    { vect_recog_pow_pattern, "pow" },
>    { vect_recog_popcount_clz_ctz_ffs_pattern, "popcount_clz_ctz_ffs" },
> +  { vect_recog_ctz_ffs_pattern, "ctz_ffs" },
>    { vect_recog_widen_shift_pattern, "widen_shift" },
>    { vect_recog_rotate_pattern, "rotate" },
>    { vect_recog_vector_vector_shift_pattern, "vector_vector_shift" },
> --- gcc/testsuite/gcc.dg/vect/pr109011-1.c.jj	2023-04-19 11:14:17.458843682 +0200
> +++ gcc/testsuite/gcc.dg/vect/pr109011-1.c	2023-04-19 20:59:52.080597720 +0200
> @@ -4,7 +4,6 @@
>  /* { dg-additional-options "-mavx512cd" { target { { i?86-*-* x86_64-*-* } && avx512cd } } } */
>  /* { dg-additional-options "-mavx512vpopcntdq" { target { { i?86-*-* x86_64-*-* } && avx512vpopcntdq } } } */
>  /* { dg-additional-options "-mpower8-vector" { target powerpc_p8vector_ok } } */
> -/* { dg-additional-options "-mpower9-vector" { target powerpc_p9vector_ok } } */
>  /* { dg-additional-options "-march=z13 -mzarch" { target s390_vx } } */
>  
>  void
> @@ -28,21 +27,3 @@ bar (long long *p, long long *q)
>  
>  /* { dg-final { scan-tree-dump-times " = \.CLZ \\\(" 1 "optimized" { target { { i?86-*-* x86_64-*-* } && avx512cd } } } } */
>  /* { dg-final { scan-tree-dump-times " = \.CLZ \\\(" 1 "optimized" { target { powerpc_p8vector_ok || s390_vx } } } } */
> -
> -void
> -baz (long long *p, long long *q)
> -{
> -#pragma omp simd
> -  for (int i = 0; i < 2048; ++i)
> -    p[i] = __builtin_ctzll (q[i]);
> -}
> -
> -/* { dg-final { scan-tree-dump-times " = \.CTZ \\\(" 1 "optimized" { target { powerpc_p9vector_ok || s390_vx } } } } */
> -
> -void
> -qux (long long *p, long long *q)
> -{
> -#pragma omp simd
> -  for (int i = 0; i < 2048; ++i)
> -    p[i] = __builtin_ffsll (q[i]);
> -}
> --- gcc/testsuite/gcc.dg/vect/pr109011-2.c.jj	2023-04-19 13:03:20.621977340 +0200
> +++ gcc/testsuite/gcc.dg/vect/pr109011-2.c	2023-04-19 20:53:30.205003402 +0200
> @@ -0,0 +1,35 @@
> +/* PR tree-optimization/109011 */
> +/* { dg-do compile } */
> +/* { dg-options "-O3 -fno-unroll-loops --param=vect-epilogues-nomask=0 -fdump-tree-optimized" } */
> +/* { dg-additional-options "-mavx512cd -mbmi -mlzcnt -mno-avx512vpopcntdq" { target { { { { i?86-*-* x86_64-*-* } && avx512cd } && lzcnt } && bmi } } } */
> +/* { dg-additional-options "-mpower9-vector" { target powerpc_p9vector_ok } } */
> +/* { dg-additional-options "-march=z13 -mzarch" { target s390_vx } } */
> +
> +void
> +foo (int *p, int *q)
> +{
> +#pragma omp simd
> +  for (int i = 0; i < 2048; ++i)
> +    p[i] = __builtin_ctz (q[i]);
> +}
> +
> +void
> +bar (int *p, int *q)
> +{
> +#pragma omp simd
> +  for (int i = 0; i < 2048; ++i)
> +    p[i] = q[i] ? __builtin_ctz (q[i]) : __SIZEOF_INT__ * __CHAR_BIT__;
> +}
> +
> +void
> +baz (int *p, int *q)
> +{
> +#pragma omp simd
> +  for (int i = 0; i < 2048; ++i)
> +    p[i] = __builtin_ffs (q[i]);
> +}
> +
> +/* { dg-final { scan-tree-dump-times " = \.CLZ \\\(" 3 "optimized" { target { { { { i?86-*-* x86_64-*-* } && avx512cd } && lzcnt } && bmi } } } } */
> +/* { dg-final { scan-tree-dump-times " = \.CTZ \\\(" 4 "optimized" { target powerpc_p9vector_ok } } } */
> +/* { dg-final { scan-tree-dump-times " = \.CTZ \\\(" 2 "optimized" { target s390_vx } } } */
> +/* { dg-final { scan-tree-dump-times " = \.POPCOUNT \\\(" 1 "optimized" { target s390_vx } } } */
> --- gcc/testsuite/gcc.dg/vect/pr109011-3.c.jj	2023-04-19 13:13:23.524284082 +0200
> +++ gcc/testsuite/gcc.dg/vect/pr109011-3.c	2023-04-19 20:58:19.517908001 +0200
> @@ -0,0 +1,32 @@
> +/* PR tree-optimization/109011 */
> +/* { dg-do compile } */
> +/* { dg-options "-O3 -fno-unroll-loops --param=vect-epilogues-nomask=0 -fdump-tree-optimized" } */
> +/* { dg-additional-options "-mno-avx512cd -mbmi -mlzcnt -mavx512vpopcntdq" { target { { { { i?86-*-* x86_64-*-* } && avx512vpopcntdq } && lzcnt } && bmi } } } */
> +/* { dg-additional-options "-mpower8-vector -mno-power9-vector" { target powerpc_p8vector_ok } } */
> +
> +void
> +foo (int *p, int *q)
> +{
> +#pragma omp simd
> +  for (int i = 0; i < 2048; ++i)
> +    p[i] = __builtin_ctz (q[i]);
> +}
> +
> +void
> +bar (int *p, int *q)
> +{
> +#pragma omp simd
> +  for (int i = 0; i < 2048; ++i)
> +    p[i] = q[i] ? __builtin_ctz (q[i]) : __SIZEOF_INT__ * __CHAR_BIT__;
> +}
> +
> +void
> +baz (int *p, int *q)
> +{
> +#pragma omp simd
> +  for (int i = 0; i < 2048; ++i)
> +    p[i] = __builtin_ffs (q[i]);
> +}
> +
> +/* { dg-final { scan-tree-dump-times " = \.POPCOUNT \\\(" 3 "optimized" { target { { { { i?86-*-* x86_64-*-* } && avx512vpopcntdq } && lzcnt } && bmi } } } } */
> +/* { dg-final { scan-tree-dump-times " = \.CLZ \\\(" 3 "optimized" { target powerpc_p8vector_ok } } } */
> --- gcc/testsuite/gcc.dg/vect/pr109011-4.c.jj	2023-04-19 18:42:02.530527826 +0200
> +++ gcc/testsuite/gcc.dg/vect/pr109011-4.c	2023-04-19 20:57:17.813781462 +0200
> @@ -0,0 +1,35 @@
> +/* PR tree-optimization/109011 */
> +/* { dg-do compile } */
> +/* { dg-options "-O3 -fno-unroll-loops --param=vect-epilogues-nomask=0 -fdump-tree-optimized" } */
> +/* { dg-additional-options "-mavx512cd -mbmi -mlzcnt -mno-avx512vpopcntdq" { target { { { { i?86-*-* x86_64-*-* } && avx512cd } && lzcnt } && bmi } } } */
> +/* { dg-additional-options "-mpower9-vector" { target powerpc_p9vector_ok } } */
> +/* { dg-additional-options "-march=z13 -mzarch" { target s390_vx } } */
> +
> +void
> +foo (long long *p, long long *q)
> +{
> +#pragma omp simd
> +  for (int i = 0; i < 2048; ++i)
> +    p[i] = __builtin_ctzll (q[i]);
> +}
> +
> +void
> +bar (long long *p, long long *q)
> +{
> +#pragma omp simd
> +  for (int i = 0; i < 2048; ++i)
> +    p[i] = q[i] ? __builtin_ctzll (q[i]) : __SIZEOF_LONG_LONG__ * __CHAR_BIT__;
> +}
> +
> +void
> +baz (long long *p, long long *q)
> +{
> +#pragma omp simd
> +  for (int i = 0; i < 2048; ++i)
> +    p[i] = __builtin_ffsll (q[i]);
> +}
> +
> +/* { dg-final { scan-tree-dump-times " = \.CLZ \\\(" 3 "optimized" { target { { { { i?86-*-* x86_64-*-* } && avx512cd } && lzcnt } && bmi } } } } */
> +/* { dg-final { scan-tree-dump-times " = \.CTZ \\\(" 4 "optimized" { target powerpc_p9vector_ok } } } */
> +/* { dg-final { scan-tree-dump-times " = \.CTZ \\\(" 2 "optimized" { target s390_vx } } } */
> +/* { dg-final { scan-tree-dump-times " = \.POPCOUNT \\\(" 1 "optimized" { target s390_vx } } } */
> --- gcc/testsuite/gcc.dg/vect/pr109011-5.c.jj	2023-04-19 18:42:52.249824866 +0200
> +++ gcc/testsuite/gcc.dg/vect/pr109011-5.c	2023-04-19 20:58:33.845705184 +0200
> @@ -0,0 +1,32 @@
> +/* PR tree-optimization/109011 */
> +/* { dg-do compile } */
> +/* { dg-options "-O3 -fno-unroll-loops --param=vect-epilogues-nomask=0 -fdump-tree-optimized" } */
> +/* { dg-additional-options "-mno-avx512cd -mbmi -mlzcnt -mavx512vpopcntdq" { target { { { { i?86-*-* x86_64-*-* } && avx512vpopcntdq } && lzcnt } && bmi } } } */
> +/* { dg-additional-options "-mpower8-vector -mno-power9-vector" { target powerpc_p8vector_ok } } */
> +
> +void
> +foo (long long *p, long long *q)
> +{
> +#pragma omp simd
> +  for (int i = 0; i < 2048; ++i)
> +    p[i] = __builtin_ctzll (q[i]);
> +}
> +
> +void
> +bar (long long *p, long long *q)
> +{
> +#pragma omp simd
> +  for (int i = 0; i < 2048; ++i)
> +    p[i] = q[i] ? __builtin_ctzll (q[i]) : __SIZEOF_LONG_LONG__ * __CHAR_BIT__;
> +}
> +
> +void
> +baz (long long *p, long long *q)
> +{
> +#pragma omp simd
> +  for (int i = 0; i < 2048; ++i)
> +    p[i] = __builtin_ffsll (q[i]);
> +}
> +
> +/* { dg-final { scan-tree-dump-times " = \.POPCOUNT \\\(" 3 "optimized" { target { { { { i?86-*-* x86_64-*-* } && avx512vpopcntdq } && lzcnt } && bmi } } } } */
> +/* { dg-final { scan-tree-dump-times " = \.CLZ \\\(" 3 "optimized" { target powerpc_p8vector_ok } } } */
> 
> 	Jakub
> 
>