public inbox for gcc-cvs@sourceware.org
help / color / mirror / Atom feed
* [gcc r14-108] tree-vect-patterns: Pattern recognize ctz or ffs using clz, popcount or ctz [PR109011]
@ 2023-04-20  9:56 Jakub Jelinek
  0 siblings, 0 replies; only message in thread
From: Jakub Jelinek @ 2023-04-20  9:56 UTC (permalink / raw)
  To: gcc-cvs

https://gcc.gnu.org/g:705b0d2b62318b3935214f08a1cf023b1117acb8

commit r14-108-g705b0d2b62318b3935214f08a1cf023b1117acb8
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Thu Apr 20 11:55:16 2023 +0200

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

Diff:
---
 gcc/testsuite/gcc.dg/vect/pr109011-1.c |  19 --
 gcc/testsuite/gcc.dg/vect/pr109011-2.c |  35 ++++
 gcc/testsuite/gcc.dg/vect/pr109011-3.c |  32 ++++
 gcc/testsuite/gcc.dg/vect/pr109011-4.c |  35 ++++
 gcc/testsuite/gcc.dg/vect/pr109011-5.c |  32 ++++
 gcc/tree-vect-patterns.cc              | 313 ++++++++++++++++++++++++++++++++-
 6 files changed, 442 insertions(+), 24 deletions(-)

diff --git a/gcc/testsuite/gcc.dg/vect/pr109011-1.c b/gcc/testsuite/gcc.dg/vect/pr109011-1.c
index 707a82aaf43..16a58267dc8 100644
--- a/gcc/testsuite/gcc.dg/vect/pr109011-1.c
+++ b/gcc/testsuite/gcc.dg/vect/pr109011-1.c
@@ -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]);
-}
diff --git a/gcc/testsuite/gcc.dg/vect/pr109011-2.c b/gcc/testsuite/gcc.dg/vect/pr109011-2.c
new file mode 100644
index 00000000000..191af8945e5
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/vect/pr109011-2.c
@@ -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 } } } */
diff --git a/gcc/testsuite/gcc.dg/vect/pr109011-3.c b/gcc/testsuite/gcc.dg/vect/pr109011-3.c
new file mode 100644
index 00000000000..2e631fc1506
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/vect/pr109011-3.c
@@ -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 } } } */
diff --git a/gcc/testsuite/gcc.dg/vect/pr109011-4.c b/gcc/testsuite/gcc.dg/vect/pr109011-4.c
new file mode 100644
index 00000000000..ce1ee02516e
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/vect/pr109011-4.c
@@ -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 } } } */
diff --git a/gcc/testsuite/gcc.dg/vect/pr109011-5.c b/gcc/testsuite/gcc.dg/vect/pr109011-5.c
new file mode 100644
index 00000000000..51168ef8708
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/vect/pr109011-5.c
@@ -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 } } } */
diff --git a/gcc/tree-vect-patterns.cc b/gcc/tree-vect-patterns.cc
index 633998e8e3a..d1b86e8b5e0 100644
--- a/gcc/tree-vect-patterns.cc
+++ b/gcc/tree-vect-patterns.cc
@@ -1501,6 +1501,266 @@ vect_recog_widen_minus_pattern (vec_info *vinfo, stmt_vec_info last_stmt_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 (vec_info *vinfo,
 	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 (vec_info *vinfo,
 
   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_func_ptrs[] = {
   { 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" },

^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2023-04-20  9:56 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-04-20  9:56 [gcc r14-108] tree-vect-patterns: Pattern recognize ctz or ffs using clz, popcount or ctz [PR109011] Jakub Jelinek

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).