[v6,04/17] Add string vectorized find and detection functions

diff --git a/sysdeps/generic/string-extbyte.h b/sysdeps/generic/string-extbyte.h
new file mode 100644
index 0000000000..38b4674dca
--- /dev/null
+++ b/sysdeps/generic/string-extbyte.h
@@ -0,0 +1,37 @@ 
+/* Extract by from memory word.  Generic C version.
+   Copyright (C) 2023 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#ifndef _STRING_EXTBYTE_H
+#define _STRING_EXTBYTE_H 1
+
+#include <limits.h>
+#include <endian.h>
+#include <string-optype.h>
+
+/* Extract the byte at index IDX from word X, with index 0 being the
+   least significant byte.  */
+static __always_inline unsigned char
+extractbyte (op_t x, unsigned int idx)
+{
+  if (__BYTE_ORDER == __LITTLE_ENDIAN)
+    return x >> (idx * CHAR_BIT);
+  else
+    return x >> (sizeof (x) - 1 - idx) * CHAR_BIT;
+}
+
+#endif /* _STRING_EXTBYTE_H */
diff --git a/sysdeps/generic/string-fza.h b/sysdeps/generic/string-fza.h
new file mode 100644
index 0000000000..1ec51f7f3c
--- /dev/null
+++ b/sysdeps/generic/string-fza.h
@@ -0,0 +1,102 @@ 
+/* Basic zero byte detection.  Generic C version.
+   Copyright (C) 2023 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#ifndef _STRING_FZA_H
+#define _STRING_FZA_H 1
+
+#include <limits.h>
+#include <string-optype.h>
+#include <string-maskoff.h>
+
+/* This function returns non-zero if any byte in X is zero.
+   More specifically, at least one bit set within the least significant
+   byte that was zero; other bytes within the word are indeterminate.  */
+static __always_inline op_t
+find_zero_low (op_t x)
+{
+  /* This expression comes from
+       https://graphics.stanford.edu/~seander/bithacks.html#ZeroInWord
+     Subtracting 1 sets 0x80 in a byte that was 0; anding ~x clears
+     0x80 in a byte that was >= 128; anding 0x80 isolates that test bit.  */
+  op_t lsb = repeat_bytes (0x01);
+  op_t msb = repeat_bytes (0x80);
+  return (x - lsb) & ~x & msb;
+}
+
+/* This function returns at least one bit set within every byte of X that
+   is zero.  The result is exact in that, unlike find_zero_low, all bytes
+   are determinate.  This is usually used for finding the index of the
+   most significant byte that was zero.  */
+static __always_inline op_t
+find_zero_all (op_t x)
+{
+  /* For each byte, find not-zero by
+     (0) And 0x7f so that we cannot carry between bytes,
+     (1) Add 0x7f so that non-zero carries into 0x80,
+     (2) Or in the original byte (which might have had 0x80 set).
+     Then invert and mask such that 0x80 is set iff that byte was zero.  */
+  op_t m = repeat_bytes (0x7f);
+  return ~(((x & m) + m) | x | m);
+}
+
+/* With similar caveats, identify bytes that are equal between X1 and X2.  */
+static __always_inline op_t
+find_eq_low (op_t x1, op_t x2)
+{
+  return find_zero_low (x1 ^ x2);
+}
+
+static __always_inline op_t
+find_eq_all (op_t x1, op_t x2)
+{
+  return find_zero_all (x1 ^ x2);
+}
+
+/* With similar caveats, identify zero bytes in X1 and bytes that are
+   equal between in X1 and X2.  */
+static __always_inline op_t
+find_zero_eq_low (op_t x1, op_t x2)
+{
+  return find_zero_low (x1) | find_zero_low (x1 ^ x2);
+}
+
+static __always_inline op_t
+find_zero_eq_all (op_t x1, op_t x2)
+{
+  return find_zero_all (x1) | find_zero_all (x1 ^ x2);
+}
+
+/* With similar caveats, identify zero bytes in X1 and bytes that are
+   not equal between in X1 and X2.  */
+static __always_inline op_t
+find_zero_ne_low (op_t x1, op_t x2)
+{
+  return (~find_zero_eq_low (x1, x2)) + 1;
+}
+
+static __always_inline op_t
+find_zero_ne_all (op_t x1, op_t x2)
+{
+  op_t m = repeat_bytes (0x7f);
+  op_t eq = x1 ^ x2;
+  op_t nz1 = ((x1 & m) + m) | x1;
+  op_t ne2 = ((eq & m) + m) | eq;
+  return (ne2 | ~nz1) & ~m;
+}
+
+#endif /* _STRING_FZA_H */
diff --git a/sysdeps/generic/string-fzb.h b/sysdeps/generic/string-fzb.h
new file mode 100644
index 0000000000..42de500d67
--- /dev/null
+++ b/sysdeps/generic/string-fzb.h
@@ -0,0 +1,49 @@ 
+/* Zero byte detection, boolean.  Generic C version.
+   Copyright (C) 2023 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#ifndef _STRING_FZB_H
+#define _STRING_FZB_H 1
+
+#include <endian.h>
+#include <string-fza.h>
+
+/* Determine if any byte within X is zero.  This is a pure boolean test.  */
+
+static __always_inline _Bool
+has_zero (op_t x)
+{
+  return find_zero_low (x) != 0;
+}
+
+/* Likewise, but for byte equality between X1 and X2.  */
+
+static __always_inline _Bool
+has_eq (op_t x1, op_t x2)
+{
+  return find_eq_low (x1, x2) != 0;
+}
+
+/* Likewise, but for zeros in X1 and equal bytes between X1 and X2.  */
+
+static __always_inline _Bool
+has_zero_eq (op_t x1, op_t x2)
+{
+  return find_zero_eq_low (x1, x2);
+}
+
+#endif /* _STRING_FZB_H */
diff --git a/sysdeps/generic/string-fzi.h b/sysdeps/generic/string-fzi.h
new file mode 100644
index 0000000000..b1fd4d34b3
--- /dev/null
+++ b/sysdeps/generic/string-fzi.h
@@ -0,0 +1,135 @@ 
+/* Zero byte detection; indexes.  Generic C version.
+   Copyright (C) 2023 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#ifndef _STRING_FZI_H
+#define _STRING_FZI_H 1
+
+#include <limits.h>
+#include <endian.h>
+#include <string-fza.h>
+
+static __always_inline int
+clz (op_t c)
+{
+  if (sizeof (op_t) == sizeof (unsigned long))
+    return __builtin_clzl (c);
+  else
+    return __builtin_clzll (c);
+}
+
+static __always_inline int
+ctz (op_t c)
+{
+  if (sizeof (op_t) == sizeof (unsigned long))
+    return __builtin_ctzl (c);
+  else
+    return __builtin_ctzll (c);
+}
+
+/* A subroutine for the index_zero functions.  Given a test word C, return
+   the (memory order) index of the first byte (in memory order) that is
+   non-zero.  */
+static __always_inline unsigned int
+index_first (op_t c)
+{
+  int r;
+  if (__BYTE_ORDER == __LITTLE_ENDIAN)
+    r = ctz (c);
+  else
+    r = clz (c);
+  return r / CHAR_BIT;
+}
+
+/* Similarly, but return the (memory order) index of the last byte that is
+   non-zero.  */
+static __always_inline unsigned int
+index_last (op_t c)
+{
+  int r;
+  if (__BYTE_ORDER == __LITTLE_ENDIAN)
+    r = clz (c);
+  else
+    r = ctz (c);
+  return sizeof (op_t) - 1 - (r / CHAR_BIT);
+}
+
+/* Given a word X that is known to contain a zero byte, return the index of
+   the first such within the word in memory order.  */
+static __always_inline unsigned int
+index_first_zero (op_t x)
+{
+  if (__BYTE_ORDER == __LITTLE_ENDIAN)
+    x = find_zero_low (x);
+  else
+    x = find_zero_all (x);
+  return index_first (x);
+}
+
+/* Similarly, but perform the search for byte equality between X1 and X2.  */
+static __always_inline unsigned int
+index_first_eq (op_t x1, op_t x2)
+{
+  if (__BYTE_ORDER == __LITTLE_ENDIAN)
+    x1 = find_eq_low (x1, x2);
+  else
+    x1 = find_eq_all (x1, x2);
+  return index_first (x1);
+}
+
+/* Similarly, but perform the search for zero within X1 or equality between
+   X1 and X2.  */
+static __always_inline unsigned int
+index_first_zero_eq (op_t x1, op_t x2)
+{
+  if (__BYTE_ORDER == __LITTLE_ENDIAN)
+    x1 = find_zero_eq_low (x1, x2);
+  else
+    x1 = find_zero_eq_all (x1, x2);
+  return index_first (x1);
+}
+
+/* Similarly, but perform the search for zero within X1 or inequality between
+   X1 and X2.  */
+static __always_inline unsigned int
+index_first_zero_ne (op_t x1, op_t x2)
+{
+  if (__BYTE_ORDER == __LITTLE_ENDIAN)
+    x1 = find_zero_ne_low (x1, x2);
+  else
+    x1 = find_zero_ne_all (x1, x2);
+  return index_first (x1);
+}
+
+/* Similarly, but search for the last zero within X.  */
+static __always_inline unsigned int
+index_last_zero (op_t x)
+{
+  if (__BYTE_ORDER == __LITTLE_ENDIAN)
+    x = find_zero_all (x);
+  else
+    x = find_zero_low (x);
+  return index_last (x);
+}
+
+static __always_inline unsigned int
+index_last_eq (op_t x1, op_t x2)
+{
+  return index_last_zero (x1 ^ x2);
+}
+
+#endif /* STRING_FZI_H */
diff --git a/sysdeps/generic/string-maskoff.h b/sysdeps/generic/string-maskoff.h
index 6ad4e0b6f9..16d3cc2ddb 100644
--- a/sysdeps/generic/string-maskoff.h
+++ b/sysdeps/generic/string-maskoff.h
@@ -23,7 +23,6 @@ 
 #include <limits.h>
 #include <stdint.h>
 #include <string-optype.h>
-#include <string-fza.h>
 
 /* Provide a mask based on the pointer alignment that sets up non-zero
    bytes before the beginning of the word.  It is used to mask off
@@ -44,12 +43,12 @@  create_mask (uintptr_t i)
 /* Return the mask WORD shifted based on S_INT address value, to ignore
    values not presented in the aligned word read.  */
 static __always_inline op_t
-check_mask (op_t word, uintptr_t s_int)
+check_mask (op_t word, uintptr_t s)
 {
   if (__BYTE_ORDER == __LITTLE_ENDIAN)
-    return word >> (CHAR_BIT * (s_int % sizeof (s_int)));
+    return word >> (CHAR_BIT * (s % sizeof (op_t)));
   else
-    return word << (CHAR_BIT * (s_int % sizeof (s_int)));
+    return word << (CHAR_BIT * (s % sizeof (op_t)));
 }
 
 /* Setup an word with each byte being c_in.  For instance, on a 64 bits
@@ -79,7 +78,7 @@  highbit_mask (op_t m)
 static __always_inline op_t *
 word_containing (char const *p)
 {
-  return (op_t *) ((uintptr_t) p & -sizeof(p));
+  return (op_t *) ((op_t) p & -sizeof(p));
 }
 
 #endif /* _STRING_MASKOFF_H  */