v18-0004-Check-ascii-8-bytes-at-a-time-with-bitwise-opera.patch

From 6360898811851d0b5484e6aab7219cd7379c20e1 Mon Sep 17 00:00:00 2001
From: John Naylor <john.naylor@2ndquadrant.com>
Date: Sun, 18 Jul 2021 19:11:25 -0400
Subject: [PATCH v18 4/6] Check ascii 8-bytes at a time with bitwise operations

---
 src/common/wchar.c                       | 53 ++++++++++++++++++++----
 src/test/regress/expected/conversion.out |  2 +-
 src/test/regress/sql/conversion.sql      |  2 +-
 3 files changed, 48 insertions(+), 9 deletions(-)

diff --git a/src/common/wchar.c b/src/common/wchar.c
index 4ea352bcf1..bba54912f9 100644
--- a/src/common/wchar.c
+++ b/src/common/wchar.c
@@ -1932,24 +1932,63 @@ utf8_advance(const unsigned char *s)
 	return l;
 }
 
+/* Verify a chunk of bytes for valid ASCII including a zero-byte check. */
+static inline int
+check_ascii(const unsigned char *s, int len)
+{
+	uint64		chunk,
+				highbits_set,
+				highbit_carry;
+
+	if (len >= sizeof(uint64))
+	{
+		memcpy(&chunk, s, sizeof(uint64));
+
+		/* Check if any bytes in this chunk have the high bit set. */
+		highbits_set = chunk & UINT64CONST(0x8080808080808080);
+		if (highbits_set)
+			return 0;
+
+		/*
+		 * Check if there are any zero bytes in this chunk.
+		 *
+		 * First, add 0x7f to each byte. This sets the high bit in each byte,
+		 * unless it was a zero. We already checked that none of the bytes had
+		 * the high bit set previously, so the max value each byte can have
+		 * after the addition is 0x7f + 0x7f = 0xfe, and we don't need to
+		 * worry about carrying over to the next byte.
+		 */
+		highbit_carry = chunk + UINT64CONST(0x7f7f7f7f7f7f7f7f);
+
+		/* Then check that the high bit is set in each byte. */
+		highbit_carry &= UINT64CONST(0x8080808080808080);
+		if (highbit_carry == UINT64CONST(0x8080808080808080))
+			return sizeof(uint64);
+		else
+			return 0;
+	}
+	else
+		return 0;
+}
+
 static int
 pg_utf8_verifystr(const unsigned char *s, int len)
 {
 	const unsigned char *start = s;
 
 	/*
-	 * fast path when we have enough bytes left in the string to cover all
-	 * valid UTF-8 sequences
+	 * fast path when we have enough bytes left in the string to use bitwise operations
 	 */
-	while (len >= 4)
+	while (len >= 8)
 	{
 		int			l;
 
-		/* check if the first byte is both non-zero and doesn't have the high bit set */
-		if ((signed char) (*s) > 0)
+		/* fast path for ASCII-subset characters */
+		l = check_ascii(s, 8);
+		if (l)
 		{
-			s++;
-			len--;
+			s += l;
+			len -= l;
 			continue;
 		}
 
diff --git a/src/test/regress/expected/conversion.out b/src/test/regress/expected/conversion.out
index e4ab9fe765..62461063b3 100644
--- a/src/test/regress/expected/conversion.out
+++ b/src/test/regress/expected/conversion.out
@@ -141,7 +141,7 @@ with test_bytes as (
 ), test_padded as (
   select
     description,
-    (test_conv(inbytes || repeat('.', 3)::bytea, 'utf8', 'utf8')).error
+    (test_conv(inbytes || repeat('.', 8)::bytea, 'utf8', 'utf8')).error
   from test_bytes
 )
 select
diff --git a/src/test/regress/sql/conversion.sql b/src/test/regress/sql/conversion.sql
index e5a7e47958..5d0280cd1c 100644
--- a/src/test/regress/sql/conversion.sql
+++ b/src/test/regress/sql/conversion.sql
@@ -118,7 +118,7 @@ with test_bytes as (
 ), test_padded as (
   select
     description,
-    (test_conv(inbytes || repeat('.', 3)::bytea, 'utf8', 'utf8')).error
+    (test_conv(inbytes || repeat('.', 8)::bytea, 'utf8', 'utf8')).error
   from test_bytes
 )
 select
-- 
2.31.1