Speed up hex_encode with bytewise lookup

Previously, hex_encode looked up each nibble of the input
separately. We now use a larger lookup table containing the two-byte
encoding of every possible input byte, resulting in a 1/3 reduction
in encoding time.

Reviewed by Tom Lane, Michael Paquier, Nathan Bossart, David Rowley

Discussion: https://postgr.es/m/CANWCAZZvXuJMgqMN4u068Yqa19CEjS31tQKZp_qFFFbgYfaXqQ%40mail.gmail.com

diff --git a/src/backend/utils/adt/encode.c b/src/backend/utils/adt/encode.c
index 4a6fcb56cd951e5c364cea5a2b37f70022cc6416..4ccaed815d17e00194a86bda37fd7a0532ee3e6c 100644 (file)
--- a/src/backend/utils/adt/encode.c
+++ b/src/backend/utils/adt/encode.c
@@ -145,7 +145,26 @@ binary_decode(PG_FUNCTION_ARGS)
  * HEX
  */
 
-static const char hextbl[] = "0123456789abcdef";
+/*
+ * The hex expansion of each possible byte value (two chars per value).
+ */
+static const char hextbl[512] =
+"000102030405060708090a0b0c0d0e0f"
+"101112131415161718191a1b1c1d1e1f"
+"202122232425262728292a2b2c2d2e2f"
+"303132333435363738393a3b3c3d3e3f"
+"404142434445464748494a4b4c4d4e4f"
+"505152535455565758595a5b5c5d5e5f"
+"606162636465666768696a6b6c6d6e6f"
+"707172737475767778797a7b7c7d7e7f"
+"808182838485868788898a8b8c8d8e8f"
+"909192939495969798999a9b9c9d9e9f"
+"a0a1a2a3a4a5a6a7a8a9aaabacadaeaf"
+"b0b1b2b3b4b5b6b7b8b9babbbcbdbebf"
+"c0c1c2c3c4c5c6c7c8c9cacbcccdcecf"
+"d0d1d2d3d4d5d6d7d8d9dadbdcdddedf"
+"e0e1e2e3e4e5e6e7e8e9eaebecedeeef"
+"f0f1f2f3f4f5f6f7f8f9fafbfcfdfeff";
 
 static const int8 hexlookup[128] = {
    -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
@@ -165,9 +184,11 @@ hex_encode(const char *src, size_t len, char *dst)
 
    while (src < end)
    {
-       *dst++ = hextbl[(*src >> 4) & 0xF];
-       *dst++ = hextbl[*src & 0xF];
+       unsigned char usrc = *((const unsigned char *) src);
+
+       memcpy(dst, &hextbl[2 * usrc], 2);
        src++;
+       dst += 2;
    }
    return (uint64) len * 2;
 }