SL-19110 Fast hashing classes for use in place of the slow LLMD5, where speed matters. (#64)

This commit adds the HBXX64 and HBXX128 classes for use as a drop-in
replacement for the slow LLMD5 hashing class, where speed matters and
backward compatibility (with standard hashing algorithms) and/or
cryptographic hashing qualities are not required.
It also replaces LLMD5 with HBXX* in a few existing hot (well, ok, just
"warm" for some) paths meeting the above requirements, while paving the way for
future use cases, such as in the DRTVWR-559 and sibling branches where the slow
LLMD5 is used (e.g. to hash materials and vertex buffer cache entries), and
could be use such a (way) faster algorithm with very significant benefits and
no negative impact.

Here is the comment I added in indra/llcommon/hbxx.h:

// HBXXH* classes are to be used where speed matters and cryptographic quality
// is not required (no "one-way" guarantee, though they are likely not worst in
// this respect than MD5 which got busted and is now considered too weak). The
// xxHash code they are built upon is vectorized and about 50 times faster than
// MD5. A 64 bits hash class is also provided for when 128 bits of entropy are
// not needed. The hashes collision rate is similar to MD5's.
// See https://github.com/Cyan4973/xxHash#readme for details.

1 files changed, 259 insertions, 0 deletions

diff --git a/indra/llcommon/hbxxh.h b/indra/llcommon/hbxxh.h
new file mode 100644
index 0000000000..8a5f977648
--- /dev/null
+++ b/indra/llcommon/hbxxh.h
@@ -0,0 +1,259 @@
+/**
+ * @file hbxxh.h
+ * @brief High performances vectorized hashing based on xxHash.
+ *
+ * $LicenseInfo:firstyear=2023&license=viewergpl$
+ * Second Life Viewer Source Code
+ * Copyright (c) 2023, Henri Beauchamp.
+ *
+ * This 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;
+ * version 2.1 of the License only.
+ * 
+ * This 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 this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
+ * 
+ * Linden Research, Inc., 945 Battery Street, San Francisco, CA  94111  USA
+ * $/LicenseInfo$
+ */
+
+#ifndef LL_HBXXH_H
+#define LL_HBXXH_H
+
+#include "lluuid.h"
+
+// HBXXH* classes are to be used where speed matters and cryptographic quality
+// is not required (no "one-way" guarantee, though they are likely not worst in
+// this respect than MD5 which got busted and is now considered too weak). The
+// xxHash code they are built upon is vectorized and about 50 times faster than
+// MD5. A 64 bits hash class is also provided for when 128 bits of entropy are
+// not needed. The hashes collision rate is similar to MD5's.
+// See https://github.com/Cyan4973/xxHash#readme for details.
+
+// 64 bits hashing class
+
+class HBXXH64
+{
+    friend std::ostream& operator<<(std::ostream&, HBXXH64);
+
+protected:
+    LOG_CLASS(HBXXH64);
+
+public:
+    inline HBXXH64()                            { init(); }
+
+    // Constructors for special circumstances; they all digest the first passed
+    // parameter. Set 'do_finalize' to false if you do not want to finalize the
+    // context, which is useful/needed when you want to update() it afterwards.
+    // Ideally, the compiler should be smart enough to get our clue and
+    // optimize out the const bool test during inlining...
+
+    inline HBXXH64(const void* buffer, size_t len,
+                   const bool do_finalize = true)
+    {
+        init();
+        update(buffer, len);
+        if (do_finalize)
+        {
+            finalize();
+        }
+    }
+
+    inline HBXXH64(const std::string& str, const bool do_finalize = true)
+    {
+        init();
+        update(str);
+        if (do_finalize)
+        {
+            finalize();
+        }
+    }
+
+    inline HBXXH64(std::istream& s, const bool do_finalize = true)
+    {
+        init();
+        update(s);
+        if (do_finalize)
+        {
+            finalize();
+        }
+    }
+
+    inline HBXXH64(FILE* file, const bool do_finalize = true)
+    {
+        init();
+        update(file);
+        if (do_finalize)
+        {
+            finalize();
+        }
+    }
+
+    ~HBXXH64();
+
+    void update(const void* buffer, size_t len);
+    void update(const std::string& str);
+    void update(std::istream& s);
+    void update(FILE* file);
+
+    // Note that unlike what happens with LLMD5, you do not need to finalize()
+    // HBXXH64 before using digest(), and you may keep updating() it even after
+    // you got a first digest() (the next digest would of course change after
+    // any update). It is still useful to use finalize() when you do not want
+    // to store a final digest() result in a separate U64; after this method
+    // has been called, digest() simply returns mDigest value.
+    void finalize();
+
+    U64 digest() const;
+
+    // Fast static methods. Use them when hashing just one contiguous block of
+    // data.
+    static U64 digest(const void* buffer, size_t len);
+    static U64 digest(const char* str);    // str must be NUL-terminated
+    static U64 digest(const std::string& str);
+
+private:
+    void init();
+
+private:
+    // We use a void pointer to avoid including xxhash.h here for XXH3_state_t
+    // (which cannot either be trivially forward-declared, due to complex API
+    // related pre-processor macros in xxhash.h).
+    void*   mState;
+    U64     mDigest;
+};
+
+inline bool operator==(const HBXXH64& a, const HBXXH64& b)
+{
+    return a.digest() == b.digest();
+}
+
+inline bool operator!=(const HBXXH64& a, const HBXXH64& b)
+{
+    return a.digest() != b.digest();
+}
+
+// 128 bits hashing class
+
+class HBXXH128
+{
+    friend std::ostream& operator<<(std::ostream&, HBXXH128);
+
+protected:
+    LOG_CLASS(HBXXH128);
+
+public:
+    inline HBXXH128()                           { init(); }
+
+    // Constructors for special circumstances; they all digest the first passed
+    // parameter. Set 'do_finalize' to false if you do not want to finalize the
+    // context, which is useful/needed when you want to update() it afterwards.
+    // Ideally, the compiler should be smart enough to get our clue and
+    // optimize out the const bool test during inlining...
+
+    inline HBXXH128(const void* buffer, size_t len,
+                    const bool do_finalize = true)
+    {
+        init();
+        update(buffer, len);
+        if (do_finalize)
+        {
+            finalize();
+        }
+    }
+
+    inline HBXXH128(const std::string& str, const bool do_finalize = true)
+    {
+        init();
+        update(str);
+        if (do_finalize)
+        {
+            finalize();
+        }
+    }
+
+    inline HBXXH128(std::istream& s, const bool do_finalize = true)
+    {
+        init();
+        update(s);
+        if (do_finalize)
+        {
+            finalize();
+        }
+    }
+
+    inline HBXXH128(FILE* file, const bool do_finalize = true)
+    {
+        init();
+        update(file);
+        if (do_finalize)
+        {
+            finalize();
+        }
+    }
+
+    ~HBXXH128();
+
+    void update(const void* buffer, size_t len);
+    void update(const std::string& str);
+    void update(std::istream& s);
+    void update(FILE* file);
+
+    // Note that unlike what happens with LLMD5, you do not need to finalize()
+    // HBXXH128 before using digest(), and you may keep updating() it even
+    // after you got a first digest() (the next digest would of course change
+    // after any update). It is still useful to use finalize() when you do not
+    // want to store a final digest() result in a separate LLUUID; after this
+    // method has been called, digest() simply returns a reference on mDigest.
+    void finalize();
+
+    // We use an LLUUID for the digest, since this is a 128 bits wide native
+    // type available in the viewer code, making it easy to manipulate. It also
+    // allows to use HBXXH128 efficiently in LLUUID generate() and combine()
+    // methods.
+    const LLUUID& digest() const;
+
+    // Here, we avoid an LLUUID copy whenever we already got one to store the
+    // result *and* we did not yet call finalize().
+    void digest(LLUUID& result) const;
+
+    // Fast static methods. Use them when hashing just one contiguous block of
+    // data.
+    static LLUUID digest(const void* buffer, size_t len);
+    static LLUUID digest(const char* str);    // str must be NUL-terminated
+    static LLUUID digest(const std::string& str);
+    // Same as above, but saves you from an LLUUID copy when you already got
+    // one for storage use.
+    static void digest(LLUUID& result, const void* buffer, size_t len);
+    static void digest(LLUUID& result, const char* str); // str NUL-terminated
+    static void digest(LLUUID& result, const std::string& str);
+
+private:
+    void init();
+
+private:
+    // We use a void pointer to avoid including xxhash.h here for XXH3_state_t
+    // (which cannot either be trivially forward-declared, due to complex API
+    // related pre-processor macros in xxhash.h).
+    void*   mState;
+    LLUUID  mDigest;
+};
+
+inline bool operator==(const HBXXH128& a, const HBXXH128& b)
+{
+    return a.digest() == b.digest();
+}
+
+inline bool operator!=(const HBXXH128& a, const HBXXH128& b)
+{
+    return a.digest() != b.digest();
+}
+
+#endif // LL_HBXXH_H