Add 128 bit murmur3 hash function.

Signed-off-by: Reimar Döffinger <Reimar.Doeffinger@gmx.de>

1 files changed, 191 insertions, 0 deletions

diff --git a/libavutil/murmur3.c b/libavutil/murmur3.c
new file mode 100644
index 0000000000..cab9a98f93
--- /dev/null
+++ b/libavutil/murmur3.c
@@ -0,0 +1,191 @@
+/*
+ * Copyright (C) 2013 Reimar Döffinger <Reimar.Doeffinger@gmx.de>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg 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.
+ *
+ * FFmpeg 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 FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+#include <stdint.h>
+#include "mem.h"
+#include "intreadwrite.h"
+#include "murmur3.h"
+
+typedef struct AVMurMur3 {
+    uint64_t h1, h2;
+    uint8_t state[16];
+    int state_pos;
+    uint64_t len;
+} AVMurMur3;
+
+AVMurMur3 *av_murmur3_alloc(void)
+{
+    return av_mallocz(sizeof(AVMurMur3));
+}
+
+void av_murmur3_init_seeded(AVMurMur3 *c, uint64_t seed)
+{
+    memset(c, 0, sizeof(*c));
+    c->h1 = c->h2 = seed;
+}
+
+void av_murmur3_init(AVMurMur3 *c)
+{
+    // arbitrary random number as seed
+    av_murmur3_init_seeded(c, 0x725acc55daddca55);
+}
+
+static const uint64_t c1 = UINT64_C(0x87c37b91114253d5);
+static const uint64_t c2 = UINT64_C(0x4cf5ad432745937f);
+
+#define ROT(a, b) ((a << b) | (a >> (64 - b)))
+
+static uint64_t inline get_k1(const uint8_t *src)
+{
+    uint64_t k = AV_RL64(src);
+    k *= c1;
+    k = ROT(k, 31);
+    k *= c2;
+    return k;
+}
+
+static uint64_t inline get_k2(const uint8_t *src)
+{
+    uint64_t k = AV_RL64(src + 8);
+    k *= c2;
+    k = ROT(k, 33);
+    k *= c1;
+    return k;
+}
+
+static uint64_t inline update_h1(uint64_t k, uint64_t h1, uint64_t h2)
+{
+    k ^= h1;
+    k = ROT(k, 27);
+    k += h2;
+    k *= 5;
+    k += 0x52dce729;
+    return k;
+}
+
+static uint64_t inline update_h2(uint64_t k, uint64_t h1, uint64_t h2)
+{
+    k ^= h2;
+    k = ROT(k, 31);
+    k += h1;
+    k *= 5;
+    k += 0x38495ab5;
+    return k;
+}
+
+void av_murmur3_update(AVMurMur3 *c, const uint8_t *src, int len)
+{
+    const uint8_t *end;
+    uint64_t h1 = c->h1, h2 = c->h2;
+    uint64_t k1, k2;
+    if (len <= 0) return;
+    c->len += len;
+    if (c->state_pos > 0) {
+        while (c->state_pos < 16) {
+            c->state[c->state_pos++] = *src++;
+            if (--len <= 0) return;
+        }
+        c->state_pos = 0;
+        k1 = get_k1(c->state);
+        k2 = get_k2(c->state);
+        h1 = update_h1(k1, h1, h2);
+        h2 = update_h2(k2, h1, h2);
+    }
+
+    end = src + (len & ~15);
+    while (src < end) {
+        // These could be done sequentially instead
+        // of interleaved, but like this is over 10% faster
+        k1 = get_k1(src);
+        k2 = get_k2(src);
+        h1 = update_h1(k1, h1, h2);
+        h2 = update_h2(k2, h1, h2);
+        src += 16;
+    }
+    c->h1 = h1;
+    c->h2 = h2;
+
+    len &= 15;
+    if (len > 0) {
+        memcpy(c->state, src, len);
+        c->state_pos = len;
+    }
+}
+
+static inline uint64_t fmix(uint64_t k)
+{
+    k ^= k >> 33;
+    k *= UINT64_C(0xff51afd7ed558ccd);
+    k ^= k >> 33;
+    k *= UINT64_C(0xc4ceb9fe1a85ec53);
+    k ^= k >> 33;
+    return k;
+}
+
+void av_murmur3_final(AVMurMur3 *c, uint8_t dst[16])
+{
+    uint64_t h1 = c->h1, h2 = c->h2;
+    memset(c->state + c->state_pos, 0, sizeof(c->state) - c->state_pos);
+    h1 ^= get_k1(c->state) ^ c->len;
+    h2 ^= get_k2(c->state) ^ c->len;
+    h1 += h2;
+    h2 += h1;
+    h1 = fmix(h1);
+    h2 = fmix(h2);
+    h1 += h2;
+    h2 += h1;
+    AV_WL64(dst, h1);
+    AV_WL64(dst + 8, h2);
+}
+
+#ifdef TEST
+int main(void)
+{
+    int i;
+    uint8_t hash_result[16] = {0};
+    AVMurMur3 *ctx = av_murmur3_alloc();
+#if 1
+    uint8_t in[256] = {0};
+    uint8_t *hashes = av_mallocz(256 * 16);
+    for (i = 0; i < 256; i++)
+    {
+        in[i] = i;
+        av_murmur3_init_seeded(ctx, 256 - i);
+        // Note: this actually tests hashing 0 bytes
+        av_murmur3_update(ctx, in, i);
+        av_murmur3_final(ctx, hashes + 16 * i);
+    }
+    av_murmur3_init_seeded(ctx, 0);
+    av_murmur3_update(ctx, hashes, 256 * 16);
+    av_murmur3_final(ctx, hash_result);
+    av_free(hashes);
+    printf("result: 0x%"PRIx64" 0x%"PRIx64"\n", AV_RL64(hash_result), AV_RL64(hash_result + 8));
+    // official reference value is 32 bit
+    return AV_RL32(hash_result) != 0x6384ba69;
+#else
+    uint8_t *in = av_mallocz(512*1024);
+    av_murmur3_init(ctx);
+    for (i = 0; i < 40*1024; i++)
+        av_murmur3_update(ctx, in, 512*1024);
+    av_murmur3_final(ctx, hash_result);
+    av_free(in);
+    return hash_result[0];
+#endif
+}
+#endif