Allow user to provide a pre-hashed value

This is ~10% faster for u64 lookups. For lookups and construction the
time is constant regardless of the true length of the input (assuming
you can amortize the hashing cost somehow externally to this library).

For example, for a 128 byte string construction is ~2.5x faster and
lookups are ~1.7x faster.

Author: vlovich

Cargo.toml

@@ -30,3 +30,7 @@ harness = false
 [features]
 default = ["parallel"]
 parallel = ["rayon", "crossbeam-utils"]
+
+[profile.release]
+lto = true
+debug = 2

benches/build.rs

@@ -4,7 +4,7 @@ extern crate bencher;
 
 use bencher::Bencher;
 
-use boomphf::Mphf;
+use boomphf::{ExternallyHashed, Mphf};
 
 fn build1_ser_u64(bench: &mut Bencher) {
     let items: Vec<u64> = (0..1000000u64).map(|x| x * 2).collect();
@@ -13,13 +13,48 @@ fn build1_ser_u64(bench: &mut Bencher) {
     });
 }
 
+fn build1_ser_externally_hashed(bench: &mut Bencher) {
+    let items: Vec<ExternallyHashed> = (0..1000000u64)
+        .map(|x| ExternallyHashed(wyhash::wyrng(&mut (x * 2))))
+        .collect();
+    bench.iter(|| {
+        std::hint::black_box(Mphf::new(2.0, &items));
+    });
+}
+
 fn build1_ser_slices(bench: &mut Bencher) {
     let items: Vec<[u8; 8]> = (0..1000000u64).map(|x| (x * 2).to_le_bytes()).collect();
     bench.iter(|| {
         std::hint::black_box(Mphf::new(2.0, &items));
     });
 }
 
+fn build1_ser_long_slices(bench: &mut Bencher) {
+    let items = (0..1000000u64)
+        .map(|x| {
+            let mut long_key = [0u8; 128];
+            long_key[0..8].copy_from_slice(&(x * 2).to_le_bytes());
+            long_key
+        })
+        .collect::<Vec<_>>();
+    bench.iter(|| {
+        std::hint::black_box(Mphf::new(2.0, &items));
+    });
+}
+
+fn build1_ser_long_slices_externally_hashed(bench: &mut Bencher) {
+    let items = (0..1000000u64)
+        .map(|x| {
+            let mut long_key = [0u8; 128];
+            long_key[0..8].copy_from_slice(&(x * 2).to_le_bytes());
+            ExternallyHashed(wyhash::wyhash(&long_key, 0))
+        })
+        .collect::<Vec<_>>();
+    bench.iter(|| {
+        std::hint::black_box(Mphf::new(2.0, &items));
+    });
+}
+
 #[allow(dead_code)]
 fn build1_par_u64(bench: &mut Bencher) {
     let items: Vec<u64> = (0..1000000u64).map(|x| x * 2).collect();
@@ -38,16 +73,88 @@ fn build1_par_slices(bench: &mut Bencher) {
     });
 }
 
-fn scan1_ser(bench: &mut Bencher) {
+fn scan1_ser_u64(bench: &mut Bencher) {
     let items: Vec<u64> = (0..1000000u64).map(|x| x * 2).collect();
     let phf = Mphf::new(2.0, &items);
 
     bench.iter(|| {
-        for i in (0..1000000u64).map(|x| x * 2) {
+        for i in &items {
             std::hint::black_box(phf.hash(&i));
         }
     });
 }
 
-benchmark_group!(benches, build1_ser_u64, build1_ser_slices, build1_par_u64, build1_par_slices, scan1_ser);
+fn scan1_ser_slice(bench: &mut Bencher) {
+    let items: Vec<[u8; 8]> = (0..1000000u64).map(|x| (x * 2).to_le_bytes()).collect();
+    let phf = Mphf::new(2.0, &items);
+
+    bench.iter(|| {
+        for i in &items {
+            std::hint::black_box(phf.hash(i));
+        }
+    });
+}
+
+fn scan1_ser_externally_hashed(bench: &mut Bencher) {
+    let items: Vec<ExternallyHashed> = (0..1000000u64)
+        .map(|x| ExternallyHashed(wyhash::wyrng(&mut (x * 2))))
+        .collect();
+    let phf = Mphf::new(2.0, &items);
+
+    bench.iter(|| {
+        for i in &items {
+            std::hint::black_box(phf.hash(i));
+        }
+    });
+}
+
+fn scan1_ser_long_key(bench: &mut Bencher) {
+    let items = (0..1000000u64)
+        .map(|x| {
+            let mut long_key = [0u8; 128];
+            long_key[0..8].copy_from_slice(&(x * 2).to_le_bytes());
+            long_key
+        })
+        .collect::<Vec<_>>();
+    let phf = Mphf::new(2.0, &items);
+
+    bench.iter(|| {
+        for i in &items {
+            std::hint::black_box(phf.hash(i));
+        }
+    });
+}
+
+fn scan1_ser_long_key_externally_hashed(bench: &mut Bencher) {
+    let items: Vec<ExternallyHashed> = (0..1000000u64)
+        .map(|x| {
+            let mut long_key = [0u8; 128];
+            long_key[0..8].copy_from_slice(&(x * 2).to_le_bytes());
+            ExternallyHashed(wyhash::wyhash(&long_key, 0))
+        })
+        .collect();
+    let phf = Mphf::new(2.0, &items);
+
+    bench.iter(|| {
+        for i in &items {
+            std::hint::black_box(phf.hash(i));
+        }
+    });
+}
+
+benchmark_group!(
+    benches,
+    build1_ser_externally_hashed,
+    build1_ser_u64,
+    build1_ser_slices,
+    build1_ser_long_slices,
+    build1_ser_long_slices_externally_hashed,
+    build1_par_u64,
+    build1_par_slices,
+    scan1_ser_u64,
+    scan1_ser_slice,
+    scan1_ser_externally_hashed,
+    scan1_ser_long_key,
+    scan1_ser_long_key_externally_hashed
+);
 benchmark_main!(benches);

src/bitvector.rs

@@ -363,7 +363,7 @@ impl BitVector {
     #[inline]
     pub fn get_word(&self, word: usize) -> u64 {
         #[cfg(feature = "parallel")]
-        return self.vector[word].load(Ordering::Relaxed) as u64;
+        return self.vector[word].load(Ordering::Relaxed);
 
         #[cfg(not(feature = "parallel"))]
         return self.vector[word] as u64;