Use SSE2 SIMD in IndexTable::find_entry

src/index.rs

@@ -11,7 +11,11 @@ use crate::{
 	table::{key::TableKey, SIZE_TIERS_BITS},
 	Key,
 };
-use std::convert::TryInto;
+#[cfg(target_arch = "x86")]
+use std::arch::x86::*;
+#[cfg(target_arch = "x86_64")]
+use std::arch::x86_64::*;
+use std::{cmp::max, convert::TryInto};
 
 // Index chunk consists of 8 64-bit entries.
 const CHUNK_LEN: usize = CHUNK_ENTRIES * ENTRY_BYTES; // 512 bytes
@@ -231,8 +235,56 @@ impl IndexTable {
 		Ok(try_io!(Ok(&map[offset..offset + CHUNK_LEN])))
 	}
 
-	#[inline(never)]
 	fn find_entry(&self, key_prefix: u64, sub_index: usize, chunk: &[u8]) -> (Entry, usize) {
+		if cfg!(target_feature = "sse2") {
+			self.find_entry_sse2(key_prefix, sub_index, chunk)
+		} else {
+			self.find_entry_base(key_prefix, sub_index, chunk)
+		}
+	}
+
+	#[cfg(target_feature = "sse2")]
+	fn find_entry_sse2(&self, key_prefix: u64, sub_index: usize, chunk: &[u8]) -> (Entry, usize) {
+		assert!(chunk.len() >= CHUNK_ENTRIES * 8); // Bound checking (not done by SIMD instructions)
+		const _: () = assert!(
+			CHUNK_ENTRIES % 4 == 0,
+			"We assume here we got buffer with a number of elements that is a multiple of 4"
+		);
+
+		let shift = max(32, Entry::address_bits(self.id.index_bits()));
+		unsafe {
+			let target = _mm_set1_epi32(((key_prefix << self.id.index_bits()) >> shift) as i32);
+			let shift_mask = _mm_set_epi64x(0, shift.into());
+			let mut i = (sub_index >> 2) << 2; // We keep an alignment of 4
+			while i + 4 <= CHUNK_ENTRIES {
+				// We load the value 2 by 2
+				// Then we remove the address by shifting such that the partial key is in the low
+				// part
+				let first_two = _mm_shuffle_epi32::<0b11011000>(_mm_srl_epi64(
+					_mm_loadu_si128(chunk[i * 8..].as_ptr() as *const __m128i),
+					shift_mask,
+				));
+				let last_two = _mm_shuffle_epi32::<0b11011000>(_mm_srl_epi64(
+					_mm_loadu_si128(chunk[(i + 2) * 8..].as_ptr() as *const __m128i),
+					shift_mask,
+				));
+				// We set into current the input low parts
+				let current = _mm_unpacklo_epi64(first_two, last_two);
+				let cmp = _mm_movemask_epi8(_mm_cmpeq_epi32(current, target));
+				if cmp != 0 {
+					let position = i + (cmp.trailing_zeros() as usize) / 4;
+					if position >= sub_index {
+						// We need to check we are not reading again the same input
+						return (Self::read_entry(chunk, position), position)
+					}
+				}
+				i += 4;
+			}
+		}
+		(Entry::empty(), 0)
+	}
+
+	fn find_entry_base(&self, key_prefix: u64, sub_index: usize, chunk: &[u8]) -> (Entry, usize) {
 		assert!(chunk.len() >= CHUNK_ENTRIES * 8);
 		let partial_key = Entry::extract_key(key_prefix, self.id.index_bits());
 		for i in sub_index..CHUNK_ENTRIES {
@@ -532,6 +584,7 @@ impl IndexTable {
 #[cfg(test)]
 mod test {
 	use super::*;
+	use std::path::PathBuf;
 
 	#[test]
 	fn test_entries() {
@@ -552,4 +605,37 @@ mod test {
 
 		assert!(IndexTable::transmute_chunk(chunk2) == chunk);
 	}
+
+	#[test]
+	fn test_find_entries() {
+		let partial_keys = [1, 1 << 10, 1 << 20];
+		for index_bits in [16, 18, 20, 22] {
+			let index_table = IndexTable {
+				id: TableId(index_bits.into()),
+				map: RwLock::new(None),
+				path: PathBuf::new(),
+			};
+
+			let data_address = Address::from_u64((1 << index_bits) - 1);
+
+			let mut chunk = [0; CHUNK_ENTRIES * 8];
+			for (i, partial_key) in partial_keys.iter().enumerate() {
+				chunk[i * 8..(i + 1) * 8].copy_from_slice(
+					&Entry::new(data_address, *partial_key, index_bits).as_u64().to_le_bytes(),
+				);
+			}
+
+			for partial_key in &partial_keys {
+				let key_prefix = *partial_key << (CHUNK_ENTRIES_BITS + SIZE_TIERS_BITS);
+				assert_eq!(
+					index_table.find_entry_sse2(key_prefix, 0, &chunk).0.partial_key(index_bits),
+					*partial_key
+				);
+				assert_eq!(
+					index_table.find_entry_base(key_prefix, 0, &chunk).0.partial_key(index_bits),
+					*partial_key
+				);
+			}
+		}
+	}
 }