Remove final cfg gating where preloading can also benefit from CPU prefetching

Author: tibvdm

sa-index/src/array/mod.rs

@@ -85,14 +85,30 @@ impl SuffixArray {
     }
 
     /// Issues a non-blocking hardware prefetch hint for the cache line holding SA entry `index`.
-    #[cfg(feature = "mmap")]
     #[inline]
     pub fn prefetch_sa_index(&self, index: usize) {
-        if let SuffixArray::MmapBacked { mmap, data_offset, bits_per_value, .. } = self {
-            let byte_offset = data_offset + (index * bits_per_value) / 8;
-            if byte_offset < mmap.len() {
-                let ptr: *const u8 = &mmap[byte_offset];
-                prefetch::prefetch_read(ptr);
+        match self {
+            SuffixArray::Original(sa, _) => {
+                if index < sa.len() {
+                    prefetch::prefetch_read(unsafe { sa.as_ptr().add(index) as *const u8 });
+                }
+            }
+            SuffixArray::Compressed(ba, _) => {
+                let bpv = ba.bits_per_value();
+                let word_idx = (index * bpv) / 64;
+                let total_words = (ba.len() * bpv + 63) / 64;
+                if word_idx < total_words {
+                    let ptr = ba.get_data_slice(word_idx, word_idx + 1).as_ptr() as *const u8;
+                    prefetch::prefetch_read(ptr);
+                }
+            }
+            #[cfg(feature = "mmap")]
+            SuffixArray::MmapBacked { mmap, data_offset, bits_per_value, .. } => {
+                let byte_offset = data_offset + (index * bits_per_value) / 8;
+                if byte_offset < mmap.len() {
+                    let ptr: *const u8 = &mmap[byte_offset];
+                    prefetch::prefetch_read(ptr);
+                }
             }
         }
     }

sa-index/src/sa_searcher.rs

@@ -597,11 +597,8 @@ impl Searcher {
     /// call. One fetch will be wasted; both are free (single non-blocking CPU instruction).
     /// From iteration 2 onward the needed SA entry is already in L1/L2 cache.
     #[inline]
-    #[cfg_attr(not(feature = "mmap"), allow(unused_variables))]
     fn prefetch_binary_search_pivots(&self, lo: usize, center: usize, hi: usize) {
-        #[cfg(feature = "mmap")]
         self.sa.prefetch_sa_index((lo + center) / 2);
-        #[cfg(feature = "mmap")]
         self.sa.prefetch_sa_index((center + hi) / 2);
     }