test: try to improve throughput

fuel-vm/src/interpreter/memory.rs

@@ -1040,17 +1040,37 @@ fn slice_eq(a: &[u8], b: &[u8]) -> bool {
             return false;
         }
 
-        let chunks = a.chunks_exact(16);
-        let remainder = chunks.remainder();
-        for (chunk_a, chunk_b) in chunks.zip(b.chunks_exact(16)) {
-            let simd_a = u8x16::from_slice(chunk_a);
-            let simd_b = u8x16::from_slice(chunk_b);
-            if simd_a != simd_b {
-                return false;
-            }
+        // because we're using simd, we can parallelize by a factor of 4
+        let mut res_a = false;
+        let mut res_b = false;
+        let mut res_c = false;
+        let mut res_d = false;
+
+        let chunks_a = a.chunks_exact(16);
+        let chunks_b = b.chunks_exact(16);
+        let remainder_a = chunks_a.remainder();
+        let remainder_b = chunks_b.remainder();
+
+        for i in (0..chunks_a.len()).step_by(4) {
+            let simd_a = u8x16::from_slice(&a[i..i + 16]);
+            let simd_b = u8x16::from_slice(&b[i..i + 16]);
+            res_a |= simd_a != simd_b;
+
+            let simd_a = u8x16::from_slice(&a[i + 16..i + 32]);
+            let simd_b = u8x16::from_slice(&b[i + 16..i + 32]);
+            res_b |= simd_a != simd_b;
+
+            let simd_a = u8x16::from_slice(&a[i + 32..i + 48]);
+            let simd_b = u8x16::from_slice(&b[i + 32..i + 48]);
+            res_c |= simd_a != simd_b;
+
+            let simd_a = u8x16::from_slice(&a[i + 48..i + 64]);
+            let simd_b = u8x16::from_slice(&b[i + 48..i + 64]);
+            res_d |= simd_a != simd_b;
         }
 
-        remainder == &b[a.len() - remainder.len()..]
+        // check the remainder
+        remainder_a == remainder_b && res_a && res_b && res_c && res_d
     }
     #[cfg(not(feature = "experimental"))]
     {