slice/ascii: Optimize eq_ignore_ascii_case with auto-vectorization

Refactor the current functionality into a helper function
Use `as_chunks` to encourage auto-vectorization in the optimized chunk processing function
Add a codegen test
Add benches for `eq_ignore_ascii_case`

The optimized function is initially only enabled for x86_64 which has `sse2` as
part of its baseline, but none of the code is platform specific. Other
platforms with SIMD instructions may also benefit from this implementation.

Performance improvements only manifest for slices of 16 bytes or longer, so the
optimized path is gated behind a length check for greater than or equal to 16.

Author: okaneco

library/core/src/slice/ascii.rs

@@ -60,6 +60,18 @@ impl [u8] {
             return false;
         }
 
+        #[cfg(all(target_arch = "x86_64", target_feature = "sse2"))]
+        if self.len() >= 16 {
+            return self.eq_ignore_ascii_case_chunks(other);
+        }
+
+        self.eq_ignore_ascii_case_simple(other)
+    }
+
+    /// ASCII case-insensitive equality check without chunk-at-a-time
+    /// optimization.
+    #[inline]
+    const fn eq_ignore_ascii_case_simple(&self, other: &[u8]) -> bool {
         // FIXME(const-hack): This implementation can be reverted when
         // `core::iter::zip` is allowed in const. The original implementation:
         //  self.len() == other.len() && iter::zip(self, other).all(|(a, b)| a.eq_ignore_ascii_case(b))
@@ -78,6 +90,37 @@ impl [u8] {
         true
     }
 
+    /// Optimized version of `eq_ignore_ascii_case` which processes chunks at a
+    /// time.
+    ///
+    /// Platforms that have SIMD instructions may benefit from this
+    /// implementation over `eq_ignore_ascii_case_simple`.
+    #[cfg(all(target_arch = "x86_64", target_feature = "sse2"))]
+    #[inline]
+    const fn eq_ignore_ascii_case_chunks(&self, other: &[u8]) -> bool {
+        const N: usize = 16;
+        let (a, a_rem) = self.as_chunks::<N>();
+        let (b, b_rem) = other.as_chunks::<N>();
+
+        let mut i = 0;
+        while i < a.len() && i < b.len() {
+            let mut equal_ascii = true;
+            let mut j = 0;
+            while j < N {
+                equal_ascii &= a[i][j].eq_ignore_ascii_case(&b[i][j]);
+                j += 1;
+            }
+
+            if !equal_ascii {
+                return false;
+            }
+
+            i += 1;
+        }
+
+        a_rem.eq_ignore_ascii_case_simple(b_rem)
+    }
+
     /// Converts this slice to its ASCII upper case equivalent in-place.
     ///
     /// ASCII letters 'a' to 'z' are mapped to 'A' to 'Z',

library/coretests/benches/ascii.rs

@@ -1,3 +1,4 @@
+mod eq_ignore_ascii_case;
 mod is_ascii;
 
 // Lower-case ASCII 'a' is the first byte that has its highest bit set

library/coretests/benches/ascii/eq_ignore_ascii_case.rs

@@ -0,0 +1,56 @@
+use test::Bencher;
+
+#[bench]
+fn bench_str_under_8_bytes_eq(b: &mut Bencher) {
+    let s = "foo";
+    let other = "FOo";
+    b.iter(|| {
+        assert!(s.eq_ignore_ascii_case(other));
+    })
+}
+
+#[bench]
+fn bench_str_of_8_bytes_eq(b: &mut Bencher) {
+    let s = "foobar78";
+    let other = "FOObAr78";
+    b.iter(|| {
+        assert!(s.eq_ignore_ascii_case(other));
+    })
+}
+
+#[bench]
+fn bench_str_17_bytes_eq(b: &mut Bencher) {
+    let s = "performance-criti";
+    let other = "performANce-cRIti";
+    b.iter(|| {
+        assert!(s.eq_ignore_ascii_case(other));
+    })
+}
+
+#[bench]
+fn bench_str_31_bytes_eq(b: &mut Bencher) {
+    let s = "foobarbazquux02foobarbazquux025";
+    let other = "fooBARbazQuuX02fooBARbazQuuX025";
+    b.iter(|| {
+        assert!(s.eq_ignore_ascii_case(other));
+    })
+}
+
+#[bench]
+fn bench_long_str_eq(b: &mut Bencher) {
+    let s = "Lorem ipsum dolor sit amet, consectetur adipisicing elit, sed do eiusmod tempor \
+             incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud \
+             exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. Duis aute \
+             irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla \
+             pariatur. Excepteur sint occaecat cupidatat non proident, sunt in culpa qui \
+             officia deserunt mollit anim id est laborum.";
+    let other = "Lorem ipsum dolor sit amet, CONSECTETUR adipisicing elit, sed do eiusmod tempor \
+             incididunt ut labore et dolore MAGNA aliqua. Ut enim ad MINIM veniam, quis nostrud \
+             exercitation ullamco LABORIS nisi ut aliquip ex ea commodo consequat. Duis aute \
+             irure dolor in reprehenderit in voluptate velit esse cillum DOLORE eu fugiat nulla \
+             pariatur. Excepteur sint occaecat CUPIDATAT non proident, sunt in culpa qui \
+             officia deserunt mollit anim id est laborum.";
+    b.iter(|| {
+        assert!(s.eq_ignore_ascii_case(other));
+    })
+}

tests/codegen-llvm/lib-optimizations/eq_ignore_ascii_case.rs

@@ -0,0 +1,14 @@
+//@ compile-flags: -Copt-level=3
+//@ only-x86_64
+#![crate_type = "lib"]
+
+// Ensure that the optimized variant of the function gets auto-vectorized.
+// CHECK-LABEL: @eq_ignore_ascii_case_autovectorized
+#[no_mangle]
+pub fn eq_ignore_ascii_case_autovectorized(s: &str, other: &str) -> bool {
+    // CHECK: load <16 x i8>
+    // CHECK: load <16 x i8>
+    // CHECK: bitcast <16 x i1>
+    // CHECK-NOT: panic
+    s.eq_ignore_ascii_case(other)
+}