Optimize is_ascii

Generalize the x86-64+sse2 version of `is_ascii` to be architecture-neutral, and
rewrite it using `slice::as_chunks`. The new version is both shorter (in terms
of Rust source code) and smaller (in terms of produced assembly).

Compare the assembly generated before and after:
https://godbolt.org/z/MWKdnaYoK

Author: Kmeakin

library/core/src/slice/ascii.rs

@@ -3,7 +3,6 @@
 use core::ascii::EscapeDefault;
 
 use crate::fmt::{self, Write};
-#[cfg(not(all(target_arch = "x86_64", target_feature = "sse2")))]
 use crate::intrinsics::const_eval_select;
 use crate::{ascii, iter, ops};
 
@@ -327,175 +326,52 @@ impl<'a> fmt::Debug for EscapeAscii<'a> {
     }
 }
 
-/// ASCII test *without* the chunk-at-a-time optimizations.
-///
-/// This is carefully structured to produce nice small code -- it's smaller in
-/// `-O` than what the "obvious" ways produces under `-C opt-level=s`.  If you
-/// touch it, be sure to run (and update if needed) the assembly test.
-#[unstable(feature = "str_internals", issue = "none")]
-#[doc(hidden)]
 #[inline]
-pub const fn is_ascii_simple(mut bytes: &[u8]) -> bool {
-    while let [rest @ .., last] = bytes {
-        if !last.is_ascii() {
+const fn is_ascii_const(mut bytes: &[u8]) -> bool {
+    while let [first, rest @ ..] = bytes {
+        if !first.is_ascii() {
             break;
         }
         bytes = rest;
     }
     bytes.is_empty()
 }
 
+/// The implementation using iterators produces a tighter loop than the
+/// implementation using pattern-matching when inlined into `is_ascii_chunked`.
+/// So we have duplicate implementations of the scalar case until iterators are
+/// usable in const contexts.
+#[inline(always)]
+fn is_ascii_scalar(bytes: &[u8]) -> bool {
+    bytes.iter().all(u8::is_ascii)
+}
+
 /// Optimized ASCII test that will use usize-at-a-time operations instead of
 /// byte-at-a-time operations (when possible).
-///
-/// The algorithm we use here is pretty simple. If `s` is too short, we just
-/// check each byte and be done with it. Otherwise:
-///
-/// - Read the first word with an unaligned load.
-/// - Align the pointer, read subsequent words until end with aligned loads.
-/// - Read the last `usize` from `s` with an unaligned load.
-///
-/// If any of these loads produces something for which `contains_nonascii`
-/// (above) returns true, then we know the answer is false.
-#[cfg(not(all(target_arch = "x86_64", target_feature = "sse2")))]
 #[inline]
 #[rustc_allow_const_fn_unstable(const_eval_select)] // fallback impl has same behavior
-const fn is_ascii(s: &[u8]) -> bool {
+const fn is_ascii(bytes: &[u8]) -> bool {
     // The runtime version behaves the same as the compiletime version, it's
     // just more optimized.
     const_eval_select!(
-        @capture { s: &[u8] } -> bool:
+        @capture { bytes: &[u8] } -> bool:
         if const {
-            is_ascii_simple(s)
+            is_ascii_const(bytes)
         } else {
-            /// Returns `true` if any byte in the word `v` is nonascii (>= 128). Snarfed
-            /// from `../str/mod.rs`, which does something similar for utf8 validation.
-            const fn contains_nonascii(v: usize) -> bool {
-                const NONASCII_MASK: usize = usize::repeat_u8(0x80);
-                (NONASCII_MASK & v) != 0
-            }
-
-            const USIZE_SIZE: usize = size_of::<usize>();
-
-            let len = s.len();
-            let align_offset = s.as_ptr().align_offset(USIZE_SIZE);
-
-            // If we wouldn't gain anything from the word-at-a-time implementation, fall
-            // back to a scalar loop.
-            //
-            // We also do this for architectures where `size_of::<usize>()` isn't
-            // sufficient alignment for `usize`, because it's a weird edge case.
-            if len < USIZE_SIZE || len < align_offset || USIZE_SIZE < align_of::<usize>() {
-                return is_ascii_simple(s);
-            }
-
-            // We always read the first word unaligned, which means `align_offset` is
-            // 0, we'd read the same value again for the aligned read.
-            let offset_to_aligned = if align_offset == 0 { USIZE_SIZE } else { align_offset };
-
-            let start = s.as_ptr();
-            // SAFETY: We verify `len < USIZE_SIZE` above.
-            let first_word = unsafe { (start as *const usize).read_unaligned() };
-
-            if contains_nonascii(first_word) {
-                return false;
-            }
-            // We checked this above, somewhat implicitly. Note that `offset_to_aligned`
-            // is either `align_offset` or `USIZE_SIZE`, both of are explicitly checked
-            // above.
-            debug_assert!(offset_to_aligned <= len);
-
-            // SAFETY: word_ptr is the (properly aligned) usize ptr we use to read the
-            // middle chunk of the slice.
-            let mut word_ptr = unsafe { start.add(offset_to_aligned) as *const usize };
-
-            // `byte_pos` is the byte index of `word_ptr`, used for loop end checks.
-            let mut byte_pos = offset_to_aligned;
-
-            // Paranoia check about alignment, since we're about to do a bunch of
-            // unaligned loads. In practice this should be impossible barring a bug in
-            // `align_offset` though.
-            // While this method is allowed to spuriously fail in CTFE, if it doesn't
-            // have alignment information it should have given a `usize::MAX` for
-            // `align_offset` earlier, sending things through the scalar path instead of
-            // this one, so this check should pass if it's reachable.
-            debug_assert!(word_ptr.is_aligned_to(align_of::<usize>()));
-
-            // Read subsequent words until the last aligned word, excluding the last
-            // aligned word by itself to be done in tail check later, to ensure that
-            // tail is always one `usize` at most to extra branch `byte_pos == len`.
-            while byte_pos < len - USIZE_SIZE {
-                // Sanity check that the read is in bounds
-                debug_assert!(byte_pos + USIZE_SIZE <= len);
-                // And that our assumptions about `byte_pos` hold.
-                debug_assert!(word_ptr.cast::<u8>() == start.wrapping_add(byte_pos));
-
-                // SAFETY: We know `word_ptr` is properly aligned (because of
-                // `align_offset`), and we know that we have enough bytes between `word_ptr` and the end
-                let word = unsafe { word_ptr.read() };
-                if contains_nonascii(word) {
-                    return false;
-                }
-
-                byte_pos += USIZE_SIZE;
-                // SAFETY: We know that `byte_pos <= len - USIZE_SIZE`, which means that
-                // after this `add`, `word_ptr` will be at most one-past-the-end.
-                word_ptr = unsafe { word_ptr.add(1) };
-            }
-
-            // Sanity check to ensure there really is only one `usize` left. This should
-            // be guaranteed by our loop condition.
-            debug_assert!(byte_pos <= len && len - byte_pos <= USIZE_SIZE);
-
-            // SAFETY: This relies on `len >= USIZE_SIZE`, which we check at the start.
-            let last_word = unsafe { (start.add(len - USIZE_SIZE) as *const usize).read_unaligned() };
-
-            !contains_nonascii(last_word)
+            const CHUNK_SIZE: usize = if cfg!(all(target_arch = "x86_64", target_feature = "sse2")) {
+                4 * size_of::<usize>()
+            } else {
+                2 * size_of::<usize>()
+            };
+            is_ascii_chunked::<CHUNK_SIZE>(bytes)
         }
     )
 }
 
-/// ASCII test optimized to use the `pmovmskb` instruction available on `x86-64`
-/// platforms.
-///
-/// Other platforms are not likely to benefit from this code structure, so they
-/// use SWAR techniques to test for ASCII in `usize`-sized chunks.
-#[cfg(all(target_arch = "x86_64", target_feature = "sse2"))]
+/// Test for ASCII-ness `CHUNK_SIZE` bytes at a time.
+/// This loop should be simple enough that LLVM can auto-vectorise it.
 #[inline]
-const fn is_ascii(bytes: &[u8]) -> bool {
-    // Process chunks of 32 bytes at a time in the fast path to enable
-    // auto-vectorization and use of `pmovmskb`. Two 128-bit vector registers
-    // can be OR'd together and then the resulting vector can be tested for
-    // non-ASCII bytes.
-    const CHUNK_SIZE: usize = 32;
-
-    let mut i = 0;
-
-    while i + CHUNK_SIZE <= bytes.len() {
-        let chunk_end = i + CHUNK_SIZE;
-
-        // Get LLVM to produce a `pmovmskb` instruction on x86-64 which
-        // creates a mask from the most significant bit of each byte.
-        // ASCII bytes are less than 128 (0x80), so their most significant
-        // bit is unset.
-        let mut count = 0;
-        while i < chunk_end {
-            count += bytes[i].is_ascii() as u8;
-            i += 1;
-        }
-
-        // All bytes should be <= 127 so count is equal to chunk size.
-        if count != CHUNK_SIZE as u8 {
-            return false;
-        }
-    }
-
-    // Process the remaining `bytes.len() % N` bytes.
-    let mut is_ascii = true;
-    while i < bytes.len() {
-        is_ascii &= bytes[i].is_ascii();
-        i += 1;
-    }
-
-    is_ascii
+fn is_ascii_chunked<const CHUNK_SIZE: usize>(bytes: &[u8]) -> bool {
+    let (chunks, remainder) = bytes.as_chunks::<CHUNK_SIZE>();
+    chunks.iter().all(|chunk| is_ascii_scalar(chunk)) && is_ascii_scalar(remainder)
 }

library/core/src/slice/mod.rs

@@ -43,9 +43,6 @@ mod specialize;
 
 #[stable(feature = "inherent_ascii_escape", since = "1.60.0")]
 pub use ascii::EscapeAscii;
-#[unstable(feature = "str_internals", issue = "none")]
-#[doc(hidden)]
-pub use ascii::is_ascii_simple;
 #[stable(feature = "slice_get_slice", since = "1.28.0")]
 pub use index::SliceIndex;
 #[unstable(feature = "slice_range", issue = "76393")]