Implement manual SIMD parsing

src/day1.rs

@@ -1,10 +1,15 @@
 #![allow(
     unsafe_op_in_unsafe_fn,
     clippy::missing_safety_doc,
-    clippy::identity_op
+    clippy::identity_op,
+    clippy::zero_prefixed_literal
 )]
 
-use std::{fmt::Display, simd::prelude::*};
+use std::{
+    arch::x86_64::{__m256i, _mm256_madd_epi16, _mm256_maddubs_epi16},
+    fmt::Display,
+    simd::prelude::*,
+};
 
 #[repr(align(32))]
 #[derive(Clone, Copy)]
@@ -26,31 +31,55 @@ macro_rules! get_arr {
 }
 
 #[inline(never)]
+#[target_feature(enable = "avx2,bmi1,bmi2,cmpxchg16b,lzcnt,movbe,popcnt")]
 unsafe fn preprocess(s: &str, left: &mut [u8; 90032], right: &mut [u8; 90032]) {
     let s = s.as_bytes();
 
     let mut i = 0;
 
-    while i < 14000 {
-        let a = *s.get_unchecked(i + 0) as u32 * 10000
-            + *s.get_unchecked(i + 1) as u32 * 1000
-            + *s.get_unchecked(i + 2) as u32 * 100
-            + *s.get_unchecked(i + 3) as u32 * 10
-            + *s.get_unchecked(i + 4) as u32 * 1
-            - 543328;
-
-        *left.get_unchecked_mut(a as usize) += 1;
-
-        let b = *s.get_unchecked(i + 8) as u32 * 10000
-            + *s.get_unchecked(i + 9) as u32 * 1000
-            + *s.get_unchecked(i + 10) as u32 * 100
-            + *s.get_unchecked(i + 11) as u32 * 10
-            + *s.get_unchecked(i + 12) as u32 * 1
-            - 543328;
-
-        *right.get_unchecked_mut(b as usize) += 1;
-
-        i += 14;
+    while i < s.len() {
+        let chunk =
+            (s.get_unchecked(i) as *const _ as *const u8x32).read_unaligned() - Simd::splat(b'0');
+        let shuffled_chunk = simd_swizzle!(chunk, [
+            00, 01, 02, 03, 04, 05, 06, 07, 08, 09, 10, 11, 12, 13, 13, 13, //
+            14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 27, 27, //
+        ]);
+
+        let s1 = _mm256_maddubs_epi16(
+            shuffled_chunk.into(),
+            u8x32::from_array([
+                10, 1, 10, 1, 1, 0, 0, 0, 10, 1, 10, 1, 1, 0, 0, 0, //
+                10, 1, 10, 1, 1, 0, 0, 0, 10, 1, 10, 1, 1, 0, 0, 0, //
+            ])
+            .into(),
+        );
+
+        let s2: u16x16 = _mm256_madd_epi16(
+            s1,
+            u16x16::from_array([100, 1, 1, 0, 100, 1, 1, 0, 100, 1, 1, 0, 100, 1, 1, 0]).into(),
+        )
+        .into();
+
+        let s3: __m256i =
+            simd_swizzle!(s2, [0, 2, 4, 6, 0, 2, 4, 6, 8, 10, 12, 14, 8, 10, 12, 14]).into();
+
+        let s4: u32x8 = _mm256_madd_epi16(
+            s3,
+            u16x16::from_array([10, 1, 10, 1, 0, 0, 0, 0, 10, 1, 10, 1, 0, 0, 0, 0]).into(),
+        )
+        .into();
+
+        let a1 = s4[0];
+        let b1 = s4[1];
+        let a2 = s4[4];
+        let b2 = s4[5];
+
+        *left.get_unchecked_mut(a1 as usize - 10000) += 1;
+        *left.get_unchecked_mut(a2 as usize - 10000) += 1;
+        *right.get_unchecked_mut(b1 as usize - 10000) += 1;
+        *right.get_unchecked_mut(b2 as usize - 10000) += 1;
+
+        i += 28;
     }
 }
 
@@ -139,3 +168,25 @@ pub fn part1(s: &str) -> impl Display {
 pub fn part2(s: &str) -> impl Display {
     unsafe { inner2(s) }
 }
+
+#[cfg(test)]
+mod tests {
+    use std::fs::read_to_string;
+
+    use super::*;
+
+    #[test]
+    fn test() {
+        let s = read_to_string("./inputs/1.txt").unwrap();
+        let s = s.as_str();
+
+        assert_eq!(
+            part1(s).to_string(),
+            read_to_string("./outputs/1p1.txt").unwrap(),
+        );
+        assert_eq!(
+            part2(s).to_string(),
+            read_to_string("./outputs/1p2.txt").unwrap(),
+        );
+    }
+}