[07_weird_assembly_machine] add a faster implementation of func0

.cargo/config.toml

@@ -0,0 +1,2 @@
+[target.x86_64-unknown-linux-gnu]
+rustflags = ["-Ctarget-cpu=native"]

src/bin/07_weird_assembly_machine.rs

@@ -1,3 +1,4 @@
+#![allow(dead_code)]
 use lib::compute_async;
 
 // https://button.qedaka.de/07_weird_assembly_machine.html?name=Jonah&n=10000000000000070&x=4410539207075887818
@@ -68,8 +69,8 @@ const DATA: &[u64] = &[
     11796026101712826135, 2668163399826034890, 1114227485540495868, 12046048025420064663,
 ];
 
-const N: u64 = 10000000047000070;
-const X: u64 = 2925478411027973783;
+const N: u64 = 10000045930000000;
+const X: u64 = 16195396392266746892;
 
 fn main() {
     compute_async(|tx| {
@@ -86,16 +87,6 @@ fn main() {
     });
 }
 
-fn func0(a: u64, mut b: u64) -> u64 {
-    let mut x: u64 = 0;
-    while b != 0 {
-        let c = b.trailing_zeros();
-        b = b ^ 1u64.rotate_left(c);
-        x = x ^ a.rotate_left(c);
-    }
-    x
-}
-
 fn func1(x: u64, mut n: u64) -> u64 {
     let x = func2(x, &DATA[0..128]);
     if x <= n {
@@ -112,6 +103,32 @@ fn func2(x: u64, data: &[u64]) -> u64 {
     out
 }
 
+#[cfg(not(all(target_arch = "x86_64", target_feature = "pclmulqdq")))]
+fn func0(a: u64, mut b: u64) -> u64 {
+    let mut x: u64 = 0;
+    while b != 0 {
+        let c = b.trailing_zeros();
+        b = b ^ 1u64.rotate_left(c);
+        x = x ^ a.rotate_left(c);
+    }
+    x
+}
+
+#[cfg(all(target_arch = "x86_64", target_feature = "pclmulqdq"))]
+fn func0(a: u64, b: u64) -> u64 {
+    use std::arch::x86_64::*;
+    unsafe {
+        let a_vec = _mm_set_epi64x(0, a as i64);
+        let b_vec = _mm_set_epi64x(0, b as i64);
+        let result = _mm_clmulepi64_si128(a_vec, b_vec, 0x00);
+
+        // Extract both 64-bit halves and XOR them
+        let low = _mm_extract_epi64(result, 0) as u64;
+        let high = _mm_extract_epi64(result, 1) as u64;
+        low ^ high
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;