[DOC] Cookbook entry on auto-vectorizable dna4.

doc/cookbook/index.md

@@ -198,3 +198,13 @@ This will use `4` threads by default and can be adjusted by setting `seqan3::con
 the desired value:
 
 \snippet doc/cookbook/compression_threads.cpp example
+
+# Auto vectorized dna4 complement
+
+Our alphabet seqan3::dna4 cannot be easily auto-vectorized by the compiler.
+
+See discussion here: https://github.com/seqan/seqan3/issues/1970
+
+You can add your own alphabet that is auto-vectorizable in some use cases:
+
+\include doc/cookbook/simdyfyable_alphabet_complement.cpp

doc/cookbook/simdyfyable_alphabet_complement.cpp

@@ -0,0 +1,73 @@
+#include <seqan3/alphabet/nucleotide/nucleotide_base.hpp>
+#include <seqan3/alphabet/nucleotide/rna4.hpp>
+
+class simd_dna4 : public seqan3::nucleotide_base<simd_dna4, 256>
+{
+private:
+    using base_t = nucleotide_base<simd_dna4, 256>;
+
+    friend base_t;
+    //!\cond \brief Befriend seqan3::alphabet_base.
+    friend base_t::base_t;
+    //!\endcond
+    friend seqan3::rna4;
+
+public:
+    constexpr simd_dna4() noexcept = default;
+    constexpr simd_dna4(simd_dna4 const &) noexcept = default;
+    constexpr simd_dna4(simd_dna4 &&) noexcept = default;
+    constexpr simd_dna4 & operator=(simd_dna4 const &) noexcept = default;
+    constexpr simd_dna4 & operator=(simd_dna4 &&) noexcept = default;
+    ~simd_dna4() noexcept = default;
+
+    using base_t::base_t;
+
+    template <std::same_as<seqan3::rna4> t> // template parameter t to accept incomplete type
+    constexpr simd_dna4(t const & r) noexcept
+    {
+        assign_rank(r.to_rank());
+    }
+
+    static uint8_t constexpr alphabet_size = 4;
+
+    constexpr simd_dna4 & assign_rank(rank_type const c) noexcept
+    {
+        base_t::assign_rank(c == 0 ? 'A' : (c == 1 ? 'C' : (c == 2 ? 'G' : 'T')));
+        return *this;
+    }
+
+    constexpr simd_dna4 & assign_char(char_type const c) noexcept
+    {
+        base_t::assign_rank(c);
+        return *this;
+    }
+
+    constexpr char_type to_char() const noexcept
+    {
+        return base_t::to_rank();
+    }
+
+    constexpr rank_type to_rank() const noexcept
+    {
+        char_type c{to_char()};
+        char_type upper_char = c & 0b0101'1111; // to_upper
+        rank_type rank{};
+        rank = (upper_char == 'T') * 3 + (upper_char == 'G') * 2 + (upper_char == 'C');
+        return rank;
+    }
+
+    constexpr simd_dna4 complement() const noexcept
+    {
+        char_type rank{to_char()};
+        rank ^= rank % 2 ? ('C' ^ 'G') : ('A' ^ 'T');
+
+        simd_dna4 ret{};
+        static_cast<base_t &>(ret).assign_rank(rank);
+        return ret;
+    }
+};
+
+int main()
+{
+    // ...
+}