[bonnhatnguyen]

Description

Adds the first C++ implementation (1.cpp) for the fasta benchmark, enabling C++ on the leaderboard.

Key Architectural Highlights

• Direct $O(1)$ Lookup Table: Rather than doing linear scanning or binary searches over cumulative probabilities per generated nucleotide, cumulative distribution thresholds are precomputed into a 136.7 KB direct lookup array (char lookup[IM]). Each pseudo-random integer in $[0, IM)$ immediately indexes the corresponding nucleotide in $O(1)$ with zero floating-point arithmetic or branch mispredictions in the hot loop.
• High-Throughput Block Buffered Output: Employs a 256 KB contiguous output buffer (out_buf) with batched fwrite, minimizing system-call overhead while remaining fully compliant with standard single-threaded output constraints.
• Double-Buffer ALU Sequence: Concatenates sequence prefix to avoid modulo arithmetic during repeated wrap operations.

Verification

• Bit-exact verification against bench/algorithm/fasta/1000_out ($N = 1000$).
• Bit-exact verification against Rust 1.rs canonical reference across 25,000 iterations ($N = 25000$).
• Clean compilation under GCC / Clang with -pipe -Wall -Wextra -O3 -fomit-frame-pointer -std=c++20.