===============================================================================
                     RIJNDAEL GUARD - VERSION HISTORY
===============================================================================
[v2.57] - 2026-09-23
-------------------------------------------------------------------------------
MAJOR HARDWARE ACCELERATION & STREAMING ENGINE OVERHAUL

* Hardware Crypto Engine (Intel AES-NI)
  - Integrated 14-round 256-bit AES-NI hardware vector acceleration via inline
    x86 assembly using raw machine opcode byte emission ($EMIT / db opcodes
    0x66 0x0F 0x38 0xDC / 0xDD).
  - Achieved full hardware processing throughput for CFB-256 block mode,
    bypassing high-frequency software lookup tables and byte-shift emulation.

* Hardware Detection & Reliability (Win32 SEH Probe)
  - Implemented a zero-dependency Structured Exception Handling (SEH) hardware
    probe (ProbeAESNI) via standard 32-bit Win32 TIB chain (FS:[0]).
  - Safely tests raw silicon instruction execution at runtime, correctly
    identifying active hardware pipelines even when masked by BIOS policies,
    hypervisors, or incomplete CPUID feature flags.
  - Guarantees instant, seamless fallback to 32-bit software emulation on legacy
    processors (Core 2 Duo, Pentium IV) with zero crash dialogs.

* Stream Pipeline & I/O Optimizations
  - Expanded primary thread streaming buffer from 64 KB to 1 MB (1,048,576 bytes)
    to leverage native Windows file system read-ahead and write-behind caching.
  - Inlined the 256-bit CFB assembly transformation directly into the stream 
    processing loop, eliminating millions of high-frequency subroutine call 
    frames and stack setup/teardown cycles.
  - Implemented the `g_aesArgs` global parameter bridge to guarantee thread-safe,
    fail-safe memory operand passes into NASM blocks across all compilation passes.
  - Throttled GUI window invalidations (InvalidateRect) and thread yielding
    to execute strictly on 1% progress boundary changes, freeing the CPU thread
    for unthrottled streaming.

* Performance Metrics (Baseline: Intel Core i7-3520M @ 2.90 GHz)
  - 480 MB File Stream: Reduced encryption/decryption time from ~6 minutes
    (software emulation) down to ~2.8 seconds (AES-NI accelerated).
  - Average CPU Load: Dropped from 25% single-core saturation to <1% idle load.
  - Throughput: Scaled effective I/O processing bandwidth from ~1.32 MB/s up to 
    160+ MB/s (bus-speed bound).
    
-------------------------------------------------------------------------------
Version 2.56 [Build Fix] - September 2026
-------------------------------------------------------------------------------
[FIXED] High-Volume Stream Stack Corruption / Crash
        - Resolved an issue where processing large files (>137 MB) caused an
          unhandled access violation and application crash during sustained
          cipher iterations.
        - Cause: Local stack buffer `tmpState` inside `CipherEngine()` was
          vulnerable to index overshoots during shift-row transformations on 
          32-bit block configurations, corrupting the call stack over 
          millions of loop executions.
        - Fix: Expanded internal `tmpState` scratch buffer allocation and
          implemented strict index boundary clamping (`idx < 64`) prior to
          state assignments.

[OPTIMIZED] Core Cipher Engine Memory Overhead
        - Verified zero-leak memory footprint ($O(1)$ constant spatial complexity)
          holding flat at ~20 MB RAM during multi-gigabyte processing passes.
        - Validated full stream inversion accuracy and byte-for-byte fidelity 
          on large ISO container decryption.

-------------------------------------------------------------------------------