Version 1.0:
 - high resolution time measurement ok.
 - detection of L1 cache ok.
 - detection of L2 cache ok.
 - CPU type detection ok.
 - partial implementation of video performance tests.

Version 1.1:
 - L1 cache code improved.
 - L2 cache code improved.
 - AT cache detection added.
 - CPU type detection reworked.
 - full implementation of video performance tests.

Version 1.2:
 - L2 cache code improved.
 - DRAM page size detection added.
 - cache line size detection added.
 - video info added.

Version 1.3:
 - different device driver used to be WARP compatible

Version 1.3a:
 - nasty bug in preloading code pages => total system lockup   (fixed)

Version 1.4:
 - L2 cache code improved.
 - CPU type detection improved.
 - DRAM page size detection improved.
 - DRAM interleave detection added.
 - results layout depends on presence of L1/L2 cache.

Version 1.4a:
 - CPU-ID bug fixed

Version 1.5
 - CPU type detection reworked (more CPU types, more reliably)
 - more robust to different environments:
   works without device driver -> less details
   works from a boot floppy -> no video testing
 - much smaller critical code with disabled interrupts should remove any
   chance for lockups due to page faults
 - fixed bug in L2 cache code (affected Pentium only)

Version 1.6
 - another CPU-ID bug fixed
 - minor code reordering to improve L1 cache detection on Pentiums
 - added testing of opcode fetch from L2 cache
 - added capability of opcode fetch testing on CPUs with harvard architecture
   (sorry, this needs much bigger code)
 - added 'more' option

Version 1.6a
 - bug in CPU type detection fixed (could not separate 386 type CPUs)
 - added test for FDIV bug

Version 1.6b
 - FDIV test introduced a weird bug, fixed (sorry, I don't have a Pentium)
 - video chipsets now detected according to the *.DSC files, not SVGA.EXE

Version 1.6c
 - UMC U5D, INTEL 486DX2-WB-SLe, AMD 486DX4 and AMD 486-SLe detection added
 - INTEL 486DX4 detection fixed
 - CYRIX detection code changed, no reply from testers so far
 - ALI graphics chip set detection added
 - italian language version added

Version 1.6d
 - CYRIX detection code fixed, works now
 - changed opcode fetch test according to new infos on Pentiums
 - reworked CPU elimination code
 - improved bus size detection code

Version 1.7
 - major code shrink
 - the results display contained many unnecessary figures
 - the figure for memory -> video transfers was wrong in case of cache misses
 - CYRIX CPUs have peak performance with 4 byte opcode fetches, INTEL with 6;
   test both, report maximum
 - check, if virtual mode extensions in INTEL CPUs are present

Version 1.8
 - small bug fix
 - improved alignment of test code and test memory
 - enhanced opcode fetch testing
 - check for presence of MSRs and subtle execution oddities (separates 386 CPUs
   manufactured by INTEL from those of other brands).
 - changed detection code for IBM CPUs (have MSRs built in).
 - changed detection code for CYRIX CPUs according to manufacturer provided code.
 - should detect CYRIX M1sc (5x86)
 - added NexGen NX586 detection code according to manufacturer provided code.
 - should detect new NexGen CPUs
 - added AMD 486/5x86 detection code according to manufacturer provided code.
 - should detect AMD 5x86 CPUs
 - detects INTEL Pentium Pro

Version 1.8a
 - v1.8 introduced a heavy bug on AMD CPUs, fixed
   (I hate this 5x86 marketing stuff)
 - improved testing of DRAM page size
 - should detect INTEL Overdrives
 - should detect Cyrix 486DX4
 - should detect TI 486DX4
 - should detect IBM/ST 5x86

Version 1.8b
 - small bug affecting 5x86 and Nx586 CPUs fixed

Version 1.8c
 - works under Warp 4
 - AMD 486DX4 was incorrectly titled 486SX-SLe, fixed
 - in seldom cases, write back L2 caches were incorrectly detected as write
   through; improved sensitivity should fix this but may render test less
   reliable
 - fixed annoying 'fractional' bus sizes
 - problems in CYRIX detection could havoc the system control chips -> fixed
 - detects CYRIX 6x86
 - detects INTEL P55 (Pentium w/MMX)
 - should detect INTEL Pentium Pro w/MMX, AMD K5 and K6 (most probably wrong
   speed detect due to lacking calibration -> reports needed)
 - more video chips

Future:
 - overall improvement :-)
 - less bugs :-)
 - fancy GUI interface
 - any other good and implementable idea
