One more benchmark/demo for the Geneve

[vol]

I have just added results for the Geneve 9640 into my project .  The Geneve shows itself very good, its results are much better than results for the Apple IIgs, IBM PC XT, or Sinclair QL.  It is only slightly slower than the Amiga 500 or MSX turboR.  Maybe it is possible that somebody can write more optimized code for the TMS9995 and overrun the Amiga.  However the main loop is only 14 assembly instructions so this is rather unlikely.  The TMS9995 also shows an excellent code density, only the DEC T-11 main loop is one instruction shorter.
Sorry I was lazy to make a better palette so generated pictures look rather rawish.
I can't help but compare several speed results with results of the old good Fractal 2.0 program that can generate and save very impressive images.
We can compare the performance for the same Mandelbrot parameters: x in [-2.25, .75], y in [-1.24, 1.24], iterations = 37.  Both programs use GRAPHIC6 (512x212, 16 colors), screenshots are below: Fractals 2.0 - about 42 minutes; Super Fast Mandelbrot Generator - about 14 seconds.  The generated images are not exactly the same and don't exactly know why.

 
It would be great if someone could run my programs on a real machine (I have to use MAME/MESS) and check several of the results with the results published in the main table of my project.  Several screenshots would also be great.  My programs don't use delays between using the VDP ports so it a question, can they work at all? 😄

[WhataKowinkydink]

Impressive!!  Thanks for sharing.  Reminds me for the second time today that I need to get my old Geneve MESS/MAME setup working again.

[Nick99]

Where can I download the program, to try it on my Geneve?

Edited December 3, 2022 by Nick99

[vol]

1 hour ago, Nick99 said:

Where can I download the program, to try it on my Geneve?

Sorry, the link needs several clicks to get it from the benchmark page.  You can run each program individually or run AUTOEXEC to get a menu.

Edited December 3, 2022 by vol

[+mizapf]

By the way, I recommend to use the "semimonotonic" coloring in Fractals; it will deliver nicer pictures. This mode does not cycle the colors but it advances monotonically for all even and odd iterations. Thus, the color sequence is

 

1-2-1-2-1-2-...-2-1-2-3-2-3-2-3-2-...-2-3-4-3-4-3-4-3... -15 (highest iteration) - 0 (Mandelbrot set).

 

Fractals uses a fixed point arithmetic that I optimized for the on-chip RAM; it uses 8 bytes, with 2 bytes integer and 6 bytes fraction. The coloring mode does not affect the performance. Also, I followed the brute force approach to run the iteration for each point on the screen without any further optimizations (beyond the math routines).

 

This is the semimonotonic coloring:

 

This is the cyclic coloring:

 

The semimonotonic coloring shows its advantages in deep zooms:

[+mizapf]

The longer I look at our Christmas tree, the more I see a Mandelbrot set. Weird.

 

[+OLD CS1]

1 hour ago, mizapf said:

The longer I look at our Christmas tree, the more I see a Mandelbrot set. Weird.

Chaos.

[vol]

On 12/21/2022 at 9:20 PM, mizapf said:

The longer I look at our Christmas tree, the more I see a Mandelbrot set. Weird.

 

I have rotated a 512x424 interlaced picture that was generated by my program and got an Xmas tree! 😀 Happy Holidays for everybody!

Edited December 25, 2022 by vol