Friday, October 21, 2016

New tests and added Fibonacci sunflower

So, I'm new owner of MiST FPGA, I used various cores and do new tests.

First was tested Commodore 16 core for compare with VICE emulator. Result was generally different:
VICE Plus/4 version 2.4.29 = 8:32:35
Commodore 16 core 160811 = 10:17:00
It's near 2 hour difference

I tested also Commodore 64 core for compare with Commodore 128 in C64 mode:
Here was big sursprise - time is identical - 9:54:29
C64 core 160627 and C128 mode 64 at 1MHz

New tested was also Amstrad CPC - thanks to boys from +4World - they converted program.
Running Z80A at 4MHz, MiST core 160917-r005.8.2:


Result was 5:23:50 what is excellent - of course comparing 1-2MHz vs 4MHz is not accurate.
In Mandelbrot tests is still missing Mega 65.

At +4World was also written about Fibonacci sunflower... so - why don't try?
First was tested Commodore 128:

VIC-IIe MultiColor 160×200 Commodore 128 mode Basic 7.0 MOS 8502 at:
1MHZ = 8:00.5 (in minutes)
2MHz = 3:52.8
with WDC 65816 (SuperCPU) at 20MHz = 0:22.0

then:


VIC-IIe MultiColor 160×200 Commodore 64 mode of Commodore 128 Basic 3.5
MOS 8502 at 1MHz = 7:29.0
WDC 65816 at 20MHz = 0:19.0

and M65:



Mega 65 (tested was only Commodore 64 mode 'cause using VIC-II MultiColor mode) in Basic 3.5
1MHz = 6:36.0
48MHz = 0:10.3 - this is really visible difference between SuperCPU and GS4510...

Last tested were MiST cores.

Commodore 64 core 160627

MOS 6502 at 1MHz in Basic 3.5 = 7:28.2

and Commodore 16 core 160811

MOS 8501 at 1,76MHz in Basic 3.5 = 6:05.5

At this blog were compared only Basic programs, so it's not real accurate compare. My last task  and test is testing Mandelbrot on Mega 65.

Thursday, August 25, 2016

Commodore 64 mode of Mega 65 test

I tested this morning Mega 65 in C64 mode at 48MHz. Surprise was that it was not so faster vs SuperCPU
So, results of Mega 65 mode C64 vs SuperCPU128

M65 mode C64 at 1MHz = 08:54:31
M65 mode C64 at 48MHz = 00:14:17 - only ~1 min. faster than SCPU
C128 VIC-IIe at 20MHz = 00:15:02
C128 mode C64 VIC-IIe at 20MHz = 00:27:36 - much slower than M65
C128 VDC at 20MHz = 01:20:22 - for correction time must to be divided by 2 = ~00:40:11 - even much slower...

Miro

Wednesday, August 24, 2016

New two tests

Today I tested Commodore 128 VIC-IIe at 2MHz and Mega 65 in Commodore 64 mode in Basic 3.5
C128 program in Basic 7.0 was modified only by writting FAST command into program, so I had no any visible progress, after was complete it switched back to SLOW mode, turned VIC-IIe on and I saw drawn picture. After keypress I saw result - on C128 forum was written ~6 hours 30 mins... I believed not, but it is true: 06:10:56... Also on forum was written that progress status can be visible by border flash each line or so - it cannot! Why? Test couldn't to be compared - other tests have no command more...

So, compare C128:
VDC at 2MHz = 12:25:49
VDC at 20MHz = 01:20:22
VIC-IIe at 1MHz = 12:54:17*
VIC-IIe at ~1,3MHz = 09:45:54*
VIC-IIe at 2MHz = 06:10:56* - really similar to VDC at 2MHz
VIC-IIe at 20MHz = 00:15:02* - more than 2 times faster than VDC at 20MHz
*in compare to VDC multiply time by 2 'cause VDC draws 640x200 and VIC-IIe 320x200

Mega 65 in C64 mode
Used was again Basic 3.5 and result is bit faster, but really not so big difference, approx. 1 hour.

Compare C64 modes of C128 and M65:
C128 at 1MHz = 09:54:29
C128 at 20MHz = 00:27:36
M65 at 1MHz = 08:54:31

I add Plus/4 TED emulation to complete
TED at 1,76MHz = 08:32:35 (emulation by VICE 2.4.29)

I must to find GRAPHIC command syntax in Basic 10.0 to compare Mega 65 at 48MHz vs SCPU128, also to do test C64 mode at 48MHz.
For compare to Commodore 65 I have no one and only MESS can to be compared...

Nice can be if any Commodore 64 user and maybe SCPU with C64, also Plus/4 or TED Series user test it and reports result... I know it's lot of time, so chances are low.

Miro

Tuesday, August 16, 2016

MIRKOSOFT and Robert Willie (aka Hydrophilic) tested Mandelbrot engine on our Commodores.

Mandelbrot Basic program was created by Robert Willie for Basic 7.80 and VDC, corrected for VIC-IIe and TED with Basic 7.0 and 3.5 by Miro Karkus (MIRKOSOFT).

Tested were:

Commodore 128 in two versions:
- VIC-IIe in half density

- VDC in full density
They were tested with CMD SuperCPU at 20MHz, slow mode 1MHz, VDC version in fast mode 2MHz and VIC-IIe also overclocked ~1,3MHz
Results are here:
VDC 640x200 MOS8502 2MHz Basic 7.80 = 12:25:49
VDC 640x200 WDC65816 20MHz Basic 7.80 = 01:20:22
VIC-IIe 320x200 MOS8502 1MHz Basic 7.0 = 12:54:17
VIC-IIe 320x200 MOS8502 ~1,3MHz Basic 7.0 = 09:45:54
VIC-IIe 320x200 WDC65816 20MHz Basic 7.0 = 00:15:02
You can see that used were two Basic versions:
Basic 7.0 - standard Commodore 128 Basic
Basic 7.80 - extended Commodore 128 Basic 7.0 to VDC graphics (created by Robert Willie)
When we compare VDC vs VIC-IIe versions - VIC-IIe version time must to be multiplied twice.

Commodore 64
- really Commodore 128 in C64 mode and VIC-IIe

Here's opened door - can anybody test it on real Commodore 64 with VIC-II and maybe SCPU64?
Only download D81 image and post results in comments (Basic 3.5 for C64 is included).

Used was Basic 3.5 for Commodore 64.
Results of C64 mode are here:
VIC-II 320x200 MOS8502 1MHz Basic 3.5 = 09:54:29
VIC-II 320x200 WDC65816 20MHz Basic 3.5 = 00:27:36
When is compared result of VIC-IIe in native mode, C64 mode is faster at 1MHz, but slower at 20MHz. Again comparing to VDC requires multiplying

Commodore Plus/4
- really VICE emulator only, YAPE tried but failed at drawing.

Here is door opened totally: Can anybody test mandelbrot on any TED machine, best all - C16, C116 and Plus/4 for compare between and to C64 and C128 ?
Download D81 and we will be thankful.

Used was VICE 2.4.29 and standard Basic 3.5 emulation of Plus/4.
TED Series has not any accelerator to compare, but here's result:
TED 320x200 MOS8501 1,76MHz Basic 3.5 = 08:32:35
Looks like fastest in compare to C64 mode 1MHz, even C128 2MHz, but really it's emulator and again comparing to VDC requires multiplying

Even - each Basic version handles interrupts and services incl. Editor by other way, so real compare is MLM program...
When we look at Basic:
2.0 - was unused, missing commands - but handles near nothing - so works faster
3.5 - in case of C64 mode speeds up cursor movement, so it looks all is handled faster
3.5 - in case of TED Series - there is handled more tasks and also Editor is not so poor - so CPU helps
7.0 - C128 Basic handles more than other compared machines, so slower
7.80 - extension of Basic 7.0 adds no one more handling task - so regular to 7.0

So, here's disk image, download and test - then post results in comments here!

Thank you all very much!

Miro Karkus and Robert Willie or MIRKOSOFT and Hydrophilic