Line

Page 1/10
| 2 | 3 | 4 | 5 | 6

By PingPong

Prophet (2783)

PingPong's picture

03-08-2012, 19:10

Hi, have someone data about msx2 vdp speed when drawing lines?
Say n. of lines/seconds when doing 212x212 lines, or 212x106 lines or 106x212 lines or 212x1 lines or 1x212 lines?
while i'm sure emulators are accurate in box like operations i'm not sure about lines or srch commands.
So what i need is data from real machines....

Login or register to post comments

By ARTRAG

Enlighted (5645)

ARTRAG's picture

03-08-2012, 20:14

Write a test for lines and I'll run it reporting its results

By Manuel

Ascended (12917)

Manuel's picture

03-08-2012, 22:35

I'm also interested in the results, to see whether it can be used to improve emulator timing.

I don't know how accurate emulators are in this respect at the moment, though.

By PingPong

Prophet (2783)

PingPong's picture

03-08-2012, 22:40

thx, the problem is really this. i've my real machine unavailable, and i need collaboration from you to get the results.
Basically the test can be done by doing a loop and measuring the time taken, maybe on active area and with blanked screen. Do you have further suggestions about the approach i can use to measure performances?

By ARTRAG

Enlighted (5645)

ARTRAG's picture

04-08-2012, 20:42

Better if you do the code for your needs
I can run it on my TR

By hit9918

Prophet (2543)

hit9918's picture

05-08-2012, 09:37

Maybe start with something simple like filling the screen with horizontal lines. Every second frame a different color to get some flicker feedback, do a hundred frames or so, that stuff takes around ten second and one can press stopwatch.

Maybe another version doing the fills with LMMV, too line by line so it is comparable. It would be a nice surprise when the dotrate is like LMMV.

Or maybe do both tests with vertical lines. Wasn't there something about small width LMMV having some overhead. The test would get another figure.

I suspect it is same as with other locigal blits, screen 8 is same speed.

By hit9918

Prophet (2543)

hit9918's picture

05-08-2012, 09:40

So one got a figure that is an average of ops in visible / invisible area,
and then can run same test again with blank screen or sprites off and see how much it changes.

By hit9918

Prophet (2543)

hit9918's picture

05-08-2012, 09:45

p.s. I don't know, did I get this right that some blitter registers are shadowed and can already be loaded with the parameters of the next blit while a blit is still running?

By PingPong

Prophet (2783)

PingPong's picture

05-08-2012, 11:08

hit9918 wrote:

p.s. I don't know, did I get this right that some blitter registers are shadowed and can already be loaded with the parameters of the next blit while a blit is still running?

no shadow, if you modify registers during command execution, you will corrupt the drawing. However, some registers, like nx,ny have the requirement to be "reloaded" at the end of the line fill or copy, so i suspect that there is some shadow copy somewhere.

I will recover my pc virtual machine, where the "msx dev environment" is. then i will start coding a simple test.

By Meits

Prophet (4041)

Meits's picture

05-08-2012, 14:18

This topic made me curious to find out differences.
I did a little test on both real hardware and openMSX both being a Sony HB-700P on 60Hz

The test program is here:
1 time=0
2 screen 5: c=2
3 for i=0 to 211
4 line(0,i)-(256,i),c
5 c=c+1:ifc=16thenc=2
6 next
7 screen 0
8 print time

The emulator constantly gives me 256 and the real machine always gives me 254...
Now what about that Tongue

By mars2000you

Enlighted (5018)

mars2000you's picture

05-08-2012, 14:22

Quote:

a Sony HB-700P on 60Hz

Question Real European machine is on 50Hz

Page 1/10
| 2 | 3 | 4 | 5 | 6
My MSX profile