icon.library 46.4 test versions

[AGS]

But the original code with muls and divs ist correct?

(d2 alpha, a0 image, a1 screen)

		move.b	(a0)+,d0
		move.b	(a1),d1
		sub.w	d1,d0
		muls.w	d2,d0
		divs.w	d3,d0
		add.w	d1,d0
		move.b	d0,(a1)+

[daxb]

@PeterK:
The log looks good to me. If the same works with DOpus5 for you then I don`t know what happens wrong because I cannot test WB. However, I havn`t changed the WB test loop. I just moved it to a function.

About the subdirs, it was requested to test more icon sets at once and it is an easy solution. The old way you can only test one set at once. For me its all the same (I don`t need it). If you really wish the old way then I`ll change that, no problem.

[Michael]

@daxb:
The new script does not work here in DOPUS, in WB or in WB+DOPUS environment.
Can't find directory that was just copied to RAM:
Also how it will work if both WB and DOPUS are running ?

Maybe it's better to split the scripts for different environments, at least for now,
so that we can get each one working first, and later join it in to one.
Autodetect the environment, and if both are available, ask which to use.
Getting complicated ? ;-)

[PeterK]

Originally Posted by AGS

But the original code with muls and divs is correct?

No, it's not rounding correctly for all negative values of a*(i-s).
It needs a correction by subtracting 254 from a*(i-s) for negative cases.

        move.b    (a0)+,d0
        move.b    (a1),d1
        sub.w     d1,d0   ; i-s
        muls.w    d2,d0   ; a*(i-s)
        ;
        BPL.S     .fits
        SUB.W     #254,D0
        ;
.fits   divs.w    d3,d0
        add.w     d1,d0
        move.b    d0,(a1)+

@daxb
The current script is not working, neither on WB nor on DOpus5. And I don't like the subdir solution, as I already said before. I'm very sorry for giving these negative comments here, because I'm really happy and satisfied with your previous script like it can be found in my Aminet release and I appreciate your work a lot. Thanks again for that nice benchmark. As soon as I have some time I will try to add the DOpus5 support to your older script by using the icon.library for copying the basedir icon to RAM:. That should let it work on DOpus5 with PNG icons, too.

[daxb]

Next testversion available. Changed back to old behaviour (only the selected dir will used for test). Thanks michael for the hint with running DOpus5 and Workbencht the same time (never thought about it). If both is running user is asked for choose WB or DOpus.

@PeterK:
Don`t waste your time with the script except testing my version. I`ll change/fix it until it works for you as expected. Shouldn`t be so difficult.

[PeterK]

@daxb

Very nice, now it works on WB and DOpus5 almost as it should do. Even with PNG icons on DOpus5. Thanks!

There is only one problem left: Changing to a new directory has no effect until the system is rebooted. The new drawer is not shown anywhere in the log file, but after a reboot it is used as the new default dir.

[daxb]

I guess you mean when test/copy from ram: is used and start the script a second time it don`t use the new selected directory. This is what I want to change too. At the moment I just check if "OpenWinTemp" dir exist, if yes then nothing change. So it will always test the first chosen dir with wrong text output. Easiest solution is to delete the icons from ram: after test but that you don`t wanted if I remember right. I`ve a look tomorrow

[PeterK]

Yes, that's exactly the point. Thanks for improvements.

[Retrofan]

Please excuse my absolute ignorance. How is supossed I can use/integrate Daxb script with DOpus or Wb? I've never used it. And what does it make exactly/when to use it?

[pawelek]

Is icon.library 46.4 still requires FBlit to use the icons FastRAM?

How icon.library detect FBlit? (looking for open port FBlit?)

If Fblit is required, what should have set?

[daxb]

Originally Posted by Retrofan

Please excuse my absolute ignorance. How is supossed I can use/integrate Daxb script with DOpus or Wb? I've never used it. And what does it make exactly/when to use it?

To start you just have to double click the icon. Don`t care about DOpus/WB because the script cares for you. OpenWin.rexx is for speed testing different icon.library versions. If you choose a drawer with icons via filerequester then it just open/close the drawer and measures the time.

[PeterK]

Originally Posted by pawelek

Is icon.library 46.4 still requires FBlit to use the icons FastRAM?

Yes, if you don't have a gfx-card and can use P96 or CGX instead then you need FBlit to load the icons or any other images into FastRam.

How icon.library detect FBlit? (looking for open port FBlit?)

Exactly

If Fblit is required, what should have set?

There are two FBlit.cfg files in my Aminet release. You can use them for the basic settings. But I'd recommend to read the FBlit.guide carefully and then maybe start FBlitGUI to optimize the configuration for your needs. For example, you could add more task names to the include list, like your internet browser or a picture viewer.

@daxb
PERFECTLY !!
Your script works on WB, on DOpus5 and also with DOpus5 on WB like a charm now. No problems with changing the directories anymore. Just disable the debugging output and it's done.

[daxb]

Originally Posted by PeterK

Just disable the debugging output and it's done.

Ok, trace output removed. Further (cosmetic change), the output value for "Average" (total/passes) is now truncated without rounding after third decimal point. I think that is more then enough info because single pass has only 2 digits after decimal point with at least 0.02 sec. inaccuracy. What do you think, is 3 digits (eg. 2.942) ok?

[PeterK]

Yeah, I think that's accurate enough. Sorry, didn't have the time to check it out. However, a big THX !

[PeterK]

Update to icon.library 46.4.380 and 380a:

These two versions were buggy. Please delete them.

[daxb]

Small fix. Update: Sorry, new version after user request.

[PeterK]

Update to icon.library 46.4.381:

Unfortunately, I found a stupid bug in 380. The byte in D0 holding the transparent color number was trashed. Thus, only the transparent border at the top of the image was removed. I hope that 46.4.381 works correctly now.

Meanwhile, I had to realize that my plan to remove the transpareny also from the left and right icon borders turned out to be more complex than expected. Therefore, only the transparent top and bottom lines are skipped by the direct drawing now. This reduces the amount of image data that has to be transferred from and to the gfx-card.

Please compare the speed of 376, 378 and 381 on your system with the script from daxb and report the results. THX. Without getting more feedback I may discontinue posting updates here.

[pawelek]

Ken's Icons v4 PNG icon collection (Blizz1260/64mhz, Radeon):

icon.library 46.4.376
Total: 48.80 s
Average: 9.76 s
---------------------
icon.library 46.4.378
Total: 48.70 s
Average: 9.74 s
---------------------
icon.library 46.4.381
Total: 48.72 s
Average: 9.744 s
---------------------
icon.library 46.4.382
Total: 48.70 s
Average: 9.74 s
---------------------
icon.library 46.4.383
Total: 48.64 s
Average: 9.72 s

CrystalIcons icon collection:

icon.library 46.4.376
Total: 10.02
Average: 1.002
---------------------
icon.library 46.4.378
Total: 10.88
Average: 0.989
---------------------
icon.library 46.4.381
Total: 10.82
Average: 0.983
---------------------
icon.library 46.4.382
Total: 10.76
Average: 0.978
---------------------
icon.library 46.4.383
Total: 10.70
Average: 0.972

[PeterK]

Thank you, pawelek !
Your results show that 381 is at least not really slower on your system. But 381 performes already much better than 378 on Amithlon where 376 seems to be the fastest:
http://www.a1k.org/forum/showpost.ph...postcount=1273

I still hope to get some feedback from other users and results from other systems and maybe I will find a smarter solution to skip transparency on the left and right icon borders, too.

[PeterK]

Update to icon.library 46.4.382:

Finally, I could successfully change the direct drawing routine in order to skip the transparent columns on the left and right borders of the icons, too.

But I did a little compromise. Although I could reduce the width of the image I did not change the value of bytes/row. I guess, this would mean that probably still the complete width has to be transfered as data, but I hope that the cybergraphics.library will process only the reduced image line length. I need your benchmark results to see what really happens.


Please compare the speed of 376, 378, 381 and 382 on your system with the script from daxb and report the results. THX.