NetSurf AGA optimizing

[hooverphonique]

Originally Posted by arti

Nice, gui colors are good now, cursor doesn't leave white marks on it, only on grey border (last image).

your AGA amiga runs a 1920x1080 screen?

[crazyc]

Originally Posted by NovaCoder

Cool, ready for another BETA release soon?

So looking forward to this...

[Retrofan]

Originally Posted by crazyc

So looking forward to this...

But you must know you'll need Fpu till it won't need Ixemul.

Yes, I want to try a beta when it will be ready, but I don't mind to wait for months as long as they will be still on it, and they are giving as much as they can, I'm sure.

I take the opportunity to say a big thanks to all the people involved here

You can see there are a lot of people following this thread.

[crazyc]

I have an fpu. 😃. (Blizzard 1260@80)

I would also like to say thanks for all the hard work and effort

[arti]

@NovaCoder

When linked with new sdl_aga NetSurf returns to WB on AmiKit, but on WB Classic Lite not.

Uploaded BETA 2 to The Zone! , please make tests with BETA 1.

[NovaCoder]

Originally Posted by arti

@NovaCoder

When linked with new sdl_aga NetSurf returns to WB on AmiKit, but on WB Classic Lite not.

It must be a patch in AmiKit that's upsetting it....

Does AmiKit work with BETA 1?

[arti]

Yes and with BETA 2 with old sdl_aga.

[Retrofan]

It works with WinUAE and Fpu with a 030 (Beta 2), but it takes 32.5 secs to open EAB. Later it says 11.1 secs to open a thread inside it.

[arti]

That because page is saved to disk.

[NovaCoder]

Originally Posted by Retrofan

It works with WinUAE and Fpu with a 030 (Beta 2), but it takes 32.5 secs to open EAB. Later it says 11.1 secs to open a thread inside it.

You should compare it to BETA 1 on the same set up (and same web page).

Hopefully it's a bit quicker

Would also love to hear some results from people running it on a real Amiga.

[attila06]

Tried it quickly before going working.
On real amiga, same web page no difference between Beta1 and beta2, it's not quicker .
Someone can confirm ? (hoping i wasn't well awake)

[arti]

This Beta won't load pages faster (maybe these with many images will) but scrolling should be.

[PeterK]

Originally Posted by arti

Nice, gui colors are good now, cursor doesn't leave white marks on it, only on grey border (last image).

Indeed, looks better now. Thanks a lot for being so patient with all my experiments, Arti. It seems that my job is almost done now.

The mouse pointer tracks should not be caused by my palette, I would say, but I will try to find the reason by comparing the original source with the current code regarding everything related to these two cursor functions which should preserve and restore the background under the mouse pointer (mentioned in my previous post).

Does NetSurfAGA really need the latest version of ixemul.library (v63) ? Or could this requirement be reduced to the latest available non-FPU version of ixemul.library ?

[attila06]

@arti
you're right, scrolling is faster no doubt .

[PeterK]

Hi Arti,
to get finally rid of the mouse tracks I made a new effort by handling the pure white now before the RGB level pushing and the cube processing are done.

Furthermore, I did some small optimizations and also moved the complete code of nsfb_palette_best_match_dither() from palette.c to 8bpp.c into colour_to_pixel() in order to avoid the overhead of the unnecessary function call. Removed nsfb_palette_free() which is not really required since this can be done by free() directly.

[arti]

Updated whole post:

Yes, it works but I had to correct the code:

added "extern uint8_t table_for_cube_676[834];" to pallete.h
and changed removed 'static' before uint8_t table_for_cube_676[834]; in pallete.h (sorry, should be .c here)

However, all colors are not ok. Attached second image is from original version for comparison.
Also, browser is not filling whole screen (grey area).

[PeterK]

Please use the following line in palette.h:

extern uint8_t *table_for_cube_676;

In palette.c should be the same as before:

static uint8_t table_for_cube_676[834];

[PeterK]

Ok, I give up, since I cannot compile the source code myself.

[arti]

This gives:

amigaos3/libnsfb/palette.c:21:16: error: conflicting types for ‘table_for_cube_676’
amigaos3/libnsfb/palette.h:21:17: note: previous declaration of ‘table_for_cube_676’ was here

Mine version works, so it should be ok?

[arti]

Please tell me what messages you get during compilation.
I'm sure we will sort it out.