Login
Username:

Password:

Remember me



Lost Password?

Register now!

Sections

Who's Online
26 user(s) are online (19 user(s) are browsing Forums)

Members: 0
Guests: 26

more...

Support us!

Headlines

Forum Index


Board index » All Posts (thellier)




Re: GCC disabling the instructions reordering ?
Not too shy to talk
Not too shy to talk


Thanks for all that infos

volatile didnt worked as it stored the var in the stack so using the result too early

other options didnt seems to exists in my old gcc for 68k

but I just right here successfully compile it ok by specifying the registers for each variable and using separate registers for fu2 fv2

The result is here 100% perfect:
(I am very pleased when my C is exactly like ASM )

// register struct point3D *P1 __asm("a1");
LBB125:
// register struct point3D *P2 __asm("a2");
// register union FixPoint3D *E __asm("a3");
//
// register LONG x;
// register LONG l;
// register LONG z;
// register LONG dx;
// register LONG dl;
// register LONG dz;
// register LONG dy;
//
// register float fu __asm("fp0");
// register float fv __asm("fp1");
// register float fw __asm("fp2");
// register float fdu __asm("fp3");
// register float fdv __asm("fp4");
// register float fdw __asm("fp5");
// register float fu2 ;
// register float fv2;

[...]
L566:
// while(dy)
// {

// fu2=fu;
// fu2=fu2/fw;
fsmovex fp0,fp7
fsdivx fp2,fp7
// fv2=fv;
// fv2=fv2/fw;
fsmovex fp1,fp6
fsdivx fp2,fp6
//
// E->L.x=x;
movel d2,a3@
// x= x + dx;
addl a5@(-16),d2
// E->L.z=z;
movel d4,a3@(8)
// z= z + dz;
addl a4,d4
// E->L.l=l;
movel a5@(-12),a3@(40)
// l= l + dl;
movel a0,d5
addl d5,a5@(-12)
// E->L.fw=fw;
fmoves fp2,a3@(60)
// fw= fw + fdw;
fsaddx fp5,fp2
// E++;
addw #128,a3
//
// fu= fu + fdu;
fsaddx fp3,fp0
// E[-1].L.fu=fu2;
fmoves fp7,a3@(-76)
// fv= fv + fdv;
fsaddx fp4,fp1
// E[-1].L.fv=fv2;
fmoves fp6,a3@(-72)
//
// dy--;
subql #1,d3
// }
jne L566

Go to top


GCC disabling the instructions reordering ?
Not too shy to talk
Not too shy to talk


I am using an old GCC with optimisation option O2 and I made this code for the Vampire 68080
it have been designed with the pipeline in mind so the two fdiv are computed way before their result is used so fpu "have the time" to make the computation
But gcc reorder it badly so the move using the result is just following the div in the generated ASM

Is there a way to disable the reordering in this specific C function ?

In the code all variables are registers so C and generated ASM should looks like identical

while(dy)
{
fuv=fu;
fuv=fuv/fw; // div here

E->L.x=x;
x= x + dx;
E->L.z=z;
z= z + dz;
E->L.l=l;
l= l + dl;

E->L.fu=fuv; // result used here

fuv=fv;
fuv=fuv/fw; // other div here

fw= fw + fdw;
E->L.fw=fw;
fu= fu + fdu;
fv= fv + fdv;

E->L.fv=fuv; // result used here
E++;
dy--;
}

Go to top


Re: Recompiling MiniGL 68k ?
Not too shy to talk
Not too shy to talk


Ok Thanks to all

Go to top


Recompiling MiniGL 68k ?
Not too shy to talk
Not too shy to talk


Hello

I am currently working on Wazp3D for Maggie/Vampire (BlitzQuake works 95 %)
But at some point we will need to recompile MiniGL to optimize it for Vampire too

So does someone ever recompiled MiniGL 68k ?
What compiler ?
What sources ?
Does using the last sources a good idea if they are only an adaptation to Warp3D V5 for NG ?

Thanks

Alain Thellier

Go to top


Re: Tutorial : How to install Warp3DNova, Bridge, MiniGL, Ogles2, and WarpOS emu
Not too shy to talk
Not too shy to talk


Perhaps can also install old StormMesa for 68k 3D games to run on OS4
But I am not sure if they would run anyway on OS4...

https://aminet.net/package/util/libs/StormMesa2010

Go to top


Re: compositing engine technical talks
Not too shy to talk
Not too shy to talk


CompositeTags() can use Amiga's "bitmap" as output and also as input (=textures) that give it more flexibility than Warp3D/OpenGL

Also it is more simple to draw with this single function than to setup Gallium3D/Nova thant need (say) 50 functions before even draw something

CompositeTags works in hardware on all my NG machines : Sam440,Sam460,X5000

IMHO The main point of CompositeTags is to draw blits with scale/rotation/transparency/filtering = same as SDL2 or Cairo but in a single/simple function

http://aminet.net/package/dev/src/CompositePOC

Go to top


Re: MilkyTracker 1.03
Not too shy to talk
Not too shy to talk


(duplicate)

Go to top


Re: MilkyTracker 1.03
Not too shy to talk
Not too shy to talk


@Maijestro

I am no more coding: I dont have time to do coding

Go to top


Re: MilkyTracker 1.03
Not too shy to talk
Not too shy to talk


@Maijestro

You are right bad colors in 15/16 bits mode are allways due to unknown pixel format
Lots of pixel formats are defined in Wazp3D but some "not really" Amiga apps (or OSes like Aros) may use unknown ones

Alain Thellier - Wazp3D author

Go to top


Re: Is there some "scaling" patch/option/etc for the prefs:screen ?
Not too shy to talk
Not too shy to talk


@kas1e
Thanks for feedback but I didnt code for months...
Bug1) seems normal to me : act like two mirrors facing so works as expected
2) I know the bug but the cause is unknown: seems the os mask the icons somehow ...
3) works for me: certainly one of yours windows is "special" and cant be grabbed
4) some msg are copied well but some are missing : I dont know enough amigaos to find the fix
5) i know but dont know how to fix
6) i didnt knew but ditto
7) ditto

Go to top


Re: Is there some "scaling" patch/option/etc for the prefs:screen ?
Not too shy to talk
Not too shy to talk



Go to top


Re: AmigaOs 4.1FE Classic Voodoo 3 libraries
Not too shy to talk
Not too shy to talk


>shadow in wipeout2097

Is a bug in wipeout2097: bad parameters for a warp3d fonction = only works for old badly written warp3d drivers that dont understand the parameters = not a warp3d bug

There is an env-var parameter to patch this

Go to top


Re: The OpenGL ES 2.0 thread
Not too shy to talk
Not too shy to talk


Perhaps is it possible to add some kinda "include" to the shaders : a text file containing all (Nova) missing fonctions like min(), noise(), etc.. defined as macros
This "include" then just need to be concatenated with the shader
macro.txt + shader.frag = tmp.frag
Then compile/use tmp.frag

Go to top


Re: The MiniGL thread
Not too shy to talk
Not too shy to talk


I also made a waZp3d with an hardware Nova renderer
So can help for machines without WaRp3d drivers
Sadly it is very slow with minigl apps (cause unknown...) son was never released
http://thellier.free.fr/Wazp3D57c.zip

Go to top


Re: The OpenGL ES 2.0 thread
Not too shy to talk
Not too shy to talk


Does those shaders works if those unintialized vars are set manually ?
( I mean edit the shader)

Go to top


Re: Shaderjoy 1.12
Not too shy to talk
Not too shy to talk


Hello
Perhaps using the same texsampler for all textures may economize some registers as each texture units use several registers

I mean if all textures are (say) linear then juste create one texsampler and use it for all textures (vs creating a texsampler each time we create a texture)

Go to top


Re: MiniGL4GL4ES v3.5 and Blender v248.7 issues
Not too shy to talk
Not too shy to talk


There was the same problem with blender/wazp3d
Was long ago but I remember blender need "direct to bitmap" drawing and "no indirect mode" = I mean each menu is fully drawn then next menu, so need a "flush" after each "drawprimitive"
Looks like menu drawing in blender is poorly coded and should have included a kinda "flush" for this drawing operation

Go to top


Re: Multiple Monitors and screen orientation
Not too shy to talk
Not too shy to talk


Almost OT remarks: but you can rotate a window with my aminet/clonewindow

Go to top


Re: Help me in compiling latest pixman and cairo graphics
Not too shy to talk
Not too shy to talk


As Nova can use bitmaps as source and dest it may be used to accelerate things too.
But Nova is so hard to setup even for basic drawings...

Go to top


Re: Fractal Nova
Not too shy to talk
Not too shy to talk


@SinanSam460

Certainly a SPE version wont change speed as it is certainly only the GPU that do all the computation in the frag shader "program"

Go to top



TopTop
(1) 2 3 4 ... 14 »




Powered by XOOPS 2.0 © 2001-2024 The XOOPS Project