forums.ps2dev.org

MrMr[iCE] · Joined: 03 Oct 2005 Posts: 43

I've been working on a general purpose math lib to replace most of libm's implementations of major math functions. I'm almost done with it but here's a taste of what to expect, in cycle times:

Tinnus · Joined: 29 Jul 2006 Posts: 67

I think it should be useful to add functions like sinfd, sinft and sinfq (for all functions) to speedup calculations in the case the user wants/can calculate 2, 3 or 4 values at the same time. Should be faster than calling the function multiple times :)

I know the main purpose is to just replace libm, but why not enable a little (more) optimization as well? :)
_________________
Let's see what the PSP reserves... well, I'd say anything is better than Palm OS.

hlide · Joined: 10 Sep 2006 Posts: 750

well if you were on #pspdev, you should probably know that to enable more optimization you need a gcc "aware" of vfpu registers (single or vector ones). So such optimizations for sinfd/t/q would not really be efficient and very cumbersome to use.

MrMr[iCE] · Joined: 03 Oct 2005 Posts: 43

v2 is up for grabs

http://mrmrice.fx-world.org/files/libpspmath_v2.zip

This one adds quite a few quaternion functions, check pspmath.h for the full list. atan2f is available as well.

to install:

cools · Joined: 04 Mar 2006 Posts: 46

This is pretty cool! It can make using the vfpu a lot easier for people who dont know how to use the vfpu...

MrMr[iCE] · Joined: 03 Oct 2005 Posts: 43

v3 is up now

http://mrmrice.fx-world.org/files/libpspmath_v3.zip

This adds one new function: vfpu_ease_in_out. This function is mainly for animation control, when you want to smoothly interpolate between 2 points with an accleration and deceleration curve.

Ive included a sample in the zip that shows how to use the quaternion functions, including how to interpolate between 2 quaternions, how to make a random 3d starfield and some other features. like the vfpu random number generator.

enjoy =)

Bytrix · Joined: 14 Sep 2005 Posts: 72 Location: England

Great work MrMr[iCE]. It'll be interesting to see how these functions speed up my games. Especially in collision detection.

pegasus2000 · Joined: 12 Jul 2006 Posts: 160

Xfacter · Joined: 28 Feb 2007 Posts: 9

Great lib, but there's a problem with vfpu_atan2f. An x and y of zero will cause a crash (divide by zero). Most libraries I've seen return 0 when this is encountered, easy fix.

hlide · Joined: 10 Sep 2006 Posts: 750

KickinAezz · Joined: 03 Jun 2007 Posts: 328

http://mrmrice.fx-world.org/files/libpspmath_v4.zip -<<< Is also available

Version 4!??

hlide · Joined: 10 Sep 2006 Posts: 750

Art · Joined: 09 Nov 2005 Posts: 647

This is a noob question, but if you define floats in your app, and perform
math operations on them, does that automatically mean that the FPU is used
to calculate them in the PSP?

Raphael · Joined: 17 Jan 2006 Posts: 646 Location: Germany

Yes. Else it would have to be done in software emulation (as IS done with doubles) and this would be hella slow.
_________________
<Don't push the river, it flows.>
http://wordpress.fx-world.org - my devblog
http://wiki.fx-world.org - VFPU documentation wiki
Alexander Berl

Art · Joined: 09 Nov 2005 Posts: 647

Thanks, I suspected so.

Xfacter · Joined: 28 Feb 2007 Posts: 9

I don't know if you meant to do this or what, but vfpu_logf is actually natural log... Here's log:

hlide · Joined: 10 Sep 2006 Posts: 750

DELETED

hlide · Joined: 10 Sep 2006 Posts: 750

Xfacter · Joined: 28 Feb 2007 Posts: 9

headness13 · Joined: 21 Oct 2007 Posts: 2

very nice library, but there are a few functions missing, like:

rotate_matrixX(M44, angle)
rotate_matrixY(M44, angle)
rotate_matrixZ(M44, angle)
rotate_matrixXYZ(M44, anglex, angley, anglez)
multiply_matrix(M44, M44, M44)
vector_cross_multiply (for FVector3 and FVector4)

could you help me with this things ?! - i can make them in C code, but asm is so much faster - and speed is a ishue at what i-m doing
thanx if you decide to help me MrMr[iCE]

snowsquirrel · Joined: 24 Feb 2008 Posts: 51

hello,

I am interested in the vector proecessing funciton in the lb. But I notice that they are v4's and not v3's. To use this with v3's, do I have to assign x,y,z in my v4 to xyz in my v3. Or do most people just use v4's for everything, ignoring the 'w' when only xyz are needed?

~S

hlide · Joined: 10 Sep 2006 Posts: 750

well, as an example :

snowsquirrel · Joined: 24 Feb 2008 Posts: 51

yes, I have looked at the code, but it really means nothing too me.

I suspected that the instructions were optimized for v4's not v3's. I think that is pretty common.

But I am not 100% sure what you are saying?

hlide · Joined: 10 Sep 2006 Posts: 750

snowsquirrel · Joined: 24 Feb 2008 Posts: 51

Ok I understand everything but this line:

hlide · Joined: 10 Sep 2006 Posts: 750

have a look upon http://www.ibiblio.org/gferg/ldp/GCC-Inline-Assembly-HOWTO.html .

hlide · Joined: 10 Sep 2006 Posts: 750

snowsquirrel · Joined: 24 Feb 2008 Posts: 51

Heimdall · Joined: 10 Nov 2005 Posts: 259 Location: Netherlands

Hi guys,

From v3 to v4 I see that there was a change on the API and the demo on v3 doesn't build, does anyone know what changed from v3 to v4? (so i can be lazzy and don't diff the code ;))

Heimdall · Joined: 10 Nov 2005 Posts: 259 Location: Netherlands

nevermind it was straightforward!