forums.ps2dev.org

ashleydb wrote:You do have to calculate which one to use each frame for each mipmap textured object though.

Wrong, PSP handles that automatically, it can even trilinear filter between mipmaps per pixel.

I miss the way the PC version faded into blue electronic-circuit flowers.... :)

memon, in case you haven't noticed, "make kxploit" is your friend (or did you use it?) :)

Oh and btw, if this is the same effect as the one in that ASM invitro, it's one of my favourite demo effects ever. Just beautiful....

Yeah the advantage of my format is how ridiculously fast you can load the textures: a single sceRead and a few pointer fixups and you're done, complete with mipmaps, and being swizzled for max efficiency. I really should do some sample code to use textures from my tool from libGu instead of just my ...

That's true, the only hit you may take is that the palette upload will be slightly slower.

Though I have not actually benchmarked and seen if there is a measurable difference.

ashleydb, do you flush the texture cache?

Then I'm out of ideas because that should work :)

Try making some trivial test models and textures where it's easy to see what should go where, and tweak it until it works :)

Only V.

3DS max exports texture coordinates weirdly: do a v=1.0f-v for all the texture coordinates.

I agree with Holger, a sacrifice of an average of 4 bytes per memory allocation is definitely worth it for the convenience.

I can tell you right now that ROAM is NOT the right algorithm for any modern graphics hardware, it's been dead for the last 5 years :) PSP is just like a modern gaming PC, only smaller: it has a VERY strong GPU compared to the CPU. You should therefore shift as much work as possible onto the GPU and...

Thanks, I'll try it out later.

So, wanted to upgrade my pspsdk (it was about a month old) and to be sure I deleted my usr\local\pspdev folder, and downloaded and ran the latest toolchain.sh. After quite a while of compiling, I get: o sim-module.o sim-options.o sim-profile.o sim-signal.o sim-trace.o sim-utils.o sim-watch.o cp1.o i...

Nice one!
Excellent for lens flares ;)

These functions belong in the SDK, imho.

i believe the s p t q mean single, pair, triplet, quad.

AudioMonster wrote:There is probably a way to simplify that though.

Yes, you can express all of these operations as 3x2 matrices, and multiply them all together into one. Though depending on your previous level of math knowledge, that may or may not be a simplification to you :)

Haha yeah i forgot to hide the arm/thumb option when the current CPU is not an ARM :) It has a basic ARM recompiler and can almost run a couple of gba games too but that part of it sucks... It's the very beginnings of a "flexible" emu, kind of like MAME, and it's easy to add new systems, c...

It's buggy as hell, contains nasty hacks and is missing a lot of sce* functions.
But it kinda works for some pspsdk samples and some various homebrew.
Lots of homebrew crashes for various reasons though.

Also, remember to flush the cache or to "uncache" the pointer.. or does sceGuGetMemory do that nowadays?

Doing a perfectly optimal model slicer sounds suspiciously like an NP-complete problem :)

You're probably better off with a more pragmatic solution...

blah, it can't take more than an afternoon to do a reasonably decent model slicer..

WAB version changer has one really good use though: preventing accidental upgrades (if you let a friend use it or something). I used it to change my 1.50 psp to "2.00" with no bad side effects. Because it thinks it's 2.0, the 2.0 upgrader will not be able to run :)

CACHE, people, CACHE!!! Seems like we're getting three questions per day asking why some vertices are distorted, not realizing that they're probably still sitting in the CPU cache, not in memory where the gfx chip can read them. Flush your cache or write your data with an "uncached" pointe...

Yup, you don't flush your vertex data out of the cpu cache, so the gfx chip can't see all of it in memory properly.

The width of your image has to be a multiple of 32 bytes. You can't just try to swizzle a 23x20 pixel image and expect it to come out right..

Just make a camera matrix and set the VIEW matrix to it, instead of identity like the samples do.

EDIT: oops too slow

Try this function: it is the original function chp optimized for the Wiki submission and I did a small few changes to it (no divides, but shifts instead, etc...). Most compilers replace division by power-of-two integer numbers by shifts. Check the compiler or objdump output to be sure. That's true ...

Those problems come from the order you render your objects. For alpha blending to work correctly (and this applies to every platform except the dreamcast), you have to: 1. Render all opaque objects. 2. Make a list of your visible alpha blended objects. Roughly sort it by Z (a single iteration of a b...

Perhaps you forgot to flush the cache after writing the index array to memory, before drawing it?

Turn off depth test and draw your 2D stuff last?