DQB and CosmoX fans, benchmarks...

[bobby]

No, it does work on any memory, VRAM or RAM, I did test that
when I was coding, an NO, no one did care about this on their
ASM libs, just aligned their data using 'align dword' or things like
that, but they didnt care about it when doing copies, transfers and
things like that.

Maybe they do not have a big design like you said, but one thing
you're wrong is about libs are about desing, NO, these libs where
designed to fit a need, routines that did stuff fast and easy, not a
complex system, thats part of a design, just because YOU dont
like their design, it does not mean they dont have one.

I would do some tests, but I dont see any point, I did it 2 years
ago, I did a big demo and I think that's the only thing I can do
for it but releasing a new version, and I just dont plan to get on
a war with you just because of this. Ill just let people who use
my library do it, after all, they are in a better position to talk
about it than you and me.

CosmoX fits some needs, UGL fits some others, maybe CosmoX
is better than UGL for somethings maybe DQB is better than
CosmoX for some other, maybe UGL for another, just dont act as
a kid and let people decide that for themselves.

[Blitz]

Well bobby, are you sure really did test on both sysram and vram ? Or did you just read the intel docs? Or some other doc perhaps, we actually tried out aligned/non aligned write to sram/vram. And all the system we tried it the aligning gave quite a bit of speed up to vram, but made no difference to sysram. And UGL, does align, otherwise it wouldn't be able to get such speeds to vram, could it?

I tried some other libs too. You can get bin+source here http://dotnet.zext.net/libbench_putmsk.rar

-----------------------------------
AMD Athlon 750 mhz, 100 mhz SDR SDRAM
CSX, Transparent pixels per second: 27986904
DQB, Transparent pixels per second: 27212180
FLB, Transparent pixels per second: 26749388
GSL, Transparent pixels per second: 28086857
REL, Transparent pixels per second: 27986904
UGL, Transparent pixels per second: 51067013
-----------------------------------
Intel Celeron 466 mhz, 66 mhz SDR SDRAM
CSX, Transparent pixels per second: 13809166
DQB, Transparent pixels per second: 12414081
FLB, Transparent pixels per second: 9323438
GSL, Transparent pixels per second: 11642221
REL, Transparent pixels per second: 10663485
UGL, Transparent pixels per second: 14699664
-----------------------------------
Intel 486DX 100 mhz, 33 mhz DRAM
CSX, Transparent pixels per second: 4623351
DQB, Transparent pixels per second: 3804702
FLB, Transparent pixels per second: 2734465
GSL, Transparent pixels per second: 4661719
REL, Transparent pixels per second: 4661719
UGL, Transparent pixels per second: 6953422
-----------------------------------

[Blitz]

Actually, i think i'm in a good position to talk about it. Besides wasting my time coding in qb, i do allot of win32 coding and even some *nix. I've come by and know many APIs, so i would say that i'm in a pretty good position to judge library design.

[bobby]

Heh I did not mean that you can not judge library design, what
I meant there is that no one could say anything about any
library that people who are really using it.

I have code for many systems too, and Im not telling the world
that Im a good designer just for that do I?

This is my last post about this, if you want to keep acting as
a kid then just keep it.

[Blitz]

Yeah, you're right. I appologise

[v3cz0r]

Someone is using blitz's account, that can't be true, nor after 2 years and 2 months :P.

Became catholic or what?

[ShadowWolf]

why are people responding to a two year old argument on somthing that is pretty much meaningless now with the advent of freebasic ?

[Dr_Davenstein]

Someone threatened to hack the forums a while back because they need updating. :???:


Hopefully not. Maybe Blitz is drunk. :lol: