New Nvidia drivers are preferable due to performance improvements. I was really impressed with the performance gain the latest driver had to offer.
Custom D3D Version with massive performance boost and a lot of rendering fixes
|
06-11-2014, 01:12 PM
galop1n is the latest code on your github? because I compiled it and wave flickering in Wind waker are back when using Fast depht calc
06-11-2014, 01:20 PM
(06-11-2014, 03:55 AM)Link_to_the_past Wrote:(06-11-2014, 03:51 AM)rokclimb15 Wrote: I have a GT 430 I can use to test. Will that suffice? What driver release does it need? Ok. I'll try it out tomorrow. I have to pull the card and install it in a better system since the CPU is awful in that box. 06-11-2014, 08:25 PM
galop1n , are you interested in PCSX? I hope you would a take a look at it's Direct X backend and maybe implement hardware based mipmapping.
https://github.com/PCSX2/pcsx2/tree/master/plugins/GSdx 06-11-2014, 09:44 PM
(06-11-2014, 01:12 PM)purpasmart96 Wrote: galop1n is the latest code on your github? because I compiled it and wave flickering in Wind waker are back when using Fast depht calc No, i still have in local all the fix i made after the optimization pass. Now that the depth thing is nearly complete, i will push it soon with the other fixes and continue my stuff on Dolphin ( i have a fully working gpu texture decoder now ^^ ). (06-11-2014, 08:25 PM)Gir Wrote: galop1n , are you interested in PCSX? I hope you would a take a look at it's Direct X backend and maybe implement hardware based mipmapping. Not interested, but generating mipmap is not difficult, so you should just give it a try yourself. 06-12-2014, 12:05 AM
I added your builds to my performance testing. Considering this is using Backend_Optimization_95 that doesn't include any of the stuff you're currently working on, it's pretty amazing.
https://docs.google.com/spreadsheets/d/1...edit#gid=0 06-12-2014, 08:21 AM
Man I sure would love if people would give OpenCL a second chance, galop1n, I love to hear your input on OpenCL.
06-12-2014, 08:25 AM
06-12-2014, 08:39 AM
What about OpenMP?
That option slows down Legend of Zelda NES for some reason. 06-12-2014, 08:45 AM
(06-12-2014, 08:39 AM)Gir Wrote: What about OpenMP? OpenMP cannot fight with a gpu solution because : 1. CPU, take the raw memory, do untiling ( reorder pixels ) and conversion ( 565 to 8888, read final value in a lookup table, …) from various format to uncompressed RGBA 32 bits, then upload the FAT uncompressed texture to gpu 2. GPU, take the raw memory ( far less than raw texture ), upload it to GPU, run a gpu program that convert to uncompressed, that's quite similar to do a rendering reading a texture, so it is quite effective and match with how GPU works. What do you think is better ? |
« Next Oldest | Next Newest »
|
Users browsing this thread: 5 Guest(s)