Dolphin, the GameCube and Wii emulator - Forums

Full Version: Custom D3D Version with massive performance boost and a lot of rendering fixes
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
Pages: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39
New Nvidia drivers are preferable due to performance improvements. I was really impressed with the performance gain the latest driver had to offer.
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, 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?

I just used the latest drivers from nvidia. Try to compare this version with master and efb set to ram.

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.
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, 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.

https://github.com/PCSX2/pcsx2/tree/master/plugins/GSdx

Not interested, but generating mipmap is not difficult, so you should just give it a try yourself.
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/1o3duCVu6Y13Yq8ozCD6_dqxhRIZ4ZmKDv1PBVcdZVP0/edit#gid=0
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, 08:21 AM)purpasmart96 Wrote: [ -> ]Man I sure would love if people would give OpenCL a second chance, galop1n, I love to hear your input on OpenCL.

OpenCL sucks, but my Compute Shader version seems to give good result on the JMC47 computer Smile
What about OpenMP?

That option slows down Legend of Zelda NES for some reason.
(06-12-2014, 08:39 AM)Gir Wrote: [ -> ]What about OpenMP?

That option slows down Legend of Zelda NES for some reason.

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 Smile ?
Pages: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39