The special edition of the patch seemed to work well from my testing so far.
[PATCH] DSP LLE faster masked math
|
12-02-2010, 06:16 PM
(This post was last modified: 12-02-2010, 08:43 PM by blazej123123.)
Hi
could someone upload this plugin with all patches for x32 verion? 12-03-2010, 07:00 AM
soooo, will this be commited soon? I mean, it is an improvement over the past LLE plugin, right?
Intel i5-4690k (Devil's Canyon) @ 3.5 GHz (+Scythe Mugen) / Gainward GTX 1070 Phoenix (OC'd) / ASUS Z97 PRO GAMER / 16GB G.Skill DDR3-2400 CL10 TridentX / X-Fi XtremeMusic / Win10 Pro 64bit / Dell S2716DG Monitor / 3x original WiiMote+MotionPlus+Nunchuk
12-09-2010, 08:06 AM
(This post was last modified: 12-09-2010, 08:35 AM by RPG Advocate.)
If this works correctly, it may make Tales of Symphonia playable with correct sound, though apparently with a performance hit.
Edited to add: I just tested ToS on the 6505M build posted last week. This DOES restore the missing instruments, but whenever the FPS goes below 60, the sound skips like crazy. This is a problem since the LLE plugin is very demanding. I get a constant 60FPS with the HLE plugin, but only get 45-50 with LLE. 12-10-2010, 12:47 AM
12-10-2010, 01:03 AM
(This post was last modified: 12-10-2010, 01:04 AM by Xtreme2damax.)
Mylek's patches work for the x86 (32-bit version) of Dolphin, and I offer builds with the patched plugin and other improvements. However there is no x86 compatible code for the past two LLE commits that have recently been committed to svn, according to skid this is the reason why:
Quote:It is possible, but will require a little bit of work. The reason is the DSP has 40bit registers which do not fit into the 32bit registers of x86. We'll need to fit the registers somehow. 12-12-2010, 09:30 AM
(This post was last modified: 12-12-2010, 10:55 AM by kiesel-stein.)
(11-29-2010, 04:13 PM)RetroX Wrote: http://home.amis.net/mpuljar/dolphin/AR_crap.7z Thanks, this test data is very helpful. After two weeks of producing some really weird looking perl-code, it all condensed down to these functions: Code: u16 addarn(u16 ar, s16 ix, u16 wr) { This C variant was translated on the spot and therefore is completely untested, but the perl-implementation works for every testcase known to me(even the wr==0 cases). The only issue i have with this is the subtle differences between addarn and subarn. I had hoped that i could somehow call addarn from subarn, but these differences make that impossible. Edit: Fixed the code 12-15-2010, 01:39 AM
The extracted test-cases can be found here, the big file is extracted from the AR_crap archive, the small file contains the tests from the dolphin sources.
My perl scripts: collector.pl was used to extract the data from the sources/results, gen-tests.pl generates about 200 MB of tests covering the known behavior for WR=1..31, tester.pl can be fed with these test files. Note: If you can't figure out how the scripts work, don't use them. I won't help you. 12-16-2010, 01:52 AM
is this pretty much the "official" DSPLLE plugin thread? ^^
just wanted to report that enabling JIT Dynarec makes Donkey Kong Jungle Beat freeze on the "wiimote screen".
Intel i5-4690k (Devil's Canyon) @ 3.5 GHz (+Scythe Mugen) / Gainward GTX 1070 Phoenix (OC'd) / ASUS Z97 PRO GAMER / 16GB G.Skill DDR3-2400 CL10 TridentX / X-Fi XtremeMusic / Win10 Pro 64bit / Dell S2716DG Monitor / 3x original WiiMote+MotionPlus+Nunchuk
|
« Next Oldest | Next Newest »
|
Users browsing this thread: 1 Guest(s)