JitIL: Reverted r6135 again. Please refer r6110, r6131, r6135, and Issue 3097 about the problem. I will check the problem after I get MP2.
Further optimised the JIT cache lookup. Attempted to fix the Wii games in ICC builds.
JitIL: Restored some instruction handlers reverted in r6131. If this revision does not introduce the bugs reported in Issue 3097 and r6127, the cause of the bugs is the wrong implementation of «mtcrf» which was fixed in r6134. If not, there are other bugs. I will revert again. Any way, they were my fault. I’m sorry for all.
JitIL: Fixed the wrong implementation of «mtcrf». (The bit order was reverse…) This commit may fix the bugs reported in Issue 3097 and r6127. (I could not test them because I don’t have the games which are reported.)
Fix RTC to report the correct system time in Wii and GC titles as reported in Issue 1817
Modify GetLocalTimeSinceJan1970 to account for DST.
GetGCTime() returns only GC epoch time(used by most Wii titles.)
IPL-DEV subtracts Wii bias before copying to m_RTC(mostly used by homebrew.)
My first commit!
Done some refactoring around the PowerPC stuff. Main changes:
* Declared CPUCoreBase class
* Wrapped the interpreter into a class and made both it and JitBase inherit from CPUCoreBase
* Avoided reinitializing interpreter’s and Jit’s tables
A great deal of this work was done by nodchip, so kudos to him :-)
To finish, please forgive me for my not-so-great english.
JitIL: Partially reverted r6110 (andcx, norx, orcx, nandx, and eqvx) to fix the bug reported in Issue 3097.
JitIL: Commented out folding rules which are not tested. This commit is for the check of the bug reported in Issue 3097. The bug was introduced in r6110 by myself. In r6110, I added a function to convert andx, nandx, andcx, orx, norx, orcx, xorx, and eqvx into IRs. It is clear that the function has the bug. However I could not find the bug in the function. Then I thought there may be bugs in folding rules in IRBuilder because there are some untested folding rules which fold and, or, xor, etc. (This is also written by myself. I’m sorry.) I tested some folding rules and commented out the other untested rules.
Optimised the JIT cache lookup in JIT and JITIL. Gives a <5% speed-up in GameCube games. Wii games and GC games+MMU get a smaller speed-up. Cleaned up some code in the JIT and memory functions. Скачать:
JitIL: Extracted local variables to prevent calling regLocForInst()/fregLocForInst() many times. It was pointed out in r6127.
JitIL: Improved the register usage of some IL instructions (SExt8, ICmp*, DupSingleToMReg, InsertDoubleInMReg) for the speed improvement.
some experiments with lighting, please test an let me know if is a improvement or make things worst
Changed WII IPC HLE to continue execution after the game attempts to open unknown devices.
The allows Prince of Persia: The Forgotten Sands to be played.
Fixes issue 2684
small number mistake
9/10-bit accelerometer data interleaved with buttons
some useless IR pointer math… It still doesn’t make rotation work
a sign fix when tilting along y axis (twisting the wiimote)
* Improved constants folding in load/store instructions
* Merged load instructions
This is almost the same commit as r6076/r6077 but x64 build has been fixed.
Thanks a lot to skidau and BHaaL!!
Reverted r6112… The iPhone client was submitted to Cydia with the apropiate patches
JitIL: Reverted r6117 because I could not find the case js.downcountAmount == 0. I’m sorry for confusion.
Невероятно красивый анимационный рекламный ролик новой игры METROID: Other M с демонстрацией геймплея.