Commit Graph

13301 Commits

Author SHA1 Message Date
magumagu 1429fccb97 Initial unoptimized JITIL flag optimization. 2014-07-30 21:41:17 -07:00
Pierre Bourdon 5506e57ab8 CR: Replace some magic values with constants. 2014-07-30 21:41:17 -07:00
Pierre Bourdon 0ff1481494 Optimize PPC CR emulation by using magic 64 bit values
PowerPC has a 32 bit CR register, which is used to store flags for results of
computations. Most instructions have an optional bit that tells the CPU whether
the flags should be updated. This 32 bit register actually contains 8 sets of 4
flags: Summary Overflow (SO), Equals (EQ), Greater Than (GT), Less Than (LT).
These 8 sets are usually called CR0-CR7 and accessed independently. In the most
common operations, the flags are computed from the result of the operation in
the following fashion:
  * EQ is set iff result == 0
  * LT is set iff result < 0
  * GT is set iff result > 0
  * (Dolphin does not emulate SO)

While X86 architectures have a similar concept of flags, it is very difficult
to access the FLAGS register directly to translate its value to an equivalent
PowerPC value. With the current Dolphin implementation, updating a PPC CR
register requires CPU branching, which has a few performance issues: it uses
space in the BTB, and in the worst case (!GT, !LT, EQ) requires 2 branches not
taken.

After some brainstorming on IRC about how this could be improved, calc84maniac
figured out a neat trick that makes common CR operations way more efficient to
JIT on 64 bit X86 architectures. It relies on emulating each CRn bitfield with
a 64 bit register internally, whose value is the result of the operation from
which flags are updated, sign extended to 64 bits. Then, checking if a CR bit
is set can be done in the following way:
  * EQ is set iff LOWER_32_BITS(cr_64b_val) == 0
  * GT is set iff (s64)cr_64b_val > 0
  * LT is set iff bit 62 of cr_64b_val is set

To take a few examples, if the result of an operation is:
  * -1 (0xFFFFFFFFFFFFFFFF) -> lower 32 bits not 0       => !EQ
                            -> (s64)val (-1) is not > 0  => !GT
                            -> bit 62 is set             =>  LT
            !EQ, !GT, LT

  *  0 (0x0000000000000000) -> lower 32 bits are 0       =>  EQ
                            -> (s64)val (0) is not > 0   => !GT
                            -> bit 62 is not set         => !LT
            EQ, !GT, !LT

  *  1 (0x0000000000000001) -> lower 32 bits not 0       => !EQ
                            -> (s64)val (1) is > 0       =>  GT
                            -> bit 62 is not set         => !LT
            !EQ, GT, !LT

Sometimes we need to convert PPC CR values to these 64 bit values. The
following convention is used in this case:
  * Bit 0 (LSB) is set iff !EQ
  * Bit 62 is set iff LT
  * Bit 63 is set iff !GT
  * Bit 32 always set to disambiguize between EQ and GT

Some more examples:
  * !EQ, GT, LT -> 0x4000000100000001 (!B63, B62, B32, B0)
                -> lower 32 bits not 0          => !EQ
                -> (s64)val is > 0              =>  GT
                -> bit 62 is set                =>  LT
  * EQ, GT, !LT -> 0x0000000100000000
                -> lower 32 bits are 0          =>  EQ
                -> (s64)val is > 0 (note: B32)  =>  GT
                -> bit 62 is not set            => !LT
2014-07-30 21:41:17 -07:00
Lioncash f507827399 Merge pull request #681 from phire/non-sse4_1
Fix PPC_FP on non-sse4.1 code paths.
2014-07-31 00:31:02 -04:00
Scott Mansell 8c857b45f8 Fix PPC_FP on non-sse4.1 code paths.
The Invalid bit on the x87 fpu is sticky, so once a single NaN goes
through the old code on CPUs without sse4.1 all future floats are
mutilated.

Patch to emulate PTEST by Fiora.

Fixes issue 7237 and issue 7510.
2014-07-31 16:00:27 +12:00
Lioncash cd37af8590 DolphinWX: Remove the Projection Hack UI 2014-07-30 19:32:41 -04:00
Pierre Bourdon 83838a645f Merge pull request #690 from Armada651/d3dfullscreen_fixes
Exclusive fullscreen fixes
2014-07-30 16:28:56 -07:00
Pierre Bourdon 9b9817f927 x64Emitter: Fix REX encoding for SETcc
Previously using the new "lower 8 bits" registers (SIL, SPL, ...) caused SETcc
to write to other registers (for example, SETcc SIL would generate SETcc DH).
2014-07-30 06:41:29 -07:00
Jules Blok 3b5625c76b VideoConfig: Ignore Borderless Fullscreen setting when the backend does not support exclusive fullscreen.
This was expected to be handled by VerifyValidity(), but that only verifies the validity of the INI files.
2014-07-30 12:15:58 +02:00
Jules Blok 5bbd34637b CFrame: Don't check the video config fullscreen setting.
Checking this flag could sometimes incorrectly have the UI assume fullscreen is already off when we're still exiting.
2014-07-30 12:15:32 +02:00
Jules Blok 4501aeefbe CFrame: Check borderless fullscreen setting before enabling exclusive fullscreen in the video config.
Fixes a bug where "Use Fullscreen" would initialize into exclusive fullscreen regardless of the borderless fullscreen setting.

Also relieves the need for the video renderer to check the borderless fullscreen setting each time.
2014-07-30 12:15:26 +02:00
Lioncash 22e9d6977b Common: State OS instead of 32/64 bit in the netplay lobby 2014-07-30 02:04:17 -04:00
Matthew Parlane 5516b0382c Merge pull request #699 from lioncash/enum
Core: Use an enum for the Gekko exception flags instead of defines
2014-07-30 13:53:58 +12:00
Matthew Parlane a3ca3b0424 Merge pull request #696 from lioncash/more-forward-decls
Convert some more header inclusions into forward declarations
2014-07-30 13:53:48 +12:00
Lioncash ad3ade1510 Core: Use an enum for the Gekko exception flags instead of defines 2014-07-29 21:51:30 -04:00
Pierre Bourdon 4b66c6c65a Merge pull request #698 from lioncash/missed-workaround
Really get rid of the MSVC 2005 workaround completely
2014-07-29 18:23:04 -07:00
Lioncash b03c12764d Really get rid of the MSVC 2005 workaround completely 2014-07-29 21:20:43 -04:00
jimbo1qaz fe9b7fa4f3 Fix D3D Real XFB texture sampling. 2014-07-29 18:15:01 -07:00
Ryan Houdek 91647f5944 Merge pull request #697 from lioncash/unused-android
Android: Remove an unused variable from MainAndroid.cpp
2014-07-29 20:08:06 -05:00
Lioncash abc7845e0c Android: Remove an unused variable from MainAndroid.cpp 2014-07-29 21:05:57 -04:00
Lioncash 522a5c35ad Convert some more header inclusions into forward declarations 2014-07-29 20:55:07 -04:00
Lioncash 3d95ed93f5 Core: Use a std::string in EXI_DeviceIPL instead of a char buffer
Allows getting rid of extra code.
2014-07-29 20:05:57 -04:00
Pierre Bourdon c0e8d9879a Merge pull request #678 from lioncash/overflow
Fix a possible overflow in EXI_DeviceIPL.
2014-07-29 16:42:58 -07:00
Pierre Bourdon f6fe299c31 Merge pull request #694 from lioncash/interp-defines
Core: Remove defines used to work around an MSVC 2005 bug
2014-07-29 16:39:05 -07:00
Lioncash 412196a055 Core: Remove defines used to work around an MSVC 2005 bug 2014-07-29 19:33:08 -04:00
Pierre Bourdon 8ff3cf1838 Merge pull request #684 from lioncash/forward-decls
Core: Turn some includes into forward declarations.
2014-07-28 20:50:27 -07:00
Pierre Bourdon 062bce8b7e Merge pull request #659 from Parlane/classic_controller_fix2
Classic controller fix
2014-07-28 20:34:00 -07:00
Pierre Bourdon 9f86d3efcf Merge pull request #689 from Shadoxfix/master
Update framelimit tooltip text because the Audio option has been removed...
2014-07-28 10:28:23 -07:00
Oussama Danba 313a743cee Update framelimit tooltip text because the Audio option has been removed. 2014-07-28 19:08:43 +02:00
Tony Wasserka 3fb829ca73 Merge pull request #589 from LPFaint99/gcifolder-fifo
Gcifolder use correct region for fifologs
2014-07-28 14:44:01 +02:00
Tony Wasserka 9c7d4b6408 Merge pull request #667 from RachelBryk/remove-audio-limit
Remove audio frame limit.
2014-07-28 14:38:35 +02:00
Matthew Parlane 3554a15180 Merge pull request #680 from lioncash/fakepoll-removal
Remove fakepoll.h
2014-07-28 21:06:37 +12:00
Matthew Parlane 86db0040b7 Merge pull request #682 from lioncash/style-oversight
Fix a style guide oversight
2014-07-28 11:39:01 +12:00
Lioncash e225e3679a Merge pull request #685 from lioncash/invalid-reg
Core: Use the enum constant for an invalid reg in JitRegCache
2014-07-27 19:37:43 -04:00
Lioncash 0bb84a1f65 Core: Use the enum constant for an invalid reg in JitRegCache 2014-07-27 19:29:52 -04:00
Lioncash eb3a1de3f6 Core: Turn some includes into forward declarations. 2014-07-27 13:37:09 -04:00
Tony Wasserka 83c2c99a8c Merge pull request #605 from neobrain/mmio_cleanup
MMIO: Cleanup Mapping class by using templates instead of macros.
2014-07-27 19:27:21 +02:00
Tony Wasserka 38c8a4efb2 MMIO: Cleanup Mapping class by using templates instead of macros. 2014-07-27 19:23:19 +02:00
Lioncash 2791f6053e Clarify the preferred prefix form to be within loops in the style guide 2014-07-27 07:45:34 -04:00
Lioncash adf3543a57 Fix a style guide oversight
Fixes an oversight which says to not use mult-line comments.
2014-07-27 06:25:51 -04:00
Lioncash 4fa71dd59e Remove fakepoll.h.
It was only used for Windows XP and lower.

This also bumps the _WIN32_WINNT define in the stdafx precompiled headers to set the minimum version as Windows Vista.
2014-07-26 22:53:40 -04:00
Lioncash 551cf4b2a2 Core: Remove the unused cycle parameter from DSPHLE update calls 2014-07-26 20:04:15 -04:00
Lioncash 0b8a6f852b Core: Use character literals within EXI_DeviceIPL 2014-07-26 19:07:19 -04:00
Lioncash 70ba8cfbc2 Core: Fix a possible overflow in EXI_DeviceIPL
m_szBuffer has a size of 256, but addressable range is 0 - 255
2014-07-26 19:06:23 -04:00
Pierre Bourdon 96cfbd1bb0 Merge pull request #677 from Armada651/use_fullscreen
Frame: Only prevent exclusive fullscreen switches when the emulator is paused
2014-07-27 00:14:56 +02:00
Jules Blok 1f24122d35 Frame: Only prevent exclusive fullscreen switches when the emulator is paused.
This fixes the "Use Fullscreen" setting in which case the switch is done while the emulator is uninitialized.
2014-07-27 00:08:27 +02:00
Pierre Bourdon 957a39942e Merge pull request #675 from lioncash/movie-params
Core: Use param padState instead of a global in SetInputDisplayString
2014-07-26 23:30:26 +02:00
Lioncash d1e487ae00 Core: Use param padState instead of a global in SetInputDisplayString 2014-07-26 17:14:54 -04:00
Lioncash 40273634a4 Merge pull request #674 from lioncash/loldsp
Core: Get rid of the void handle parameter for DSP initialization
2014-07-26 17:00:02 -04:00
Lioncash e8d0a910da Core: Get rid of the void handle parameter for DSP initialization 2014-07-26 16:54:36 -04:00