Commit Graph
2455 Commits
Author SHA1 Message Date
Tillmann Karras 5fc3cb2379 Interpreter: deduplicate integer comparisons 2021-06-07 22:39:15 +01:00
JosJuice 4ef4ac41d1 JitArm64: Fix clobbering LR in FakeLKExit
Fixes Need for Speed: Carbon accessing invalid memory when
branch following and BLR optimization are enabled simultaneously.
2021-06-06 22:43:57 +02:00
Tilka 8139967768 Merge pull request #9787 from JosJuice/jitarm64-slwx-top
JitArm64: Discard top 32 bits in slwx result
2021-06-06 08:46:16 +01:00
Tilka 89af7b82f2 Merge pull request #9770 from JosJuice/jits-accidental-gt
Jits: Fix accidentally setting GT in CR when clearing EQ
2021-06-06 05:17:25 +01:00
Tilka 6c0180fc61 Merge pull request #9748 from JosJuice/fma-accuracy
Interpreter/Jit64: Emulate FMA accurately in more cases
2021-06-06 02:29:42 +01:00
JosJuice 86de3df072 JitArm64: Discard top 32 bits in slwx result
srwx would like to be able to assume that the top 32 bits
are zero, and cmpl is already doing so.
2021-06-05 23:15:08 +02:00
JMC47 638909aec6 Merge pull request #9751 from JosJuice/jitarm64-fcmpx-ftz
JitArm64: Make fcmpX with flush-to-zero enabled less bad
2021-06-05 05:27:52 -04:00
JosJuice b6cc3c4b6e JitArm64: Make fcmpX with flush-to-zero enabled less bad
See the added code comment for details. Fixes Pokémon Battle
Revolution not progressing past the title screen.
2021-06-02 20:15:33 +02:00
JosJuice de3fed6093 Jits: Fix accidentally setting GT in CR when clearing EQ
https://bugs.dolphin-emu.org/issues/12526
2021-05-30 22:07:53 +02:00
JosJuice e0c81ae54a JitArm64: Fix MSVC warnings 2021-05-28 15:34:08 +02:00
JosJuice 2c38d6419e Jit64: Emulate FMA accurately when determinism is enabled
When determinism is enabled, we either want all CPUs to use FMA or
we want no CPUs to use FMA. Until now, Jit64 has been been doing
the latter. However, this is inaccurate behavior, all CPUs since
Haswell support FMA, and getting JitArm64 to match the exact
inaccurate rounding used by Jit64 would be a bit annoying. This
commit switches us over to using FMA on all CPUs when determinism
is enabled, with older CPUs calling the std::fma function.
2021-05-25 12:37:52 +02:00
JosJuice 9bc5bd83a9 Interpreter: Use std::fma for emulating FMA
This is more accurate to the original hardware's rounding behavior.
2021-05-25 10:54:00 +02:00
Léo Lam 51671921c4 Merge pull request #9441 from skylersaleh/master
Apple M1 Support for MacOS
2021-05-24 12:39:01 +02:00
JosJuice ed331291d1 Merge pull request #9747 from JosJuice/jitarm64-sync
JitArm64: Jump to dispatcher on downcount <= 0, not < 0
2021-05-24 12:02:20 +02:00
JosJuice 86d2c2a1c3 JitArm64: Jump to dispatcher on downcount <= 0, not < 0
All the other CPU emulation cores do it on <= 0.
2021-05-24 10:42:04 +02:00
Skyler Saleh 76ed9310f2 Apple M1: RAII Wrapper for JITPageWrite*Execute*()
Added RAII wrapper around the the JITPageWriteEnableExecuteDisable() and
JITPageWriteDisableExecuteEnable() to make it so that it is harder to forget to
pair the calls in all code branches as suggested by leoetlino.
2021-05-22 15:25:18 -07:00
Skyler Saleh f567fd93b9 Apple M1: Removed unavailable CPU core dialog box
Removed the unavailable CPU core dialog box that asked users to change their
selected CPU core to one that is available. Instead, Dolphin now just overrides
the core to the default, and logs that it performed the override.
2021-05-22 15:25:18 -07:00
Skyler Saleh f92ccd5058 Apple M1: Fix bug that could cause crash with MMU
Added a Common::JITPageWriteDisableExecuteEnable() that could be missed when
a memory exception is triggered by the running game.
2021-05-22 15:25:18 -07:00
Skyler Saleh 948764d37b Apple M1: Build, Analytics, and Memory Management
Analytics:
- Incorporated fix to allow the full set of analytics that was recommended by
  spotlightishere

BuildMacOSUniversalBinary:
- The x86_64 slice for a universal binary is now built for 10.12
- The universal binary build script now can be configured though command line
  options instead of modifying the script itself.
- os.system calls were replaced with equivalent subprocess calls
- Formatting was reworked to be more PEP 8 compliant
- The script was refactored to make it more modular
- The com.apple.security.cs.disable-library-validation entitlement was removed

Memory Management:
- Changed the JITPageWrite*Execute*() functions to incorporate support for
  nesting

Other:
- Fixed several small lint errors
- Fixed doc and formatting mistakes
- Several small refactors to make things clearer
2021-05-22 15:25:17 -07:00
Skyler Saleh 4ecb3084b7 Apple M1 Support for MacOS
This commit adds support for compiling Dolphin for ARM on MacOS so that it can
run natively on the M1 processors without running through Rosseta2 emulation
providing a 30-50% performance speedup and less hitches from Rosseta2.

It consists of several key changes:

- Adding support for W^X allocation(MAP_JIT) for the ARM JIT
- Adding the machine context and config info to identify the M1 processor
- Additions to the build system and docs to support building universal binaries
- Adding code signing entitlements to access the MAP_JIT functionality
- Updating the MoltenVK libvulkan.dylib to a newer version with M1 support
2021-05-22 15:25:17 -07:00
Scott Mansell 610613ee76 Use correct mask for Fake VMem
Shouldn't have any behaviour change for regular usage as both masks are 32MB
by default.
But fixes theoretical buffer overrun when memory size override is used.
2021-05-23 05:54:02 +12:00
JosJuice 68a5fc55d2 Interpreter: Fix fctiwx rounding
The interpreter implementation of fctiwx was treating rounding
mode 0 as "round to nearest, ties towards zero", which is not
an actual IEEE-754 rounding mode. The IBM document mentioned
in a comment at the top of the function, on the other hand,
treats rounding mode 0 as "round to nearest, ties to even",
which makes more sense.

This fixes one of JMC's console-recorded F-Zero GX replays on
JitArm64. (JitArm64 uses an interpreter fallback for fctiwx.)
2021-05-22 17:28:04 +02:00
Mai M 1054abc9cc Merge pull request #9712 from JosJuice/jitarm64-fmul-rounding
JitArm64: Fix fmul rounding issues
2021-05-20 10:25:02 -04:00
Mai M 5949a19fe6 Merge pull request #9714 from JosJuice/jitarm64-convert-fmov
JitArm64: Prefer using FMOV when doing single/double conversion
2021-05-20 10:24:36 -04:00
Mai M 6958df5967 Merge pull request #9695 from JosJuice/jitarm64-fres
JitArm64: Implement fres and frsqrte
2021-05-20 10:23:49 -04:00
Mai M 539c2cb00e Merge pull request #9667 from Sintendo/jit64divwx2
Jit64: Minor divwx optimizations
2021-05-20 10:22:54 -04:00
JosJuice 11be2314fe JitArm64: Fix fmul rounding issues
This is a port of 4f18f60 to JitArm64.
2021-05-15 23:27:34 +02:00
JosJuice 66e912a252 PPCAnalyst: Treat frspx output as single 2021-05-15 23:27:33 +02:00
JosJuice 77afb0f4c3 PPCAnalyst: Apply "bitexact" analysis to fprIsSingle
This lets us set fprIsSingle to true in more cases.
2021-05-15 23:27:33 +02:00
JosJuice e5f2dcd891 JitArm64: Implement FPRF updates for fres+frsqrte 2021-05-15 19:21:17 +02:00
JosJuice 4b3fda7906 JitArm64: Implement frsqrte 2021-05-15 19:21:15 +02:00
JosJuice 85226e09f0 JitArm64: Implement fres 2021-05-15 19:16:32 +02:00
JosJuice 8c12068a03 JitArm64: Prefer using FMOV when doing single/double conversion
FMOV is faster than INS and ties UMOV. (On all CPUs I checked,
at least. It certainly shouldn't be slower, though.)
2021-05-15 18:56:40 +02:00
JosJuice b980797a16 PPCAnalyst: Fix broken bitexact analysis
`code` points to the first instruction in the block, not the
current instruction.
2021-05-15 18:19:04 +02:00
Mat M 24b9a64c11 Merge pull request #9690 from Sintendo/jit64divwux
Jit64: divwux - Prefer three-operand IMUL
2021-05-13 06:42:14 -04:00
JosJuice bfe8b1068d JitArm64: Implement FPRF updates 2021-05-13 11:51:00 +02:00
Sintendo 2cafa0a960 Jit64: divwux - Prefer three-operand IMUL
By taking advantage of three-operand IMUL, we can eliminate a MOV
instruction. This is a small code size win. However, due to IMUL sign
extending the immediate value to 64 bits, we can only apply this when
the magic number's most significant bit is zero.

To ensure this can actually happen, we also minimize the magic number by
checking for trailing zeroes.

Example (Unsigned division by 18)
Before:
41 BE E4 38 8E E3    mov         r14d,0E38E38E4h
4D 0F AF F5          imul        r14,r13
49 C1 EE 24          shr         r14,24h

After:
4D 69 F5 39 8E E3 38 imul        r14,r13,38E38E39h
49 C1 EE 22          shr         r14,22h
2021-05-06 19:54:33 +02:00
JosJuice b305e4cfc1 JitArm64: Fix JitRegister::Register call for cstd
Seems like I made a little copy-paste error.
2021-05-06 00:20:47 +02:00
Léo Lam 51bf2dca21 Merge pull request #9675 from JosJuice/jit64-div-80000000
Jit64: Fix UB/infinite loop when compiling division by 0x80000000
2021-04-26 23:50:27 +02:00
JosJuice 7d4b87e7ae Jit64: Fix UB/infinite loop when compiling division by 0x80000000 2021-04-26 23:42:03 +02:00
JosJuice ac679eb24d Merge pull request #9666 from leoetlino/jit-block-hashtable
Jit: Optimize block link queries by using hash tables
2021-04-25 18:45:41 +02:00
JosJuice 69c14d6ec3 JitArm64: Fix frspx with single precision source
I haven't observed this breaking any game, but it didn't match
the behavior of the interpreter as far as I could tell from
reading the code, in that denormals weren't being flushed.
2021-04-25 15:56:59 +02:00
JosJuice 54451ac731 JitArm64: Use ConvertSingleToDoubleLower in RW when faster 2021-04-25 15:56:59 +02:00
JosJuice 9d6263f306 JitArm64: Add unit tests for single/double conversion 2021-04-25 15:56:58 +02:00
JosJuice 2a9d88739c JitArm64: Skip accurate single/double conversion if store-safe 2021-04-25 15:56:58 +02:00
JosJuice 1d106ceaf5 JitArm64: Optimize ConvertSingleToDouble, part 2
If we can prove that FCVT will provide a correct conversion,
we can use FCVT. This makes the common case a bit faster
and the less likely cases (unfortunately including zero,
which FCVT actually can convert correctly) a bit slower.
2021-04-25 15:56:19 +02:00
JosJuice 018e247624 JitArm64: Optimize ConvertSingleToDouble, part 1 2021-04-25 15:56:19 +02:00
JosJuice 28e4869c43 JitArm64: Optimize ConvertDoubleToSingle 2021-04-25 15:56:19 +02:00
JosJuice 6e0a5876ef JitArm64: Use accurate single/double conversions
Our old conversion approach became a lot more inaccurate when
enabling flush-to-zero, to the point of obviously breaking games.
2021-04-25 15:56:19 +02:00
JosJuice 39eccf6603 JitArm64: Call RW before FCMPE in fselx
Needed because the next commit will make RW clobber flags.
2021-04-25 15:56:19 +02:00