steam-frame/FEX-Emu--FEX
1
0 Fork 0
mirror of https://github.com/FEX-Emu/FEX.git synced 2026-10-14 23:00:44 +02:00
Code Issues Packages Projects Releases 67 Wiki Activity
9,533 Commits 11 Branches 71 Tags
c28824f94de6cb3e83becbbcd9473cfc9d2bfd22
Commit Graph
1283 Commits
Author SHA1 Message Date
Alyssa Rosenzweig 974baca09c OpcodeDispatcher: allow upper garbage with rcl/rcr smaller 
we're masking immediately to something smaller

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
 2024-01-30 22:22:57 -04:00
Alyssa Rosenzweig f22094a493 OpcodeDispatcher: use a branch for 8/16-bit rotate flags 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
 2024-01-30 22:22:57 -04:00
Alyssa Rosenzweig d979b3a1da OpcodeDispatcher: note idea to further optimize rcl 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
 2024-01-30 22:22:57 -04:00
Alyssa Rosenzweig 6d82c957fa OpcodeDispatcher: fuse orlshl in rcl 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
 2024-01-30 22:22:57 -04:00
Mai fa3352004e Merge pull request #3381 from alyssarosenzweig/opt/masking 
Allow upper garbage on a bunch of instructions
 2024-01-30 10:07:53 -05:00
Ryan Houdek ce2924731e vixl/simulator: Enlarge simulator stack size 
Simulator stack size defaults to 8KB. This new unit test requires at
least 15360 stack size. Just push it up to 8MB.
 2024-01-29 19:48:38 -08:00
Ryan Houdek bc67910ee4 Merge pull request #3382 from pmatos/TypoFix 
Fix typos; NFC
 2024-01-29 16:10:14 -08:00
Mai 31a4158957 Merge pull request #3383 from alyssarosenzweig/opt/ptest 
Optimize PTEST and VTESTP
 2024-01-29 13:30:53 -05:00
Mai 58f3d3caf5 Merge pull request #3380 from alyssarosenzweig/opt/pdep 
Optimize PDEP
 2024-01-29 13:27:15 -05:00
Alyssa Rosenzweig ae48228943 OpcodeDispatcher: optimize vtestps/vtestpd 
I don't really care about AVX but do the same thing we did for vptest.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
 2024-01-29 13:24:11 -04:00
Alyssa Rosenzweig e8e35e48c7 OpcodeDispatcher: optimize ptest with tst 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
 2024-01-29 13:20:17 -04:00
Alyssa Rosenzweig 8b8f27a88f OpcodeDispatcher: optimize ptest with umaxv 
to check if the vector is zero, umaxv its elements and check if the reduced
scalar is zero.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
 2024-01-29 13:20:17 -04:00
Alyssa Rosenzweig 8e7906a665 IR: add UMaxV 
will be used to accelerate ptest

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
 2024-01-29 13:19:22 -04:00
Paulo Matos 027fbbf051 Optimize CDQOp 2024-01-29 17:18:02 +00:00
Paulo Matos ca31a0404c ConstProp should generate 32bit constants when required 2024-01-29 17:15:47 +00:00
Paulo Matos f644959c7c Fixing some typos; NFC 2024-01-29 17:14:53 +00:00
Alyssa Rosenzweig 16a54742e6 OpcodeDispatcher: optimize 32-bit tzcnt 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
 2024-01-29 13:11:25 -04:00
Alyssa Rosenzweig ad9aa0bc87 OpcodeDispatcher: optimize 32-bit lzcnt 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
 2024-01-29 13:11:25 -04:00
Alyssa Rosenzweig bd2b3f35a3 OpcodeDispatcher: optimize 32-bit popcnt 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
 2024-01-29 13:11:25 -04:00
Alyssa Rosenzweig 50169ce640 OpcodeDispatcher: optimize 32-bit pext 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
 2024-01-29 13:11:25 -04:00
Alyssa Rosenzweig baae2d68f9 OpcodeDispatcher: optimize 32-bit bextr 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
 2024-01-29 13:11:25 -04:00
Alyssa Rosenzweig ad8d038b8a OpcodeDispatcher: optimize 32-bit blsi 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
 2024-01-29 13:11:25 -04:00
Alyssa Rosenzweig 820932e3c7 OpcodeDispatcher: optimize 32-bit blsmsk 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
 2024-01-29 13:11:25 -04:00
Alyssa Rosenzweig 6f11f2e6f4 OpcodeDispatcher: optimize 32-bit blsr 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
 2024-01-29 13:11:25 -04:00
Alyssa Rosenzweig f5ad7682c3 OpcodeDispatcher: optimize 32-bit pdep 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
 2024-01-29 13:06:56 -04:00
Alyssa Rosenzweig 04805f351b JIT: rewrite pdep implementation 
- use better algorithm that is O(# set bits) instead of O(# total bits)
- eliminate spilling by careful management of our temporaries
- fix nzcv clobber bug (whoops)

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
 2024-01-29 13:06:56 -04:00
Mai 750b0b70bc Merge pull request #3356 from Sonicadvance1/modify_code_lock 
Jitarm64: Implements spin-loop futex for JIT blocks
 2024-01-23 13:46:59 -05:00
Ryan Houdek 56d8080ec9 Merge pull request #3345 from Sonicadvance1/fix_syscall_registers 
OpcodeDispatcher: Fixes syscall rcx/r11 generation
 2024-01-22 15:21:13 -08:00
Ryan Houdek c0be974272 Merge pull request #3368 from bylaws/preprcr 
FEXCore: Fix RCL/RCR shift wraparound behaviour
 2024-01-21 13:44:49 -08:00
Billy Laws e323938173 FEXCore: Fix RCL/RCR shift wraparound behaviour 
This ends up being cleaner to handle outside of
CalculateFlags_ShiftVariable as constant masking is only needed for
RCL/RCR.
 2024-01-21 18:15:50 +00:00
Billy Laws 407e26bfee FEXCore: Use TMP1-4 for values that need preserving across spills 
The ARM64EC SRA layout will use x0-3 for x86_64 registers, as such any
arguments passed to C ABI functions need to proxy their arguments
through the temporaries and move as appropriate.
 2024-01-21 16:21:13 +00:00
Ryan Houdek e18453cb57 Jitarm64: Implements spin-loop futex for JIT blocks 
This will ensure that multiple concurrent SIGBUS handlers in the same
code block doesn't modify the same code.
 2024-01-17 10:19:38 -08:00
Ryan Houdek f956f008ea Merge pull request #3372 from alyssarosenzweig/opt/cmpxchg-review 
Optimize GPR cmpxchg
 2024-01-15 05:11:12 -08:00
Ryan Houdek 1f7a619c79 OpcodeDispatcher: Fixes syscall rcx/r11 generation 
Noticed this while writing #3342.

Fixes #3343

The syscall instruction is defined in the documentation that it will set
RCX to the next instruction's RIP and R11 to be RFLAGS. We entirely
skipped this which I noticed while writing unit tests.

Adds unittests to test both 32-bit and 64-bit behaviour because our
helper shares code with both.

I don't know if anything actually relied on this behaviour but we should
definitely support it.
 2024-01-12 19:14:30 -08:00
Alyssa Rosenzweig 58127bd0e8 OpcodeDispatcher: optimize trivial cmpxchgs 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
 2024-01-12 12:23:34 -04:00
Alyssa Rosenzweig e8945dfb6d OpcodeDispatcher: optimize gpr cmpxchg 
NZCV stuff.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
 2024-01-12 12:03:28 -04:00
Ryan Houdek 8c3163096b Merge pull request #3363 from Sonicadvance1/fix_label_allocations 
ArmEmitter: Support single use forward labels
 2024-01-12 00:26:31 -08:00
Ryan Houdek 615cfe0246 Merge pull request #3361 from Sonicadvance1/decompose_std_function 
FEXCore: Decompose some std::function usage to regular pointers
 2024-01-10 16:55:29 -08:00
Ryan Houdek 3d5f876585 Fixes some new glibc allocations that cropped up 
I guess this was handled by brk things before.
 2024-01-09 13:55:04 -08:00
Ryan Houdek 37102400b5 Arm64: Switches uses of forward label over to SingleUse if possible 
Primary goal for this is to ensure that the delinker doesn't need to
allocate any memory. This delinker can end up getting hit heavily with
JIT code so we don't want it to be allocating memory.
 2024-01-08 22:18:20 -08:00
Ryan Houdek c01e6283ae CodeEmitter: Support a single use forward label 
Currently all uses of the forward label calls in to jemalloc to allocate
memory. This allows a forward label that doesn't require any memory
allocation, which is the common case in FEX.
 2024-01-08 22:18:20 -08:00
Ryan Houdek 248dc97993 FEXCore: Decompose some std::function usage to regular pointers 
The delinker step of the JIT was using std::function with capture
lambdas that required memory allocation when unnecessary.
Because the compiler can't see through our std::function usage it could
never decompose these by itself.

By passing the Thread's frame and record to the function as arguments
then we can have the signature be a raw function pointer.

This fixes an area of concern from:
https://github.com/FEX-Emu/FEX/blob/main/docs/ProgrammingConcerns.md#stdfunction-and-lambdas
 2024-01-06 19:39:54 -08:00
Ryan Houdek 743df8dfae Merge pull request #3327 from Sonicadvance1/remove_syscall_indirection 
Arm64: Removes a vtable indirection in syscalls
 2024-01-04 18:19:40 -08:00
Ryan Houdek 4b3792196f Merge pull request #3303 from Sonicadvance1/initial_runtime_longmode_switch 
OpcodeDispatcher: Initial support for runtime long-mode switch
 2024-01-04 18:17:54 -08:00
Ryan Houdek db7d7a6bd7 Merge pull request #3349 from Sonicadvance1/revert_frontend_ownership 
Revert "FEXLoader: Moves thread management to the frontend"
 2024-01-03 14:25:04 -08:00
Alyssa Rosenzweig 04a88ed3ab Merge pull request #3353 from Sonicadvance1/public_interface_cleaning 
FEXCore interface cleaning
 2024-01-03 15:14:54 -04:00
Alyssa Rosenzweig 9da08b40bd Merge pull request #3344 from Sonicadvance1/xbyak_upstream 
Externals: Update xbyak to v7.02 and switch away from fork
 2024-01-03 15:13:58 -04:00
Alyssa Rosenzweig 5467c3e478 Merge pull request #3357 from Sonicadvance1/remove_non_sra 
FEXCore: Removes SRA option, it's now permanently enabled
 2024-01-03 15:10:04 -04:00
wannacu 4e7bab849c JIT: Fixes broken register in VTBX1 
If the Dst register is allocated as VectorIndices or VectorTable,
using Dst as an operand to perform the tbx operation will result in an error.
For example:
%131(FPR0) i128 = LoadNamedVectorIndexedConstant u8:Tmp:RegisterSize, #0x6, #0xaa0
%132(FPR0) i128 = VTBX1 u8:Tmp:RegisterSize, %129(FPRFixed6) i32v4, %126(FPRFixed10) i16v8, %131(FPR0) i128
Since the tbx instruction's destination register is also the original operand,
this is consistent with the semantics of VTBX1. Therefore,
directly using VectorSrcDst as the destination operand for the tbx instruction is safe.
 2023-12-29 16:18:40 +08:00
Ryan Houdek d098545c20 FEXCore: Removes SRA option, it's now permanently enabled 2023-12-28 18:28:02 -08:00