Commit Graph

12505 Commits

Author SHA1 Message Date
comex
6695b5acce Fix backwards #ifdef. 2014-09-17 22:30:20 -04:00
comex
97c9cb5882 Add missing push wrapper around UpdatePerformanceMonitor 2014-09-17 21:10:43 -04:00
comex
7ad9027593 Be pedantic about stack overflow on Linux and OS X.
Add some magic to the fault handler to handle stack overflow due to BLR
optimization, and disable the optimization if fastmem is not enabled.
2014-09-17 20:08:09 -04:00
comex
755bd2c445 Reorganize backpatching a bit. Untested on ARM.
Rather than *MemTools.cpp checking whether the address is in the
emulated range itself (which, as of the next commit, doesn't cover every
kind of access the JIT might want to intercept) and doing PC
replacement, they just pass the access address and context to
jit->HandleFault, which does the rest itself.

Because SContext is now in JitInterface, I wanted JitBackpatch.h (which
defines it) to be lightweight, so I moved TrampolineCache and associated
x64{Analyzer,Emitter} dependencies into its own file.  I hate adding new
files in three places, two of which are MSVC...

While I'm at it, edit a misleading comment.
2014-09-17 19:57:06 -04:00
comex
7b0fdb52cd Run exception handlers on an alternate stack on Linux.
*Completely untested.*  Someone please test.
2014-09-17 19:57:04 -04:00
comex
bd4e75e69a Shorten the blr stub a bit. 2014-09-17 19:56:58 -04:00
comex
b597ec3e08 Opportunistically predict BLR destinations using RET.
When executing a BL-type instruction, push the new LR onto the stack,
then CALL the dispatcher or linked block rather than JMPing to it.  When
executing BLR, compare [rsp+8] to LR, and RET if it's right, which it
usually will be unless the thread was switched out.  If it's not right,
reset RSP to avoid overflow.

This both saves a trip through the dispatcher and improves branch
prediction.

There is a small possibility of stack overflow anyway, which should
be handled... *yawn*
2014-09-17 19:56:09 -04:00
comex
558dee84ca Wrap some function calls in ABI_Push|PopRegistersAndAdjustStack(0, 0);
These calls are made outside of JIT blocks, and thus previously did not
read any protection - register use is taken into account and the outer
dispatcher stack frame is sufficient.  However, if data is to be stored
on the stack, these calls must reserve stack shadow space on Windows to
avoid clobbering it.
2014-09-17 19:56:09 -04:00
Fiora
76c8bb40e8 JIT: update some rather outdated comments 2014-09-16 23:42:22 -07:00
shuffle2
978a855d3f Merge pull request #1090 from shuffle2/dolphin-qt
DolphinQt -- initial commit
2014-09-16 22:21:16 -07:00
skidau
2c233c4976 Merge pull request #686 from FioraAeterna/fiora
JIT: Optimize JitAsmCommon, Float, and PS implementations
2014-09-17 14:06:14 +10:00
Fiora
d3dee1d7ed GekkoDisassembler: fix some float opcodes 2014-09-16 02:06:40 -07:00
comex
7eea7080d9 Fix missing "return" 2014-09-16 00:25:21 -04:00
skidau
8361d2b1da Merge pull request #805 from FioraAeterna/storerefactor
JIT: support immediate stores
2014-09-16 13:31:39 +10:00
Dolphin Bot
bef2016909 Merge pull request #1091 from FioraAeterna/fixdisasm
GekkoDisassembler: fix/improve disassembly for a few instructions
2014-09-16 03:53:18 +02:00
Fiora
7368c2ee9e GekkoDisassembler: fix/improve disassembly for a few instructions 2014-09-15 18:48:54 -07:00
Shawn Hoffman
9d57ac68a4 Add Qt submodule for windows. 2014-09-15 15:07:42 -07:00
Shawn Hoffman
ae3a5ce9e3 Qt: Add msvc support 2014-09-15 15:07:33 -07:00
Augustin Cavalier
16c6a19190 DolphinQt: initial commit.
This adds the beginning of the DolphinQt user interface. It doesn't
do anything useful yet and only builds via CMake.
2014-09-15 15:06:05 -07:00
comex
847f78e4cc Merge pull request #1089 from FioraAeterna/fixemittersilentfail
x64Emitter: fix silent failure if WriteNormalOp is passed two memory operands
2014-09-15 14:53:21 -04:00
Fiora
764ce4b513 JIT: add fselx implementation
Based on a patch by Tilka.
2014-09-15 08:33:41 -07:00
Fiora
7ab820c6f8 JIT: Various JitAsmCommon optimizations
Use some SSE4 instructions in on CPUs that support them.
Use float instructions instead of int where appropriate (it's a cycle faster
on CPUs with arithmetic unit forwarding penalties).
2014-09-15 08:33:40 -07:00
Fiora
7b0f559ae1 JIT: various float optimizations 2014-09-15 08:33:40 -07:00
Fiora
34287b8042 JIT: some paired singles optimizations 2014-09-15 08:33:39 -07:00
Fiora
d02b7c7755 JIT: support immediate stores 2014-09-15 07:25:32 -07:00
Fiora
02dce5dbbf x64Emitter: fix silent failure if WriteNormalOp is passed two memory operands
Should now fail loudly and clearly instead.
2014-09-15 07:08:08 -07:00
Fiora
a248e49c97 Revert "Two small JIT optimizations" 2014-09-15 07:01:51 -07:00
Ryan Houdek
20af50b1c4 Merge pull request #1077 from FioraAeterna/integeropts2
Two small JIT optimizations
2014-09-15 07:09:11 -05:00
skidau
1458b59156 Merge pull request #1065 from Sonicadvance1/AArch64-fix-regcache
Fix all the current issues with the AArch64 register cache.
2014-09-15 13:02:16 +10:00
Fiora
3845c27155 JIT: save an instruction in psq_l 2014-09-14 15:12:57 -07:00
Fiora
af471d0a84 JIT: optimize andi(s)_rc
We usually don't need to do a sign-extend for the resulting flags.
2014-09-14 15:12:14 -07:00
comex
74f8a48ee6 Merge pull request #1063 from FioraAeterna/fastermmufix
MMU: properly check MEM1 range on Gamecube games
2014-09-14 15:17:09 -04:00
comex
db7617248f Merge pull request #1021 from FioraAeterna/optimizeca3
JIT: Carry optimizations!
2014-09-14 15:08:08 -04:00
Lioncash
b307bb68d7 Merge pull request #1080 from lioncash/wx-crap
Kill off the wx casts within InputCommon and GLInterface.
2014-09-14 13:21:34 -04:00
Ryan Houdek
4e7f284a81 Merge pull request #1064 from Sonicadvance1/AArch64-Fix-MOVI2R
Fix AArch64 MOVI2R helper function.
2014-09-14 09:26:02 -05:00
Ryan Houdek
4e16abd742 Merge pull request #1083 from FioraAeterna/lzcnt
Add LZCNT support, use in cntlzw
2014-09-14 09:18:10 -05:00
Fiora
40b18f09b2 JIT: use LZCNT in cntlzw 2014-09-14 05:36:23 -07:00
Fiora
997c5c2d0e x64Emitter: add LZCNT/TZCNT support and detection
Also add a unit test.
2014-09-14 05:31:22 -07:00
Ryan Houdek
1f7871f9e2 Merge pull request #1082 from FioraAeterna/fixccagain
JIT: fix JIT with conditional continue off, again
2014-09-14 03:22:03 -05:00
Fiora
653140fcfa MMU: properly check MEM1 range on Gamecube games
Fixes at the very least Rogue Squadron II crashes.
2014-09-14 00:16:37 -07:00
Fiora
9d2b2cae4a JIT: fix JIT with conditional continue off, again 2014-09-14 00:12:27 -07:00
Lioncash
357c0adc3c Kill off the wx casts within InputCommon and GLInterface.
All because someone didn't actually return the wxWindow handle for the edge case.
2014-09-14 01:07:08 -04:00
Scott Mansell
6956df79b2 Fix Fastmem in JitIL for massive speed gains.
JitIL's fastmem was stubbed out when Sonicadvance1 merged JitARMIL
into the tree. Since JitARMIL has been deleted, I simply re-arrange
the inheritance to base JitIL on Jitx86Base, so it can inherit the
backpatch function.

Povray Benchmark: 1985 seconds to 1316 seconds.
2014-09-14 16:21:54 +12:00
Lioncash
6cfc5b9849 InputCommon: Initialize NSDictionaries using literal syntax. 2014-09-13 21:32:33 -04:00
Lioncash
068799ff73 UnitTests: Fix misplaced braces in the x64Emitter tests 2014-09-13 20:45:17 -04:00
Pierre Bourdon
84a564a304 Merge pull request #1060 from rohit-n/move-using
JitArm32: Minor cleanup.
2014-09-14 01:59:39 +02:00
Pierre Bourdon
eee473ed5f Merge pull request #1066 from lioncash/breaks
DiscIO: Remove unnecessary break statements.
2014-09-14 01:58:44 +02:00
Pierre Bourdon
9607cc4159 Merge pull request #1071 from archshift/patch-1
ConfigManager.cpp: Removed three elements from platform-ifdef
2014-09-14 01:58:17 +02:00
Pierre Bourdon
439068acae Merge pull request #1055 from FioraAeterna/smallermov
X64Emitter: support shorter mov reg, imm opcodes
2014-09-14 01:57:36 +02:00
Pierre Bourdon
cb2e53e7c6 Merge pull request #1068 from rohit-n/use-vector
CBoot: Use vectors instead of raw pointers in ELF functions.
2014-09-14 01:55:01 +02:00