3957b5fb42
Call ProxyBlock in ArmJit too. Plus various cleanup and one more string function.
2013-12-19 11:45:39 +01:00
438361d0bc
Clean up code pointer naming for the jit.
...
Now it properly identifies thunk code which is actually a decent percent
when fastmem is off at least.
2013-12-18 23:57:39 -08:00
1cb7965cb1
Jit feature preparation: Introduce "proxy blocks".
...
When these are invalidated, the block they point to gets invalidated too.
Will be useful to implement various types of block merging and function inlining
without affecting correctness of cache clears etc.
Also, with this commit we can now fully inline replaced functions. fabsf() boils
down to 1-2 instructions and the block continues, for example.
2013-12-19 00:39:49 +01:00
6937719c3b
Minor fixes around function replacement, a couple of math funcs
2013-12-18 10:35:50 +01:00
1d2f1efd06
Fix function replacement on ARM
2013-12-18 00:09:08 +01:00
2eab4aa1bf
Play around with function replacement. Turned off by default of course.
2013-12-17 23:40:27 +01:00
2140892074
Initial preparations for ability to replace game functions with custom implementations.
...
Also auto-saves hashmap additions and reapplies the hashmap on function
rename so that if you rename a function that exists in several copies
they will all be labelled.
Note that actual function replacement is not activated yet.
2013-12-17 12:27:20 +01:00
2d8429ac48
Assorted cleanup in the MIPS emulation
2013-12-10 13:15:16 +01:00
030e6460cc
ARM: NEON-optimize software skinning
2013-11-24 18:03:42 +01:00
ab3037112f
Some scaffolding for a future VFPU-on-NEON implementation
2013-11-19 21:41:48 +01:00
dca457e6df
Optimize multiple sv.s and lv.s calls on ARM. Also some cleanup.
2013-11-19 21:41:47 +01:00
e3bdb3e09b
Disable LitPool as it is causing crashes with Vertex Decoder JIT. Performance seems to be almost unaffected since the IMM changes.
2013-11-15 14:12:00 +10:00
20e8a81268
Switch to compile-time ARMV7 define.
2013-11-15 11:20:39 +10:00
359110f010
x86/armjit: Add jump following (off by default.)
...
Inlines function calls up to a certain extent. Allows us to get
immediates all the way to a syscall, for example, usually.
Not sure if faster.
2013-11-10 21:59:49 -08:00
aacb31bc18
armjit: Copy over (disabled) immbranch optim.
...
This does a little loop unrolling. Costs a bit more cache space, but
avoids flushing regs for longer.
Not enabled.
2013-11-10 21:59:48 -08:00
92ecff4396
armjit: keep track of instructions in jitstate.
...
To match x86.
2013-11-10 21:59:48 -08:00
8ceaafc159
armjit: Verify free space while compiling.
2013-11-10 21:59:48 -08:00
7e46ee0b0f
armjit: Replace MOVI2R with using the regcache.
...
So that it can optimize the value with existing imms.
Not actually optimizing yet.
2013-11-10 15:50:45 -08:00
0a844ce98d
Delete functions for vsge and vslt, these have been rolled into VecDo3
2013-11-09 19:29:52 +01:00
502f772856
Add experimental mode to cache pointers in the arm jit.
...
Turned off for now as it needs more work but seems quite promising already.
2013-11-09 17:15:30 +01:00
d26692ef92
Fix bug from a couple of commits ago in ARMJit
2013-11-09 15:22:39 +01:00
04451623b9
This variant didn't seem to make much difference either (see prev commit)
2013-11-09 13:06:10 +01:00
15bc5a8db7
Add small ARM perf experiment. Did not help on ARMv7 so turned it off.
...
xsacha might want to try it on ARMv6.
2013-11-09 12:57:07 +01:00
58db79672f
Fix vmtvc on ARM, fixing issues with our prefix check. Add some logging.
...
Also improve vcmp on ARM.
2013-11-08 19:59:11 +01:00
309f904c0c
Extract JitState into its own header (arm/x86)
2013-11-08 18:51:52 +01:00
803148b8ca
ARMv6: Fix offsets > 4096 for litpool. More aggressive check.
...
Somehow Scooby Doo gets to offsets of ~4200 unless i drop the threshold down to ~3200. Not sure why the offset can jump by so much in one instruction.
Makes Scooby Doo playable now instead of showing a blue screen in the main game. Likely affects other games.
2013-11-08 16:07:05 +10:00
50e9e45d65
Check version in each DoState() func.
...
They bail on PointerWrap error or bad version.
2013-09-14 20:23:03 -07:00
324cde5a79
Let's actually use the log category mechanism. A first step.
2013-09-07 21:19:21 +02:00
78d3ee3d6a
Misc cleanup, mostly logging code
2013-09-07 13:01:19 +02:00
b558189c37
Just invalidate blocks on ClearCacheAt().
...
This makes it safe to call from a jitted syscall, etc.
2013-09-01 00:32:43 -07:00
109ad17ac6
Use a typesafe struct for opcodes.
...
Also, correctly read delayslots using Read_Instruction on ARM.
2013-08-24 15:36:24 -07:00
b37f09cedf
Make MIPSInfo a struct for typesafety.
...
Found a bug in ReadsFromReg().
2013-08-24 13:22:10 -07:00
8327cd0f8e
Clean up some inconsistency in jit branches on arm.
2013-08-16 02:02:56 -07:00
0a8f85a919
Some JIT cleanup, implement VI2F on ARM. also disabled untested impl of viim for x86.
2013-07-31 17:27:04 +02:00
51596b636a
Fix numerous ARM JIT bugs. Activate vmtvc and vscl, and vadd/vmul/vdiv/vsub for real this time.
2013-07-31 10:34:58 +02:00
e93c2abe27
x86 jit: implement vfim. Move some stuff to native. cleanup for armjit logging
2013-07-30 22:28:05 +02:00
76a937f489
ARMJIT Experiment: Keep downcount in a register. Needs benchmarking.
2013-07-27 17:27:26 +02:00
78fe42fe80
Buildfix
2013-04-27 12:04:42 +02:00
2a39a3b972
JIT: Get rid of one memory access per dispatch, and get rid of blockcodepointers.
2013-04-27 01:32:03 +02:00
9eace8a80e
Combine the two JitCache implementations (x86, ARM) into one.
2013-04-27 01:32:03 +02:00
8a904fe478
Some JIT cleanup
2013-04-27 01:32:02 +02:00
5293c152c6
Untested support for Android-x86. No idea if this actually works.
2013-03-21 20:52:33 +01:00
9561f1d9c1
iOS: Protect/unprotect isn't needed for jit.
2013-03-14 02:16:51 -07:00
ead6983ad2
Don't lock armjit memory before aligning code.
2013-03-14 01:16:28 -07:00
f88bf8bbff
Attempt at fixing JIT on iOS. Can only have PROT_WRITE or PROT_EXEC enabled. So toggle between them as needed.
2013-03-14 14:40:01 +10:00
c4ab0855b4
Make sure interpreter and jit savestates match.
2013-03-08 08:49:21 -08:00
028e85dc92
Cleanup some differences between the two jits.
2013-03-07 02:08:44 -08:00
a69b09a831
More work on TestRunner, now runs all the CPU tests.
2013-03-07 00:22:39 +01:00
d77632bfb0
Fix literal pools on games with very large code blocks (eg. Zero no Kieski).
...
Was flushing after an offset of 4088 which did not take in to account that a single MIPS instruction can turn in to numerous ARM instructions. Chose a safer value of 4020.
Was insta-flushing after reaching this offset value. Some code blocks are over 8K in size. Use a partialFlushOffset to keep track of when the next flush is required.
Was protecting flush branch manually. Can use B_CC(CC_AL) for this instead.
2013-03-07 02:25:27 +10:00
268d16bd24
Use correct args for STR(..) throughout armjit.
2013-03-07 00:59:07 +10:00
Henrik Rydgård