libretro/ppsspp
1
0
Fork 0
mirror of https://github.com/libretro/ppsspp.git synced 2025-01-07 09:00:40 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity
10,172 Commits 19 Branches 54 Tags 247 MiB
1bfce12fdd
Commit Graph

879 Commits

Author SHA1 Message Date
Unknown W. Brackets
1bfce12fdd armjit: Report some unexpected situations. 2013-11-11 23:41:18 -08:00
Unknown W. Brackets
ac5aacbd16 armjit: Spill an imm armreg back to an imm. 
We might be able to avoid the store or etc.
 2013-11-11 23:39:13 -08:00
Unknown W. Brackets
7e19933f64 x86jit: Try predicting branch continues. 
Still doesn't seem to work.  Something like a 4% gain in Star Ocean was
the best I saw...
 2013-11-10 22:50:23 -08:00
Unknown W. Brackets
bb960480c8 x86/armjit: Stop compiling on a jump to invalid. 2013-11-10 21:59:50 -08:00
Unknown W. Brackets
fd38b10ab6 x86jit: Rename imm funcs to match armjit. 2013-11-10 21:59:49 -08:00
Unknown W. Brackets
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
Unknown W. Brackets
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
Unknown W. Brackets
92ecff4396 armjit: keep track of instructions in jitstate. 
To match x86.
 2013-11-10 21:59:48 -08:00
Unknown W. Brackets
8ceaafc159 armjit: Verify free space while compiling. 2013-11-10 21:59:48 -08:00
Unknown W. Brackets
ca7b2b554b armjit: fix major typo breaking mult/multu. 2013-11-10 21:54:44 -08:00
Unknown W. Brackets
e1fffdb37a armjit: Don't reload an armreg ptr marked noinit. 2013-11-10 16:43:38 -08:00
Unknown W. Brackets
67eaa2fd1c armjit: Optimize immediate load/stores in a row. 2013-11-10 16:32:48 -08:00
Unknown W. Brackets
bc0a846475 armjit: Optimize imm addresses (could do better...) 2013-11-10 16:30:20 -08:00
Unknown W. Brackets
c63560c0dd armjit: Try to find imms to optimize a reg load. 
This way we skip the MOVW/MOVT and go for one op only.
 2013-11-10 16:20:34 -08:00
Unknown W. Brackets
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
Unknown W. Brackets
d092f7dd2d armjit: Remember reg imm values even after flush. 
This way, we can base other imm values off them, or even do imm math using
them.  We can also avoid re-flushing an imm.
 2013-11-10 15:50:14 -08:00
Unknown W. Brackets
7f9cbc0f10 armjit: Minor cleanup and logging tweaks. 2013-11-10 15:12:40 -08:00
Unknown W. Brackets
455a7e090d Compile the cache instruction to nothing. 
Was showing up in a few profiles, does nothing currently.
 2013-11-10 14:38:10 -08:00
Unknown W. Brackets
1cc68f50ca armjit: Small optimization to syscall instr. 2013-11-10 14:38:10 -08:00
Unknown W. Brackets
b30928036e armjit: Avoid flushing an imm in beq/bne/etc. 
We might be able to STMIA it instead.
 2013-11-10 14:38:10 -08:00
Unknown W. Brackets
285ec1fad5 armjit: Implement mult/multu for immediates. 
Uncommon, but may reduce instructions a bit.
 2013-11-10 14:38:09 -08:00
Unknown W. Brackets
3a8f0598c4 x86jit: Implement wsbh/wsbw. 2013-11-10 14:38:09 -08:00
Unknown W. Brackets
9bec82873c armjit: inline byteswaps of imm values. 2013-11-10 14:38:08 -08:00
Unknown W. Brackets
06c8cb9174 armjit: Do shifts with imms as much as possible. 
This may even make an imm operand2 safe that wasn't before.
 2013-11-10 14:38:08 -08:00
Unknown W. Brackets
b2c2a87511 Fix omitted CC_AL reset, fixes #4498. 
Was breaking non-fastmem lwl/lwr/etc.
 2013-11-10 09:24:40 -08:00
Unknown W. Brackets
b310edc5f8 Fix typo in ARM debug build. 2013-11-09 15:58:27 -08:00
Henrik Rydgard
d1c012d75e ARM: Open up all 32 accessible VFP registers if NEON is available. 2013-11-09 20:18:20 +01:00
Henrik Rydgard
0a844ce98d Delete functions for vsge and vslt, these have been rolled into VecDo3 2013-11-09 19:29:52 +01:00
Unknown W. Brackets
a3a061a69f armjit: Optimize a division by a power of two. 
These really happen.
 2013-11-09 08:43:53 -08:00
Unknown W. Brackets
1776c85882 armjit: Implement a software divide for divu. 
It's not actually that much code.
 2013-11-09 08:43:52 -08:00
Unknown W. Brackets
b2a240d105 armjit: Implement msub/msubu. 2013-11-09 08:43:52 -08:00
Unknown W. Brackets
3aa8706ae7 armjit: Optimize lwl/lwr against an imm address. 2013-11-09 08:43:48 -08:00
Unknown W. Brackets
4026944b02 armjit: Handle lwl/lwr (not pretty, though.) 2013-11-09 08:42:30 -08:00
Henrik Rydgård
e90f7f360d Merge pull request #4480 from unknownbrackets/perf 
Flush regs using STMIA if possible, plus imm adjustments (armjit)
 2013-11-09 08:41:25 -08:00
Henrik Rydgard
06ce01ea04 Remove erroneous comment. 2013-11-09 17:34:52 +01:00
Unknown W. Brackets
54168b173e armjit: Clean up some magic numbers. 2013-11-09 08:25:08 -08:00
Unknown W. Brackets
6038d96b46 armjit: Flush regs using STMIA where possible. 2013-11-09 08:25:07 -08:00
Unknown W. Brackets
e686ff59bf armjit: Allocate regs in preferred slots. 
This may allow better flushing.  Not sure if these are the best regs,
but if they aren't it shouldn't really hurt.
 2013-11-09 08:25:07 -08:00
Unknown W. Brackets
cb3bb73148 armjit: Improve GPR typesafety. 2013-11-09 08:24:15 -08:00
Unknown W. Brackets
945b8bf5c5 armjit: optimize reverse subtract, avoid temp imms. 
If we have a non-op2 imm, get rid of it asap.  If we have a op2 friendly
imm, keep it.
 2013-11-09 08:18:43 -08:00
Unknown W. Brackets
415f22ecac armjit: Preserve imms on min/max as well. 2013-11-09 08:18:43 -08:00
Henrik Rydgard
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
Henrik Rydgard
58c39a38ee ARM regcache: Add mechanism to keep registers converted to pointers around 2013-11-09 16:57:29 +01:00
Henrik Rydgard
5ad04a23f4 x86 jit: Rename BindToRegister to MapReg 2013-11-09 15:23:31 +01:00
Henrik Rydgard
d26692ef92 Fix bug from a couple of commits ago in ARMJit 2013-11-09 15:22:39 +01:00
Henrik Rydgard
316d23d4cc Optimize mfv/mtv/mfc1/mtc1 on x86 too 2013-11-09 14:06:45 +01:00
Henrik Rydgard
04451623b9 This variant didn't seem to make much difference either (see prev commit) 2013-11-09 13:06:10 +01:00
Henrik Rydgard
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
Unknown W. Brackets
5d46a82f43 armjit: Use a MOV for add/or with 0. 
Might skip the ALU, so might be faster.
 2013-11-08 11:41:57 -08:00
Unknown W. Brackets
b8e126e7ce armjit: Preserve imms in slt/sltu as possible. 2013-11-08 11:41:57 -08:00