Commit graph

53 commits

Author SHA1 Message Date
M4xw
b9352354c9 Masked PSP Memory support for the AArch64 Dynarec 2019-04-15 12:07:57 +02:00
Unknown W. Brackets
ae63628360 arm64jit: Statically allocate ra as well.
This doesn't seem to have a significant impact on performance, but it
improves bloat by about 5%.
2017-12-30 11:11:03 -08:00
Unknown W. Brackets
e7ac672522 arm64jit: Cleanup method names, temp discard.
This way MapDirtyIn won't accidentally discard temps.
2017-12-30 07:51:27 -08:00
Unknown W. Brackets
98ed6fab3f arm64jit: Fix spilling for more than one temp reg.
Otherwise we hang trying to spill the same reg over and over.
2017-12-29 17:30:17 -08:00
Unknown W. Brackets
c71285c970 arm64jit: Use CBZ/CBNZ for zero compare branches.
These are pretty common, so it reduces bloat decently.  Seems about the
same speed, though.
2017-12-29 17:30:15 -08:00
Unknown W. Brackets
2498ce5e3e arm64jit: Oops, properly init temp locked flag.
Fixes #10469.
2017-12-29 14:36:18 -08:00
Unknown W. Brackets
27116dcb86 arm64jit: Avoid flushing when mapping as pointer. 2017-12-28 16:04:34 -08:00
Unknown W. Brackets
1b1e2c773b arm64jit: Jit lwl/lwr with proper temp regs.
It's possible rt might overlap with w9/w10, so we really need to allocate
these properly.  This locks and spills as necessary.
2017-12-28 15:54:03 -08:00
Unknown W. Brackets
1b792c32e1 arm64jit: Attempt to reuse imms on sw/etc.
Mostly, this handles the zero case, but it may help in other cases too.
2017-12-28 12:32:12 -08:00
Unknown W. Brackets
c4c28282cf arm64jit: Allow ARMREG_AS_PTR to be dirty.
Since we can just add/sub, then it should be valid (as long as we only
offset it when dirty.)
2017-12-28 10:40:31 -08:00
Unknown W. Brackets
7c2fc90def arm64jit: Avoid MOVK elsewhere without pointerify. 2017-12-27 17:57:19 -08:00
Unknown W. Brackets
d82efc4b0b arm64jit: Allow static alloc without pointerify. 2017-12-27 17:50:15 -08:00
Unknown W. Brackets
9573a791b4 arm64jit: Skip storing spilled but not dirty.
Unless IMM, we don't need to store non-dirty mapped regs.
2017-12-27 17:15:18 -08:00
Unknown W. Brackets
cccf448ae0 arm64jit: Allow disabling pointerification.
For platforms where we can't get base aligned.
2017-12-27 17:02:19 -08:00
Florent Castelli
8c3552de74 cmake: Detect features at compile time
Instead of relying on manually passed down flags from CMake,
we now have ppsspp_config.h file to create the platform defines for us.
This improves support for multiplatform builds (such as iOS).
2016-10-19 12:31:19 +02:00
Henrik Rydgard
f6ebedba78 Fix some build errors. 2016-10-10 22:01:20 -07:00
Unknown W. Brackets
8713e532e0 arm64jit: Log if HI is set an an imm. 2016-01-20 23:26:49 -08:00
Unknown W. Brackets
387ebf3ddd arm64: Only clobber HI and LO together.
We can't clobber LO if HI is still needed.  Fixes a crash in Trails in
the Sky on new game (due to a mfhi getting the wrong result.)
2015-10-17 01:21:51 -07:00
Henrik Rydgard
42ce836ba5 Some minor renames and cleanups 2015-10-08 14:43:16 +02:00
Henrik Rydgard
5f4145238a Fix #8034. Cannot leave IMMs or REGIMM in statically allocated registers after a FlushAll. 2015-10-08 14:39:21 +02:00
Unknown W. Brackets
9b94266c4a arm64: Avoid overflow in imm math. 2015-08-30 13:00:18 -07:00
Henrik Rydgard
9fe382ad18 ARM64: Shortcut moves between gpr and vfpu when not mapped 2015-07-11 23:53:06 +02:00
Henrik Rydgard
b73920fcba ARM64: Cleanups in regcache and jit 2015-07-11 17:00:01 +02:00
Henrik Rydgard
444fc0885c Fix the crash bug (need to mark static non-pointer regs as depointerified on flush) 2015-07-11 16:59:14 +02:00
Henrik Rydgard
2a8560e522 ARM64: Another pair of fixes for static alloc. Still crashes in many games... 2015-07-11 16:59:14 +02:00
Henrik Rydgard
d1bbc1d3c0 More regcache fixes. ML_IMM works now although there is another stability issue somewhere. 2015-07-11 16:59:13 +02:00
Henrik Rydgard
1b8549b26f Couple more regcache fixes 2015-07-11 16:59:12 +02:00
Henrik Rydgard
4920f3e3c1 Work towards handling ML_IMM in static registers but doesn't work yet (enable on line 649) 2015-07-11 16:59:11 +02:00
Henrik Rydgard
f42f81a4da ARM64: Optimize small adjustments of pointerified registers 2015-07-11 16:59:11 +02:00
Henrik Rydgard
9af6abd8a1 ARM64: Support pointerified static allocs, statically allocate SP 2015-07-11 16:59:10 +02:00
Henrik Rydgard
844a3f19a9 Fix bugs and issues in ARM64 static alloc. Thanks unknown for finding the problem. 2015-07-11 16:59:09 +02:00
Henrik Rydgard
a0bf934796 ARM64: Some work on static allocation. Close to working, cube.elf runs 700 blocks but then hangs (?!) 2015-07-11 16:59:09 +02:00
Unknown W. Brackets
db3dffb44d arm64: Oops, fix flushing zero from an armreg. 2015-07-05 11:57:18 -07:00
Unknown W. Brackets
003668fe66 armjit: Fix discarding imms. 2015-07-04 07:30:32 -07:00
Unknown W. Brackets
8ea7f99072 arm64: Fix imm wasting when STP doesn't work out. 2015-07-04 07:09:47 -07:00
Unknown W. Brackets
e6a7ba3fae arm64: Bring imms along for the STP ride. 2015-07-03 16:51:33 -07:00
Unknown W. Brackets
ca1e482a56 arm64: Avoid setting a reg to zero to store it. 2015-07-03 16:05:25 -07:00
Unknown W. Brackets
90b7d135cb arm64: Flush in pairs if possible.
On an A57, this is around twice as fast (for just the STR/STR vs STP.)
2015-07-03 11:07:09 -07:00
Unknown W. Brackets
ddb955a527 arm64: Try to optimize imm stores.
If we already have a reg, we can use it.  This can happen when immediate
addresses are loaded and used as bases, although it's not super common.
2015-07-03 10:48:11 -07:00
Unknown W. Brackets
2331df8c70 arm64: Try to be more consistent in ZERO handling.
Let's keep it IMM where possible, even though we've added checks for
MIPS_REG_ZERO.
2015-07-03 10:21:24 -07:00
Unknown W. Brackets
66d85233b9 arm64: Flush only caller-saved regs before calls. 2015-07-03 10:09:43 -07:00
Unknown W. Brackets
66adc4e695 jit: Normalize CONDITIONAL_DISABLE formatting. 2015-07-02 20:31:37 -07:00
Unknown W. Brackets
fed687fb59 arm64: Meld LO and HI together for multiplies. 2015-07-02 20:31:37 -07:00
Unknown W. Brackets
1d1c80d9cf arm64: Use BFI for cfc1. 2015-07-02 20:31:35 -07:00
Henrik Rydgård
d014d420db Unify JitOptions across the backends.
This is required to make ExtractIR not a member of the various backends.
2015-04-12 11:41:26 -07:00
Henrik Rydgard
ad648baa9c ARM64 regcache: Add support to "pointerify" registers. Use in load/store to cut down instructions. 2015-04-06 18:13:32 +02:00
Henrik Rydgard
77501e220d ARM64: Enable a few more instructions, more emitter/disasm unittests 2015-04-06 18:13:13 +02:00
Henrik Rydgard
0922db6062 ARM64: Some FP work. 2015-04-06 18:13:11 +02:00
Henrik Rydgard
58b059ca14 Some casting cleanup, misc 2015-04-06 18:13:06 +02:00
Henrik Rydgard
34ab532971 ARM64 jit: Fix some pointer truncations and other fun stuff.. 2015-04-06 18:13:05 +02:00