Stefanos Kornilios Mitsis Poiitidis
8642406b8f
CPUID: Add cache information, function 0x2
2021-02-23 11:01:15 +02:00
Stefanos Kornilios Mitsis Poiitidis
e2599db3ed
Merge pull request #793 from Sonicadvance1/assert_on_missing_lock
...
Adds assert checks on missing LOCK support
2021-02-23 10:51:30 +02:00
Ryan Houdek
bb0850eab8
Adds assert checks on missing LOCK support
...
These usually don't get hit, but Geekbench4 DOES manage to hit LOCK on
BTS.
Which we just don't support right now.
2021-02-22 23:20:18 -08:00
Stefanos Kornilios Mitsis Poiitidis
1af541475a
Merge pull request #795 from Sonicadvance1/atomic_bittest_ops
...
Implements BTC, BTR, BTS atomic variants
2021-02-23 09:08:39 +02:00
Stefanos Kornilios Mitsis Poiitidis
6669bd2d4a
Merge pull request #794 from Sonicadvance1/fix_log_move_fail
...
Pass log moves on buildbot stage failure
2021-02-23 08:59:21 +02:00
Ryan Houdek
d3e98fedb4
Pulls uname nodename from host system
...
Hardcoding FEXCore as the nodename is an annoyance.
Pull the host's nodename instead.
Fixes #600
2021-02-22 22:39:05 -08:00
Ryan Houdek
8a39f4b25c
Implements BTC, BTR, BTS atomic unit tests
...
This just takes the regular non-atomic unit tests and changes them to
have lock prefixes.
These are all handled as byte sized atomics so there aren't any
alignment problems.
2021-02-22 22:29:31 -08:00
Ryan Houdek
232eeff483
Implements BTC, BTR, BTS atomic variants
...
BTS specifically is being used for threading related tasks. Which could
be why our threading has been unstable.
2021-02-22 22:27:10 -08:00
Ryan Houdek
12ef69650f
Pass log moves on buildbot stage failure
...
Passes the log movement stage to clean up the output.
Really it is only the first unit test stage that is failing, but each
subsequent log run will have failed to move since the previous stage
didn't run.
Makes it easier to scan over a failure as only the first unit test
failure step.
2021-02-22 18:42:28 -08:00
Ryan Houdek
f1349becb4
Disables the unaligned atomic memory op tests
...
These don't run on the armv8.0 runner
2021-02-22 18:33:17 -08:00
Ryan Houdek
f848957e17
Implements unit tests for the new unaligned atomics
...
Tests all the unaligned atomic ops we support now
2021-02-22 18:25:25 -08:00
Ryan Houdek
55fc47e0b7
Passes SIGBUS handler to unaligned atomic op handler for ARMv8.1
...
ARMv8.0 still unsupported here.
This allows us to support unaligned atomic ops for every atomic x86 op
that we support.
2021-02-22 18:22:10 -08:00
Ryan Houdek
4f40166902
Implements unaligned atomic memory ops for ARMv8.1+
...
This takes the existing unaligned CAS handling code and makes it more
robust to handle both true unaligned CAS and unaligned atomic memory
ops.
Specifically it inserts some lambda helpers to calculate the Desired and
Expected values inside of the CAS loops to account for CAS and memory
operation.
2021-02-22 18:17:53 -08:00
Stefanos Kornilios Mitsis Poiitidis
fba626c547
Merge pull request #790 from FEX-Emu/skmp/workaround-exit-group
...
Threading: Workaround exit_group bug
2021-02-22 17:50:59 +02:00
Stefanos Kornilios Mitsis Poiitidis
9b360e66dc
Threading: Workaround exit_group bug
2021-02-22 14:24:42 +02:00
Stefanos Kornilios Mitsis Poiitidis
3a6fd00154
Merge pull request #785 from FEX-Emu/skmp/fix-sar8-16
...
OpDisp: Imm SAR OpSize < 32 needs sign extension
2021-02-18 22:33:46 +02:00
Stefanos Kornilios Mitsis Poiitidis
3bc12de017
OpDisp: Imm SAR OpSize < 32 needs sign extension
2021-02-17 08:17:20 +02:00
Stefanos Kornilios Mitsis Poiitidis
5ae6a64800
Merge pull request #778 from FEX-Emu/skmp/x87-round-trunc
...
x87: Round, Truncate & Precision control
2021-02-17 05:27:56 +02:00
Stefanos Kornilios Mitsis Poiitidis
6615d5972f
x87: Support for FCW, rounding, truncation, precision control
2021-02-17 05:16:53 +02:00
Stefanos Kornilios Mitsis Poiitidis
d075b689a5
Merge pull request #784 from FEX-Emu/skmp/fix-cvt-cvtt
...
OpDisp: Fix cvt/cvtt mapping to dispatcher handlers
2021-02-17 05:14:44 +02:00
Stefanos Kornilios Mitsis Poiitidis
eb60bda1d3
OpDisp: Fix cvt/cvtt mapping to dispatcher handlers
2021-02-16 11:19:28 +02:00
Stefanos Kornilios Mitsis Poiitidis
8c5d88816f
Merge pull request #781 from FEX-Emu/skmp/hackfix-frem
...
X87: FPREM needs to set C2 to 0 to indicate finished iteration
2021-02-16 09:54:37 +02:00
Ryan Houdek
73e9a35e36
Merge pull request #783 from FEX-Emu/skmp/fix-selects
...
Syscalls/x86-32: Fix selects to remarshal the fd_sets
2021-02-15 23:26:33 -08:00
Stefanos Kornilios Mitsis Poiitidis
568fb3643d
Syscalls/x86-32: Fix selects to actually remarshal the sets after running the host syscall
2021-02-16 02:35:38 +02:00
Stefanos Kornilios Mitsis Poiitidis
82367f49e2
X87: FPERM needs to set C2 to 0 to indicate finished iteration
2021-02-16 02:34:52 +02:00
Stefanos Kornilios Mitsis Poiitidis
9f2d49026e
Merge pull request #768 from FEX-Emu/skmp/fix-csgo
...
Fixes for Counter Strike Global Offensive
2021-02-15 03:47:14 +02:00
Stefanos Kornilios Mitsis Poiitidis
bed00d11f1
Tests: Disable pr57275.c because it uses VMOVAPS
2021-02-14 21:36:56 +02:00
Stefanos Kornilios Mitsis Poiitidis
9f0a3dab87
Merge pull request #769 from FEX-Emu/skmp/jit-fallbacks
...
x87: Call C++ handlers instead of forcing interpreter
2021-02-14 21:27:32 +02:00
Stefanos Kornilios Mitsis Poiitidis
07bd2244c7
IR: Remove ShouldInterpret & related logic
2021-02-12 14:21:47 +02:00
Stefanos Kornilios Mitsis Poiitidis
077093eacc
JIT/x64: Use xmm0 as temp instead of xmm11, add xmm11 to RA list
2021-02-12 14:15:22 +02:00
Stefanos Kornilios Mitsis Poiitidis
428111b44b
JIT/arm64: fallback support
2021-02-12 01:23:15 +02:00
Stefanos Kornilios Mitsis Poiitidis
b564927825
IprOps: Add BCDLOAD, BCDSTORE fallbacks
2021-02-11 19:24:57 +02:00
Stefanos Kornilios Mitsis Poiitidis
b0a046dfcb
JIT/IPR: Add fallbacks for almost all x87 ops
2021-02-11 18:52:05 +02:00
Stefanos Kornilios Mitsis Poiitidis
8f5bbde4dd
JIT/x86: Implement VBSL
2021-02-11 18:51:05 +02:00
Stefanos Kornilios Mitsis Poiitidis
7dc0c275cf
Add some basic infrastructure + a few fpr ops
2021-02-11 17:20:29 +02:00
Stefanos Kornilios Mitsis Poiitidis
4e31c1d1ce
RA: Remove erratic assert
2021-02-11 14:31:53 +02:00
Stefanos Kornilios Mitsis Poiitidis
df1c38908c
JIT: Support edge case in CreateElementPair
2021-02-11 14:31:32 +02:00
Stefanos Kornilios Mitsis Poiitidis
fc4b875ff1
Frontend: Only return error if invalid op in first block
2021-02-11 14:31:02 +02:00
Ryan Houdek
40b891ca69
Merge pull request #766 from phire/brk_fix
...
BRK: don't munmap if mmap failed
2021-02-08 21:32:02 -08:00
Scott Mansell
3d6461f0ca
BRK: don't munmap if mmap failed
2021-02-09 17:29:44 +13:00
Ryan Houdek
c11b629189
Merge pull request #764 from phire/fix_cpuinfo
...
Fix the number of physical cpus reported
2021-02-08 16:13:09 -08:00
Scott Mansell
46ec7c406f
emulated-cpuid: add some comments
2021-02-09 13:03:49 +13:00
Ryan Houdek
2a0b9ae681
Merge pull request #763 from phire/bigger_spills
...
arm64: Fix blocks with over 256 spill slots.
2021-02-08 15:57:15 -08:00
Scott Mansell
a28acf5505
Fix the number of physical cpus reported.
2021-02-09 12:52:27 +13:00
Scott Mansell
c4dda4a3f3
arm64: Fix blocks with over 256 spill slots.
...
With multiblock and -n4000, at least one block in geekbench5's camera
benchmark was spilling 324 values. This meant the required stack
adjustment was 5184 bytes, which is larger than can fit into a 12bit
immediate.
Now, it's probally a bug that our RA is spilling that many values, and
we probally should be packing our spill slots closer together, but we
still shouldn't fail to run in this edge case.
So this commit adds a fallback path which uses a temp register to load
the stack adjustment.
2021-02-09 12:08:32 +13:00
Stefanos Kornilios Mitsis Poiitidis
b40131ede3
Merge pull request #759 from phire/fix_mul_op
...
Fix power of 2 OP_MUL optimisation (Fixes Geekbench result upload)
2021-02-07 08:08:37 +02:00
Scott Mansell
5b0ade0bd8
Fix power of 2 OP_MUL optimisation
...
The new left-shift that replaced the multiply takes
it's size from arg[0]. If arg[0] was 64bit and the original
OP_MUL was 32bit, then the original code would truncate
upper bits, and the replacement code wouldn't.
This bug broke the SSL cert checking code in geekbench,
causing it to fail to upload results.
Fixes #647
2021-02-07 18:49:47 +13:00
Stefanos Kornilios Mitsis Poiitidis
fcf1c2572b
Merge pull request #756 from Sonicadvance1/python_check
...
Adds python version check
2021-02-06 14:20:49 +02:00
Ryan Houdek
69891867b1
Adds python version check
2021-02-05 10:36:52 -08:00
Ryan Houdek
abefae2b58
Merge pull request #754 from Sonicadvance1/fix_artifacts_on_failure
...
Always upload artifact results even on failure
2021-02-03 16:08:51 -08:00
