Unknown W. Brackets
2ba35c6391
interp: Use an integer multiply for dots.
2019-08-04 21:17:09 -07:00
Unknown W. Brackets
08a894abde
interp: Handle NaN better in vfpu_dot.
2019-08-04 21:17:00 -07:00
Unknown W. Brackets
4bff980d34
interp: Use software dot for better accuracy.
...
This gets some vdot/similar tests to give more accurate results compared
to hardware. Also added flushing of zero and NaNs.
Currently disabled, only enabled with a compile-time flag.
2019-08-04 21:14:23 -07:00
Unknown W. Brackets
c7e83cd4fa
interp: Correct vfim for -inf and similar.
...
Was dropping the sign bit before for inf and nan.
2019-03-31 13:41:48 -07:00
Unknown W. Brackets
af3ed69144
interp: Mask moves to vfpu ctrl.
...
These bits of the registers can't be written.
2019-03-31 10:37:07 -07:00
Unknown W. Brackets
b86a6af364
interp: Properly apply mask on single lane ops.
...
When using something like vadd.s, we should still be applying the mask.
Mainly should only matter if masks are set in a conditional, or if games
nop out instructions.
2019-03-31 10:13:28 -07:00
Unknown W. Brackets
0be3213151
interp: Correct vscl prefix handling.
2019-03-31 10:13:28 -07:00
Unknown W. Brackets
fef8e71b8f
interp: Correct behavior for matrixes with size=1.
...
Games don't (shouldn't?) actually use 1x1 matrices, but they seem to have
predictable behavior in matrix ops.
2019-03-31 08:21:29 -07:00
Unknown W. Brackets
22940f0393
Debugger: Avoid asserts in disassembly.
2018-06-08 06:59:18 -07:00
Henrik Rydgard
e1e335885c
A better alternative to #8564
2016-12-01 18:23:58 +01:00
Henrik Rydgard
3cae60b320
Revert potentially dangerous optimization that may have caused #8754 , though not sure how.
2016-05-18 21:22:08 +02:00
Henrik Rydgard
2cbfb192c4
IR: Lots more VFPU support, some with SIMD
2016-05-12 12:17:25 +02:00
Bovine
2ed4c1bd6b
Fix matrix disassembly notation
...
Changed disassembler output to match gas input for operations on
transposed matrices where row != col.
2015-01-03 02:04:08 -07:00
Henrik Rydgard
29dcc0a303
Minor cleanups, warning fixes
2014-12-06 12:25:28 +01:00
Henrik Rydgard
8f016d3e48
Merge some matrix utils and stuff from the NEON branch
2014-11-29 11:37:45 +01:00
Aapo Rantalainen
ed6b8e34ab
Fix missing includes (stdio.h)
2014-01-01 22:31:03 +02:00
Henrik Rydgård
ce378b231f
Delete CPU.cpp/h , cleanup
2013-12-30 00:11:29 +01:00
Henrik Rydgard
2d8429ac48
Assorted cleanup in the MIPS emulation
2013-12-10 13:15:16 +01:00
Unknown W. Brackets
ec05146ffd
Improve vfpu disasm for a few instructions.
2013-11-29 10:07:15 -08:00
Henrik Rydgard
55500d4bb6
Reorder VFPU registers in memory so that we can flush and reload them in bulk more often.
...
Doesn't actually do that yet, that's for the NEON branch.
2013-11-28 13:27:51 +01:00
Unknown W. Brackets
e3834d5833
Avoid some unlikely uninitialized values.
2013-10-26 18:31:14 -07:00
Unknown W. Brackets
109ad17ac6
Use a typesafe struct for opcodes.
...
Also, correctly read delayslots using Read_Instruction on ARM.
2013-08-24 15:36:24 -07:00
Unknown W. Brackets
46c5a48a86
Clean up some minor aliasing.
2013-08-22 23:31:49 -07:00
Unknown W. Brackets
b733bc1a2a
Clean up some indents, add some reporting.
2013-05-18 02:09:32 -07:00
Unknown W. Brackets
e16514b50d
Avoid float copy in vfpu to workaround 32-bit bug.
...
This appears to be an MSVC bug (or just precision issue) when using
fast math, only on 32-bit builds. Anyway, memcpy or u32 * fixes it.
Unfortunately, matrix ops have similar issues and memcpy doesn't seem
to help there.
2013-05-18 02:09:32 -07:00
Henrik Rydgard
516ca8a0c4
Merge branch 'master' into armjit-fpu
...
Conflicts:
Core/MIPS/ARM/ArmJit.h
Core/MIPS/x86/CompVFPU.cpp
GPU/GLES/Framebuffer.cpp
2013-02-28 23:56:28 +01:00
Unknown W. Brackets
01f3c4ecde
Log an error if we hit a 1x1 matrix.
2013-02-19 07:46:29 -08:00
Henrik Rydgard
e32721c72a
Merge branch 'master' into armjit-fpu
...
Conflicts:
Core/MIPS/MIPSVFPUUtils.cpp
Core/MIPS/x86/CompVFPU.cpp
GPU/GLES/VertexDecoder.cpp
2013-02-19 00:50:33 +01:00
Unknown W. Brackets
179fccaff7
Tests say matrices apply mask to last col (kinda.)
...
It seems inconsistent but probably better than before. Also add an error.
2013-02-18 13:19:16 -08:00
Unknown W. Brackets
be631dea64
Disasm transposed vectors properly.
2013-02-17 17:52:58 -08:00
Henrik Rydgard
159f423135
VFPUutil style & simplification
2013-02-16 09:28:56 +01:00
Henrik Rydgard
909b768f47
Don't need separate variables for writemask. Some optimizations.
2013-02-16 09:28:55 +01:00
Henrik Rydgard
526335cacf
VFPUutil style & simplification
2013-02-15 23:09:02 +01:00
Henrik Rydgard
d22e258943
Don't need separate variables for writemask. Some optimizations.
2013-02-15 22:56:38 +01:00
Henrik Rydgard
aabc0aa9ef
Quick implementation of LV.Q and SV.Q in x86/x64 JIT
2013-01-25 19:50:30 +01:00
Henrik Rydgard
64cc573703
Switch to "GPL 2.0 or later" for various reasons. I wrote most of the code I imported from Dolphin (which is GPL2-but-not-later), so it should be OK.
2012-11-04 23:24:00 +01:00
Henrik Rydgard
4f7ad15758
Add snapshot of the whole source code.
2012-11-01 16:19:01 +01:00
Unknown W. Brackets