e6185ba6bb
Clean out NaNs and INFs from vertex coordinates
...
This enables a behavior seen on the real PSP where 0 * NaN == 0 in the
GPU (NOTE: This means the vertex transform pipeline).
However, this
doesn't touch INFs unfortunately, and we need to modify those too...
This, together with #21715 , finally fixes #20204 .
2026-06-02 10:52:39 +02:00
5e16bf907b
VertexDecoder: Improve logging for missing formats. Add missing convert function.
...
The missing function is mainly used in D3D11, which can be used on
Windows for ARM64. It's not necesssary for the other backends, which is
why it used to be missing in the ARM64 vertex decoder.
Also fix a minor memory leak in AtracCtx2.
2025-09-24 10:52:09 -06:00
d5bfc1b5eb
Finish up this cleanup (move _M_SSE compat define out of Common.h)
2024-12-19 16:23:20 +01:00
81f1b3fd95
Make handwritten vertex decoders work with non-compiled vertex decoding
2024-05-11 10:00:35 +02:00
afca3717ba
Specialized vertex decoder for GTA
2024-05-11 10:00:35 +02:00
64a810f9bf
VertexDecoder: Minor optimization for CPUs not supporting SSE4.
2023-12-05 01:23:09 +01:00
c4e44d66b0
x86/x64: Nop-align the main loop of vertex decoder loops
2023-06-12 20:39:39 +02:00
01cea7f088
Pass uvScale in as an argument to the vertex decoder
...
Cleaner than overwriting/restoring gstate_c.uvScale in the decoder
loop. A small cleanup I've been wanting to do for ages.
Expecting a negligble perf boost if any.
2023-06-12 20:25:18 +02:00
d957f6b0be
Of course got the check backwards
2023-06-12 19:45:34 +02:00
1a1462ecb0
x86 buildfix, warning fix
2023-06-12 17:46:57 +02:00
c9aa3479a4
Make vertexFullAlpha-in-register work the same as on ARM.
2023-06-12 16:08:14 +02:00
a164f77f47
VertexDecoderX86 (64-bit only): Avoid a memory access per loop iteration for alpha
2023-06-12 15:58:55 +02:00
49f6c461ad
Reporting: Fix some header includes.
...
Particularly in Common, avoid including Core/Reporting.h.
2022-12-27 14:58:20 -08:00
b2798c7ada
jit: Add more reasonable estimates for RX protect.
2022-11-20 10:55:35 -08:00
1469a32a9d
Vertex decoder: Add fallback for non-SSE4.1
...
See #16157
2022-10-03 19:06:17 +02:00
ca248e1201
softgpu: Fix s8 primitives in throughmode.
...
Also always cull no-position verts, hardware too. Matches tests.
2022-09-18 07:46:18 -07:00
4889d5285d
vertexjit: Use consistent skinInDecode.
...
Just a little cleanup.
2022-09-10 21:54:04 -07:00
9f48ddb18c
vertexjit: Clamp through float pos during decode.
...
Rather than decoding it on read, better to decode it right away.
2022-09-01 23:33:52 -07:00
367cca96b4
vertexjit: Optimize 5551 read a bit on x86/x64.
2021-11-25 19:23:59 -08:00
f32f89dd90
Global: Remove some unused variables.
2021-02-15 11:59:45 -08:00
c1fa4958d9
vertexjit: Only save extra regs on x64.
2021-02-01 07:06:18 -08:00
ddff063097
vertexjit: Correct saved registers on x64.
2021-01-31 15:05:15 -08:00
7f27f569d7
x64: Fix stack alignment in vertex decoder, just because.
2020-09-12 16:10:17 +02:00
c5e0b799d9
Remove category from _assert_msg_ functions. We don't filter these by category anyway.
...
Fixes the inconsistency where we _assert_ didn't take a category but
_assert_msg_ did.
2020-07-19 20:33:25 +02:00
7910b4029a
arm64jit: Track writable and non-writable pointers.
...
Switch uses different memory regions. We can handle this, might as well
cleanup some const abuse.
2020-05-17 00:15:12 -07:00
dbe52c8135
Remove some obsolete checks - software skinning is now supported even if morphcount isn't 1.
2018-04-10 13:09:01 +02:00
ccd594dae7
Revert "VertexLoader: Remove now-unused weights translation code"
...
This reverts commit 44100c6c1d
2018-04-10 12:30:49 +02:00
36fd2711d6
Revert "Remove further remains of hardware skinning."
...
This reverts commit 40db61a680
2018-04-10 12:22:41 +02:00
b6cbbceac9
MOVUPS to be safe.
2018-03-05 16:23:07 +01:00
91b565ae20
Minor simplification in X86/X64 vertex decoder jit
2018-03-05 13:19:36 +01:00
40db61a680
Remove further remains of hardware skinning. Fixes #10661
2018-03-05 00:03:47 +01:00
44100c6c1d
VertexLoader: Remove now-unused weights translation code
2018-03-05 00:03:47 +01:00
22e65ba80d
Get rid of ugly alignment macros and some other cruft, we now have alignas(16) from C++11
2017-08-31 01:14:51 +02:00
8d0498303a
Fix a PIC compliance bug in the VFPU. Comment other cases properly (for easy searching).
2017-08-29 11:45:12 +02:00
f7fea57bb6
More RIP addressing in software skinning..
2017-07-07 15:12:57 +02:00
837118d230
More RIP elimination
2017-07-07 15:07:56 +02:00
86396ba39b
Turning off the "close memory finder" lets us find more RIP addressing...
2017-07-07 15:07:53 +02:00
f5e110018b
Fix software skinning
2017-07-07 12:57:39 +02:00
270001e651
vminmax_sreg only needs a single element
2017-07-07 11:33:07 +02:00
cf7939d3e4
Get rid of remaining RIP addressing in vertex decoder
2017-07-07 11:33:06 +02:00
442a7012a6
Remove some rip addressing in vertex decoder
2017-07-07 11:33:06 +02:00
99d23fb021
X64/X86: Even more use of the context register
2017-07-05 13:21:35 +02:00
385204bec6
X64 vertex decoder: Use relative memory accesses when loading matrices.
2017-07-05 12:51:41 +02:00
744736b926
SoftGPU: Jit 16 bit texel decode.
2017-05-30 22:57:42 -07:00
5637ae1190
Fix typo in VertexDecoder
2017-05-13 16:27:26 +02:00
dacb776300
x86: Minor optimization for s16 through positions.
...
This takes it from 150% to 390% non-jit for pos only verts.
2017-05-07 12:20:56 -07:00
0fe927aaa7
x86: Minor memory copy perf improvement.
...
This was actually 270% -> 340% non-jit for pos-only verts.
2017-05-07 12:18:47 -07:00
b06e271ef3
x86: More optimal 4444 in vertexjit.
...
This was commented out, but works fine and goes from 320% -> 450% the speed
of non-jit for simple pos/col verts.
2017-05-07 12:14:06 -07:00
257f8dbbc6
GPU: Remove now-unused vertex decoder funcs.
...
We always convert to float now, so these functions are no longer used.
2017-05-06 18:55:16 -07:00
e9625707fa
x86: Implement Jit_TcU16Through.
...
This just ports the updateSide logic from other types.
2017-03-18 13:44:47 -07:00
Henrik Rydgård