preserve-emulation/ppsspp
1
0 Fork 0
mirror of https://github.com/hrydgard/ppsspp.git synced 2026-10-15 07:20:31 +02:00
Code Issues Packages Projects Releases Wiki Activity
48,818 Commits 99 Branches 131 Tags
b71fbaeb5cd40cd403bc92b5ca0b444f2dfa41bf
Commit Graph
92 Commits
Author SHA1 Message Date
Henrik RydgårdandClaude Opus 5.5 6096ea682c SoftGPU: Splat Vec4s in registers, and three more stalls 
Vec4<int> and Vec4<float>'s AssignToAll went through the four value
constructor, which stores the values one by one, so whatever loaded the
vector next stalled: the edge functions' setup per triangle and tile,
among others. They now splat in a register on SSE and NEON.
ReadSpanDepth loads its four depths as one vector where they're stored
together, rather than through an array, and ComputePlane is inline, as
returning its plane through memory stalled the UV planes' setup the same
way.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
 2026-10-10 15:01:49 +02:00
Henrik RydgårdandClaude Opus 5.5 c8819d5087 SoftGPU: Make arm64 and x86 render the GE probes identically 
Found by the probe regression run (ppsspp-re geprobe/regress.py), which
gave different counts on the two:

- Vec4<int> >> was a logical shift on SSE, arithmetic elsewhere. A negative
  spot factor lit pixels white on x86 (gpu/probe exp100).
- Vec4<float>::ToRGBA rounded on SSE and truncated on NEON and the generic
  path. Spline colors were one lower on ARM (exp48-51, the PSP rounds).
- Texel coordinates past the int range: the C++ sampler now gives INT_MIN
  like the x86 JIT's conversion, as the PSP does with a repeating texture (exp5).
- The viewer direction is normalized with the GE's normalization instead of
  the SIMD float one (exp2, exp3).

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
 2026-10-02 15:09:43 -06:00
Henrik Rydgård 1297958f9d Optimize ConvertMatrix4x3To3x4Transposed (by Gemini) 2026-07-13 11:25:02 +02:00
Henrik Rydgård db712016bb Comment improvements, rename a dirty-flag 2026-07-13 11:10:31 +02:00
Henrik Rydgård 0e8ba0d6fe Bugfixes, cleanup 2026-05-30 19:07:59 +02:00
Henrik Rydgård 3378faca5e Optimize morph code (non-JIT vertex decoders) 2026-01-20 13:51:24 +01:00
Henrik Rydgård e93c80db4e Cleaning up our SIMD header includes, using the new header 2024-12-19 16:08:48 +01:00
Henrik Rydgård 45cf6d423e Buildfixes 2024-12-18 13:57:29 +01:00
Henrik Rydgård c794f4bd41 Add an unrelated comment and some casts 2024-06-05 08:35:09 +02:00
Henrik Rydgård 737ec3e90b NEON buildfix 2023-11-28 18:40:10 +01:00
Henrik Rydgård 4ec2d76bc9 NEON-optimize matrix tranposes 2023-11-27 23:57:26 +01:00
Henrik Rydgård 45aae7b9da ARM32: Backport a lot of previously 64-bit-only NEON optimizations to ARM32. 2023-11-27 23:51:10 +01:00
Unknown W. Brackets d5b4c98f96 softgpu: Reduce some non-SIMD lighting math. 
Small perf improvement for vertex/lighting heavy (i.e. 3D) scenes.
 2023-07-16 10:31:44 -07:00
fp64 cd9f01c4df Remove SSE4 path from Vec4<int>::operator* 2023-06-30 22:07:26 -04:00
fp64 f133739cd0 Replace some signed divison in SoftGPU 
This also adds a few bitwise operations to Vec4<int> and further
SIMDifies it.
Also, fixes unrelated warning.
 2023-06-29 16:43:21 -04:00
Unknown W. Brackets dfe113e846 Merge pull request #17634 from fp64/macro-x86-loadu 
Streamline x86 SSE workaround
 2023-06-27 23:01:41 -07:00
M4xw 99ce3125df [Softgpu] Fix AArch64 oversight 2023-06-27 17:20:11 +02:00
fp64 436b49c4f2 Streamline x86 SSE workaround 
Seems clearer than using #ifdef's at each site. Also rationale
is clearly spelled out, one 'Go to definition' away from any instance.
 2023-06-27 00:30:01 -04:00
Henrik Rydgård 08d578dce9 Merge pull request #17618 from unknownbrackets/softgpu-opt-cast 
Optimize casts in softgpu
 2023-06-25 07:55:30 +02:00
Unknown W. Brackets 15b66ba6c0 softgpu: Make SIMD on x86_32 a bit safer. 2023-06-24 14:49:23 -07:00
Unknown W. Brackets 795de9b164 softgpu: Use SIMD for more Vec4 casts. 
A number of these were falling back to some pretty terrible code.
Thanks to fp64 for noticing.
 2023-06-24 12:36:44 -07:00
Unknown W. Brackets a7fa37d114 softgpu: Use SIMD more for dot products. 2023-06-12 19:54:32 -07:00
Unknown W. Brackets b55dbdab7f softgpu: Use NEON for some color conv. 2023-01-07 19:06:34 -08:00
Unknown W. Brackets a7b7bf7826 Global: Set many read-only params as const. 
This makes what they do and which args to use clearer, if nothing else.
 2022-12-10 21:13:36 -08:00
Henrik Rydgård 37b0c90a2d Silence address-sanitizer warnings in Math3D.h on ARM64 (not very serious but good to fix) 2022-12-09 23:47:42 +01:00
Unknown W. Brackets b2e6a086dc softgpu: Reduce size of VertexData texture coords. 
There's no real benefit to this with only two values.
Not much of a gain perf wise, but still good to transfer less data.
 2022-09-12 21:10:46 -07:00
Unknown W. Brackets b90fc7137f softgpu: Correct accuracy of fog calculation. 
This matches values from a PSP exactly, with the help of immediate mode
vertex values (since this directly allows specifying the fog factor
without any floating point math.)
 2022-09-11 08:24:40 -07:00
Henrik Rydgård cd92151de7 Add ARM64_NEON compile arch flag 
This allows doing ARM64 builds without NEON support, and allows simplifying some checks.
 2022-06-25 07:29:20 +02:00
Unknown W. Brackets 163fa352e8 softgpu: Avoid some unaligned access on x86_32. 2022-03-13 12:44:58 -07:00
Unknown W. Brackets 8a00c2d233 GPU: Allow gcc/clang/icc runtime SSE4 usage. 
All our builds before were only using SSE4 in jit...
 2022-01-08 17:09:09 -08:00
Unknown W. Brackets 43f71884ee softgpu: Clarify internal matrix multiply usage. 2022-01-07 17:53:24 -08:00
Unknown W. Brackets e7d66f2029 softgpu: Reuse SSE/NEON matrix code. 2022-01-06 21:19:47 -08:00
Unknown W. Brackets 079b67e7ed softgpu: Use common SIMD matrix multiplies. 2022-01-06 21:19:47 -08:00
Unknown W. Brackets 2d8fdd8cf4 Math3D: Allow construction from NEON vectors. 
This makes it match SSE and easier to keep things generic.  Will impact
alignment of non-packed Vec2/Vec3.
 2021-11-28 08:24:53 -08:00
Unknown W. Brackets fb6fadbbb7 softgpu: Fast path rectangles as fans. 
Some games, such as Legend of Heroes III, use fans instead of strips.
 2021-11-14 18:31:45 -08:00
Henrik Rydgård a498f164ee vmulq_laneq_f32 not supported on ARM32 2021-10-31 16:32:45 +01:00
Henrik Rydgård fdacf751ce NEON/SSE-optimize some matrix multiplications used by software transform 
Will hopefully reclaim any potential speed loss from the recent
refactor.
 2021-10-31 13:36:34 +01:00
Unknown W. Brackets 2f63f9999d GPU: Normalize 0 to 1 always in software lighting. 
See #14167.  This seems to be consistent.
 2021-02-27 23:51:45 -08:00
Henrik Rydgård 9e41fafd0d Move math and some file and data conversion files out from native to Common. 
Buildfixing

Move some file util files

Buildfix

Move KeyMap.cpp/h to Core where they belong better.

libretro buildfix attempt

Move ini_file

More buildfixes
 2020-10-04 09:12:46 +02:00
Henrik Rydgård 510229b68b SoftGPU: Detect through-mode rectangles from triangle strips 2019-10-27 20:54:36 +01:00
xebra 62aaf6336a Math3D: Something wrong with hand simd optimization in vec2<float>, so it causes very slow down. 
However, compiler optimization is faster enough, so removed it.
 2018-10-07 23:54:17 +09:00
xebra d0682d7829 [spline/bezier]Move SIMD optimization of vector operations to Math3D.h. 
Needs rebuild to avoid a dialog confirmation on Visual Studio.
 2018-10-07 23:53:43 +09:00
xebra 62ad5fe546 Fix namespace Vec2f. 2018-10-07 23:53:41 +09:00
Henrik Rydgård 45cfda4aa0 Small refactoring in VertexDecoderCommon 2018-03-05 00:03:47 +01:00
Henrik Rydgård 7bb427e6f1 Buildfix 2017-08-31 17:24:34 +02:00
Henrik Rydgård 6a1fa728d8 Remove Globals.h 2017-08-31 17:15:22 +02:00
Henrik Rydgård 91783a3281 SIMD-optimize some data conv routines used in uniform updates. 2017-08-20 11:43:35 +02:00
Unknown W. Brackets 4fb7e43af8 SoftGPU: Grab 4 S/T coords in non-through too. 2017-04-23 11:11:16 -07:00
Unknown W. Brackets 3142462ac6 SoftGPU: Rasterize triangles in chunks of 4 pixels. 
Not very optimal yet.
 2017-04-23 10:37:11 -07:00
Unknown W. Brackets 5ee062c681 Try to optimize bezier color sampling. 2015-04-18 12:47:21 -07:00