preserve-emulation/ppsspp
1
0 Fork 0
mirror of https://github.com/hrydgard/ppsspp.git synced 2026-10-14 15:00:30 +02:00
Code Issues Packages Projects Releases Wiki Activity
48,374 Commits 97 Branches 131 Tags
ae324777644f2903e8a177455b6db6e0f8213fea
Commit Graph
112 Commits
Author SHA1 Message Date
Henrik RydgårdandClaude Opus 5.5 251d42ab2e SoftGPU: Leave textures with an invalid level address to the generic samplers 
The x86 sampler JIT returned zero for a whole sample when a texture level
pointer was null, and for a mipmapped texture it checked the next level
too, even when it isn't blended in. Tactics Ogre draws its font from a
mipmapped texture whose next level has an invalid address, at level 0,
so all its text was missing on x86-64. The generic samplers read a null
level's texels as zero only where they use that level, and still apply
the texture function. Now samplers for such textures aren't compiled,
and the JIT's null checks are gone. Test: gpu/textures/mipinvalid.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
 2026-10-05 13:05:46 -06:00
Henrik RydgårdandClaude Opus 5.5 347f3668ac SoftGPU: Separate mipmap CLUTs for CLUT8 too 
With the separate-CLUT bit, mip level n puts n above the index bits, wrapped
to the CLUT: CLUT4 adds n * 16 as before, CLUT8 with a 16-bit palette adds
(n & 1) * 256, and a 32-bit palette's 256 entries wrap it back to 0
(gpu/probe exp90). Also in the x86 sampler JIT.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
 2026-10-02 15:09:42 -06:00
Henrik RydgårdandClaude Opus 5.5 15467e24c4 SoftGPU: Sprite UVs and texel addressing like the GE 
Sprites interpolate UVs with the same fixed point planes as triangles, and
texture coordinates become texels at 1/16 texel, truncated toward zero, also
in the sampler JIT. Bit exact on gpu/probe exp57 (through mode and transform
mode sprites, nearest and bilinear).

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
 2026-10-02 15:09:42 -06:00
Henrik RydgårdandClaude Opus 5.5 7d3131abc2 SoftGPU: Bilinear filter horizontally first, truncating, like the GE 
The GE lerps the top and bottom texel pairs, truncates each to 8 bits, then lerps
those vertically. Bit exact on gpu/probe exp52 (also in ppsspp-re tools/geprobe).

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
 2026-10-02 15:09:42 -06:00
Herman Semenoff 400d136f51 GPU: modernize use C++17 constexpr for precalculate compilation 2026-05-19 21:57:45 +03:00
Henrik Rydgård d5bfc1b5eb Finish up this cleanup (move _M_SSE compat define out of Common.h) 2024-12-19 16:23:20 +01:00
Henrik Rydgård e01ca5b057 Logging API change (refactor) (#19324) 
* Rename LogType to Log

* Explicitly use the Log:: enum when logging. Allows for autocomplete when editing.

* Mac/ARM64 buildfix

* Do the same with the hle result log macros

* Rename the log names to mixed case while at it.

* iOS buildfix

* Qt buildfix attempt, ARM32 buildfix
 2024-07-14 14:42:59 +02:00
Unknown W. Brackets b2798c7ada jit: Add more reasonable estimates for RX protect. 2022-11-20 10:55:35 -08:00
Unknown W. Brackets 79b1d1d35f softgpu: Better approximate slope mip level mode (#16276) 
* samplerjit: Remove unused x/y parameters.

Still need to tune the accuracy of filtering, but those were not the
right way.

* softgpu: Better approximate slope mip level mode.

This isn't exactly right, but it's closer.

* softgpu: Calculate auto from largest difference.

Direction shouldn't matter.
 2022-10-23 10:15:43 +02:00
Unknown W. Brackets 167213c746 softgpu: Cache texture bufws at 16 bit. 
Reducing the size of state a bit.
 2022-09-12 21:57:00 -07:00
Unknown W. Brackets 90e009edb9 softgpu: Clamp/wrap textures at 512 pixels. 
A texture larger than 512 is "valid", but simply wraps/clamps at 512.
Importantly, the texture coords are still calculated at the specified
size, which can be up to 32768.
 2022-09-10 20:23:09 -07:00
Unknown W. Brackets df1a91ee25 samplerjit: Correct nearest negative texture clamp. 
Was not clamping to zero when negative.
 2022-02-20 10:25:00 -08:00
Unknown W. Brackets a88c9a0680 softgpu: Remove incorrect offsetting for X/Y. 2022-02-20 09:13:20 -08:00
Unknown W. Brackets ad18833a4f samplerjit: Fix non-SSE4 bugs in jit. 2022-02-15 20:13:38 -08:00
Unknown W. Brackets 99d7703d33 samplerjit: Precalculate DXT1/3/5 offsets. 
This improves WALL-E by 8% overall.
 2022-02-05 13:04:17 -08:00
Unknown W. Brackets c91b51c8e1 samplerjit: Reduce DXT5 decode code size a bit. 2022-02-03 20:42:34 -08:00
Unknown W. Brackets c2dd59084d samplerjit: Optimize DXT calc using BMI2. 2022-02-01 00:18:56 -08:00
Unknown W. Brackets 3e4afe2a0c samplerjit: Avoid RCX gymanstics with BMI2. 2022-01-31 22:33:09 -08:00
Unknown W. Brackets 4cadcea6da samplerjit: Decode colors with BMI2. 
This only happens with nearest, though, so very small benefit.
 2022-01-31 22:05:34 -08:00
Unknown W. Brackets 1b2cf52bfe samplerjit: Fix non-shared CLUT on Linux. 
Oops, good that CI will catch this now - I've broken this more than once.
 2022-01-29 22:20:46 -08:00
Unknown W. Brackets 26a8d498d7 samplerjit: Correct level lookup in nearest. 2022-01-29 20:29:43 -08:00
Unknown W. Brackets 3387ab1711 samplerjit: Fix reg corruption in DXT funcs. 
We'd cache something in a reg, but it'd no longer be there.
 2022-01-29 20:29:08 -08:00
Unknown W. Brackets 5976cad797 samplerjit: Reduce register waste. 
A few registers were allocated longer than needed, which made requiring
stack more likely.
 2022-01-29 09:47:06 -08:00
Unknown W. Brackets eb70a90347 samplerjit: Avoid frac uv transfer to gen regs. 
It should just stay in vec, this is more convenient anyway.
 2022-01-28 23:50:54 -08:00
Unknown W. Brackets 99d6d569f0 samplerjit: Reduce transfers in nearest texel calc. 
This benefits a few games, mostly where there's lots of UI or similar.
 2022-01-24 21:28:04 -08:00
Unknown W. Brackets c1e657ed47 samplerjit: Better vectorize UV linear calc. 
Gives about 1-2% when mips are used.
 2022-01-24 20:42:07 -08:00
Unknown W. Brackets 733046962f samplerjit: Reuse XMM reg for sizes. 
Gives just under 1% overall improvement in games using mips.
 2022-01-24 19:01:23 -08:00
Unknown W. Brackets d8c5c35b1a samplerjit: Optimize texenv blending a bit. 
This reduces to a single multiply, which is much faster.
 2022-01-23 11:43:34 -08:00
Unknown W. Brackets 4262e657b4 samplerjit: Oops, forgot about 64 unpack. 
Just a minor codegen tweak.  Always forget there are more of these than
pack instructions.
 2022-01-22 10:49:36 -08:00
Unknown W. Brackets 0425b8d630 samplerjit: Fix Linux stack corruption. 
Oops, nearest was not using the red zone correctly.
 2022-01-22 10:47:32 -08:00
Unknown W. Brackets 212e730e98 samplerjit: Fix some Linux register issues. 2022-01-22 00:14:15 -08:00
Unknown W. Brackets 6ec819878a samplerjit: Reduce prolog/epilog spill. 
Track reg usage so we only push/pop what we need.
 2022-01-19 00:03:59 -08:00
Unknown W. Brackets 357e2e9d68 softjit: Simplify constant writes. 2022-01-19 00:03:59 -08:00
Unknown W. Brackets c2985bca31 softjit: Centralize some common funcs from sampler. 
No need to duplicate this code.
 2022-01-19 00:03:59 -08:00
Unknown W. Brackets 0ba2d05da5 samplerjit: Simplify AVX shift-copies. 
These have been the most common and the fallback is safe.  Let's just add
a helper.
 2022-01-17 15:15:36 -08:00
Unknown W. Brackets d6fa301ab1 softgpu: Track CLUTs as states for binning. 
This way we can have multiple CLUTs in process at once, which helps.
 2022-01-16 08:14:09 -08:00
Unknown W. Brackets edb79d968f softgpu: Cache CLUT params in sampler state. 
And now there's no more gstate for pixel drawing or sampling.  Just a
little left in rasterization.
 2022-01-15 18:09:09 -08:00
Unknown W. Brackets c0e85e6170 softgpu: Move texenv color into sampler state. 2022-01-15 17:52:40 -08:00
Unknown W. Brackets ad3635c82a softgpu: Move tex size to cached state. 2022-01-15 17:22:43 -08:00
Unknown W. Brackets b915a82c41 softgpu: Correct decal doubling without alpha. 2022-01-09 12:23:55 -08:00
Unknown W. Brackets 72aa4be879 samplerjit: Skip processing alpha if unused. 2022-01-09 12:23:55 -08:00
Unknown W. Brackets fe0b3dbd01 samplerjit: Fix alpha for 565 in linear lookup. 2022-01-09 11:08:46 -08:00
Henrik Rydgård f82f24a9bb Merge pull request #15280 from unknownbrackets/samplerjit-dxt 
Correct some recent regressions in samplerjit
 2022-01-05 09:42:30 +01:00
Unknown W. Brackets 741a9b0a4d samplerjit: Fix DXT compilation. 2022-01-05 00:00:03 -08:00
Unknown W. Brackets 19998976c7 samplerjit: Correct linear compile failure. 
It was resetting to nullptr, because `nearest` was nullptr.
 2022-01-04 23:58:07 -08:00
Unknown W. Brackets 2aa57679fa softjit: Keep mip S/T calc in SIMD. 
This is only a tiny bit faster, though.
 2022-01-03 06:45:10 -08:00
Unknown W. Brackets 26e7768a67 samplerjit: Remove old linear nearest paths. 
We only use it for DXT now, so let's not keep the dead code around.
 2022-01-02 17:28:52 -08:00
Unknown W. Brackets 5e3bef7e14 samplerjit: Avoid gather if overread could crash. 
This should be rare, but a game could easily shove a CLUT4 texture at the
end of VRAM, and then accessing the last index would segfault.
 2022-01-02 17:28:52 -08:00
Unknown W. Brackets 7806dfddea samplerjit: Use VPGATHERDD for all types. 2022-01-02 17:19:18 -08:00
Unknown W. Brackets ce6ea8da11 samplerjit: Apply gather lookup to all CLUT4. 2022-01-02 17:19:18 -08:00