105 Commits

Author SHA1 Message Date
sum2012 5954db188e Remove 2 cases for fast path for soft gpu by Deekseek
Fix #21363
https://chat.deepseek.com/share/980xe4c2ue35te8cy2
Performance impact is minimal because sprites that require depth/fog are relatively rare, and the fallback still uses the per-pixel rectangle loop (not the full triangle rasterizer)
2026-05-07 00:26:06 +08:00
Henrik Rydgård 3e62157487 Fix some assorted static analysis warnings 2025-08-25 10:14:15 +02:00
Henrik Rydgård 2e0ca023ae Build warning fixes 2025-04-17 11:05:52 +02:00
Henrik Rydgård 5326d87f9c Rename CrossSIMD to SIMDHeader, but also keep CrossSIMD.h (will have a future use) 2024-12-19 15:15:43 +01:00
Henrik Rydgård 3e198c53b2 More include cleanup 2024-12-18 13:57:26 +01:00
Henrik Rydgård 3af961f3ba Revert DrawPixel changes 2023-04-02 16:41:29 +02:00
Герман Семенов 122b63b9a8 GPU: using if constexpr C++17 optimization 2023-04-02 16:36:37 +02:00
Unknown W. Brackets a88b8a14f6 softgpu: Fix over-optimization of alpha test.
When alpha blend is off, was previously skipping the alpha test if only it
was enabled.  See #17213.
2023-03-31 23:53:37 -07:00
Unknown W. Brackets 2c5b0999e8 softgpu: Make debug-only optim more consistent.
Of course it doesn't matter when optimizations are enabled in any compiler
that can build PPSSPP...
2023-03-31 23:52:23 -07:00
Henrik Rydgård ee3618290b Typo fix in NEON code.
Fixes #16772
2023-01-10 12:32:33 +01:00
Unknown W. Brackets 62fe03dcb4 softgpu: Use NEON for texture blending. 2023-01-07 19:06:35 -08:00
Unknown W. Brackets 587a322207 softgpu: Use NEON SIMD for alpha blending, etc. 2023-01-07 19:06:34 -08:00
Unknown W. Brackets 49f6c461ad Reporting: Fix some header includes.
Particularly in Common, avoid including Core/Reporting.h.
2022-12-27 14:58:20 -08:00
Unknown W. Brackets 77c98441f4 softgpu: Expand fast path to all fb formats.
This should speed up 2D games not using 5551, and make it mroe likely any
regression is caught for Darkstalkers.
2022-12-06 18:43:21 -08:00
Henrik Rydgård 49f8f96a98 Couple more microoptimizations 2022-12-07 02:08:48 +01:00
Henrik Rydgård 50fdf67fd0 Software rect raster: Speed up memory tracking in debug mode (ztag unused, remove allocations) 2022-12-07 01:51:20 +01:00
Henrik Rydgård b51422c18b Ordering fix, should fix elimination of the DarkStalkers software blit 2022-12-07 01:50:48 +01:00
Henrik Rydgård 70b6733d0a Minor alpha bit optimization in Darkstalkers rectangle code 2022-12-07 01:50:29 +01:00
Unknown W. Brackets 51fa931e46 softgpu: Allow no alpha blend in 5551 fast path. 2022-12-06 06:37:51 -08:00
Unknown W. Brackets 60c129c52f softgpu: Use no-modulate fast path for REPLACE. 2022-12-06 06:31:30 -08:00
Unknown W. Brackets c291bd78ef softgpu: Use a template for 5551 fast path.
Just refactoring here, should be the same.
2022-12-06 06:29:19 -08:00
Unknown W. Brackets 264a81913d softgpu: Avoid checking UVs for sprites w/o tex. 2022-12-06 06:11:38 -08:00
Unknown W. Brackets 5d6b213fa7 softgpu: Avoid tri combine to rect if clipping. 2022-12-05 23:24:26 -08:00
Unknown W. Brackets d9501cb0ad softgpu: Fix w culling of triangle rects. 2022-12-05 23:09:55 -08:00
Unknown W. Brackets 4d06400548 softgpu: Fix compile hazard while running.
This prevents any clearing of cache while other threads may be using
previously cached funcs, and avoids wx exclusive hazards.
2022-11-20 12:04:02 -08:00
Unknown W. Brackets ce51942508 softgpu: Correct WX-exclusive platform hazards.
Should mainly affect BSD at this point.
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 1931fa2f5f softgpu: Make triangle fan rect detection generic. 2022-10-16 16:01:09 -07:00
Unknown W. Brackets 6282f8b05f softgpu: Expand texture coords to include q.
We'll need this to correctly project.
2022-09-26 17:13:14 -07:00
Unknown W. Brackets 8376176b2f softgpu: Split clippos out of rasterization vert.
We don't use it, except w, at all in rasterization, so no need to keep it
in the bin queue.
2022-09-26 16:50:40 -07:00
Unknown W. Brackets c47d7eab38 softgpu: Simply 5551 blending fast path.
Since it only supports multiply and add, let's just stick with that.
2022-09-24 18:55:45 -07:00
Unknown W. Brackets 1eeb4f0bcf softpu: Refactor out 5551 fast path checks.
They were duplicated, and better to organize them according to state.
2022-09-24 18:55:45 -07:00
Unknown W. Brackets f30b1d048d softgpu: Avoid fast path in another wrong case.
Seen in Kurohyo.  Missed that the alpha blend check essentially means only
standard blending can work.
2022-09-24 17:53:09 -07:00
Unknown W. Brackets cefef3b4f9 softgpu: Narrow blend check for fast path further.
See #15756, frame was largely black because of a full screen blend
rectangle intended to brighten the screen slighty (I assume.)
2022-09-20 14:29:33 -07:00
Unknown W. Brackets de080e2594 softgpu: Simplify vertex range culling.
The previous logic was harder to understand and easier to get wrong.
Just drop them when clipping the primitive.
2022-09-18 06:16:25 -07:00
Unknown W. Brackets 9b01fce5b5 softgpu: Run early Z tests in fast rect path.
Needed for some 2D games, like Criminal Girls.
2022-09-17 13:37:54 -07:00
Unknown W. Brackets 1be5f1670c softgpu: Remove mostly unused screen offset state.
This is now already cleaned up from coords, so no need to cache it.
2022-09-12 22:02:10 -07:00
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
Henrik Rydgård 2791ab3226 Merge pull request #16011 from unknownbrackets/softgpu-rect
Detect more triangles as rectangles in softgpu
2022-09-12 08:35:13 +02:00
Unknown W. Brackets 288d18447d softgpu: Detect full triangles as rectangles.
Seen in a dump from Infected, improves FPS ~25% there.
2022-09-11 22:39:32 -07:00
Unknown W. Brackets 8a6e8066bf softgpu: Store vertex colors as packed RGBA8.
No need to keep it expanded at 4x the space for both colors.
2022-09-11 18:41:06 -07:00
Unknown W. Brackets 2e3b73abaa softgpu: Restrict sprite fast path to throughmode.
Its UV checks already should have generally, but let's be safe.  It
doesn't validate state like fog, etc.
2022-09-11 08:22:37 -07:00
Henrik Rydgård d86127ac5e Merge pull request #15999 from unknownbrackets/softgpu-texsize
softgpu: Clamp/wrap textures at 512 pixels
2022-09-11 10:12:38 +02:00
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 f5f5c9ea87 softgpu: Avoid calling unordered coords tl/br.
These are just corners, we don't know if they're top or not at this point.
2022-09-10 14:58:27 -07:00
Unknown W. Brackets 13ca08b235 softgpu: Avoid over-aggressive rect conversion.
The TL and BR have to match between UVs and pos, not enough for UV to be
in order.  This was causing an artifact on Chains of Olympus' title.
2022-09-10 14:56:13 -07:00
Unknown W. Brackets e7d49cd7d0 softgpu: Allow almost flat rectangles to go fast.
Improves transform rectangles used in Chains of Olympus, for example on
the title screen.
2022-09-10 13:29:40 -07:00
Unknown W. Brackets 6004d4a459 softgpu: Refactor duplicate rectangle compat check.
This just moves the logic to a single place for changes.
2022-09-10 13:28:35 -07:00
Henrik Rydgård a91219bebc The Darkstalkers subpixel check should be == 0. Fixes #14953 again 2022-09-05 19:12:45 +02:00
Unknown W. Brackets 15b73b56bb softgpu: Correct texture check in fastpath.
This is the no textures option, so we shouldn't check sampler state at
all.  Presumably this was failing and reading uninitialized data...
2022-09-03 11:21:14 -07:00