Mirror/ppsspp
1
0 Fork 0
mirror of https://github.com/hrydgard/ppsspp.git synced 2026-10-01 14:58:14 +00:00
Code Issues Packages Projects Releases Wiki Activity
48,144 Commits 93 Branches 131 Tags
master
Commit Graph
42 Commits
Author SHA1 Message Date
Henrik RydgårdandClaude Opus 5.5 9d3022642b SoftGPU: Fix bin queue bookkeeping and dependency tracking 
- Flush() on an empty queue now trims the state and CLUT rings, since its
  callers flush because one is full and push right after.
- BinQueue::Full() uses >=, so an overshoot can't go unnoticed.
- IsExactSelfRender compares against the target the queued draws were
  binned for, not gstate, which already has the next one during a flush.
- A depth test without depth writes marks the depth buffer as read.
- The DarkStalkers untextured sprite recomputes the binner state around it.

Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
 2026-09-29 13:16:40 -06:00
Henrik RydgårdandClaude Opus 5.5 93f57a4f38 GPU: Delete copy operations on classes that own resources 
These own GPU objects, memory or refcounts in their destructors (or assert
there that they were torn down), so a copy would double-free. Nothing copies
them today; this keeps it that way. The manager base classes cover every
backend's subclass.

Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
 2026-09-29 10:33:32 -06:00
Henrik RydgårdandClaude Opus 5.5 a9c782bfcf SoftGPU: Recompute the raster state after a JIT cache clear 
A code space clear frees the functions the current state points to, but
the state was kept as long as the GE registers didn't change. Track the
clear generations and recompute, also when a compile during the state
computation clears the caches.

Also skip the binner flush for compiles on builds without the software
JIT, where Compile() does nothing.

Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
 2026-09-29 09:35:17 -06:00
Henrik Rydgård f60e27a9b7 Just some refactoring of the GPUStatistics struct, and more use of StringWriter 2026-05-29 14:40:31 +02:00
Herman Semenov b57dab2812 [GPU] Make static and const methods if possible 2024-04-05 17:04:31 +03:00
Unknown W. Brackets 59fb374c38 softgpu: Small optimization to clut updates. 2023-04-16 10:16:06 -07:00
Герман Семенов 122b63b9a8 GPU: using if constexpr C++17 optimization 2023-04-02 16:36:37 +02:00
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
Unknown W. Brackets 4d92533907 softgpu: Apply optimizations to states generically. 
This is for optimizations we can only do when we know the vertex values.
 2022-12-02 21:30:53 -08:00
Unknown W. Brackets 7e7eabf86a softgpu: Use threads on self-render if safe. 
Some games, such as Prince of Persia, self-render 1:1 to colorize, rather
than to scale or blur.  We can still use threads in these cases.
 2022-11-13 12:34:31 -08:00
Unknown W. Brackets 24560eef5c softgpu: Avoid waiting for a thread to drain. 
If we can, we want to keep the thread queues healthy, but not full.
Reduce the amount we push on a typical drain to avoid the Wait().
 2022-09-24 20:01:00 -07:00
Unknown W. Brackets 1aa6841759 softgpu: Increase queued prims. 
We made them smaller, so we can queue more of them in the same space.
Helps a little bit.
 2022-09-24 20:01:00 -07:00
Henrik Rydgård d743bfac93 Merge pull request #16085 from unknownbrackets/softgpu-vert 
softgpu: Cache reused indexed verts
 2022-09-24 00:00:26 +02:00
Unknown W. Brackets 66b6dfd0a5 softgpu: Fix self-render detect in Ridge Racer. 
When we flush we mark all pending writes zero, but we rely on this being
set to detect self-render.

TRANSFORM_ALL was wrong as well, sometimes clearing BINNER_RANGE.
 2022-09-22 20:36:15 -07:00
Unknown W. Brackets 84a3f6de71 softgpu: Remove unnecessary state param. 
Oops, meant to remove this when refactoring imm prims.
 2022-09-22 18:18:49 -07:00
Unknown W. Brackets 5f2e20d8ca softgpu: Reduce some minor bin item field sizes. 2022-09-11 18:39:14 -07:00
Unknown W. Brackets 7a83f8bab5 softgpu: Use vertType prim override for flags. 
These parameters are a real shame, was so clean before...
 2022-09-06 22:20:45 -07:00
Unknown W. Brackets 88e8f95293 softgpu: Flush on transfer to pending tex read. 
Potentially could use these for self-render, but so far we should be
detecting that so leaving it alone.
 2022-08-20 23:22:21 -07:00
Unknown W. Brackets a3f682fc5a softgpu: Reduce memory usage on 32-bit. 2022-03-13 13:34:14 -07:00
Unknown W. Brackets e68b16af69 softgpu: Enure aligned bin queues. 
On 32-bit, we still want these aligned to 16 bytes.
 2022-03-13 13:33:19 -07:00
Unknown W. Brackets cc6491342e softgpu: Prepare dedicated rectangle path. 
We're still sometimes using the slow rect-as-triangles path, let's do
something faster.  As a first step, just handle binning.
 2022-02-20 09:38:51 -08:00
Unknown W. Brackets a66377fdf1 softgpu: Remove offset from screenpos. 
This simplifies tighter calculations, and reduces the common magnitudes
we'll be dealing with.
 2022-02-19 20:38:44 -08:00
Unknown W. Brackets 2479d52202 Global: Reduce includes of common headers. 
In many places, string, map, or Common.h were included but not needed.
 2022-01-30 16:35:33 -08:00
Unknown W. Brackets 9ea5367a8c softgpu: Add dirty flags for rasterization state. 2022-01-23 08:08:41 -08:00
Unknown W. Brackets 76f9103e97 softgpu: Add a table and initial dirty flags. 
Not actually using the dirty flags to skip state, but have moved to
Execute_* functions and everything else like other graphics backends.
 2022-01-23 08:08:40 -08:00
Unknown W. Brackets c0c3f7284a softgpu: Avoid flush texturing from stride. 
This generally detects overlap more accurately using a dirty rectangles
approach.  Also detects render to self much more accurately, including
with depth.
 2022-01-20 18:39:01 -08:00
Unknown W. Brackets c4c54730bf softgpu: Remove bin asserts. 
These are active in release and used in tight loops.
 2022-01-20 16:59:38 -08:00
Unknown W. Brackets 55c11425e4 softgpu: Use persistent bin task state. 
It's constant, so it's better to avoid the copying and allocation.  A
small win, but removes new from the profile.
 2022-01-20 16:58:43 -08:00
Unknown W. Brackets 83adc44c2b softgpu: Heuristic to avoid over-draining. 
Some games (i.e. VC3) benefit from an early drain, since they get more
done while processing more verts.  Others finish the draw quickly, and
then cause significant overhead in queueing new threads.

This attempts to balance the two, and improves Call of Duty and Blade
Dancer.
 2022-01-16 21:09:28 -08:00
Unknown W. Brackets 1764111a4b softgpu: Reduce wasted memory. 2022-01-16 11:49:41 -08:00
Unknown W. Brackets cb5ac04d16 softgpu: Tune some queue sizes for perf. 
Using a chunk of RAM for this, but mostly with many threads.
 2022-01-16 11:27:43 -08:00
Unknown W. Brackets d95475e021 softgpu: Expose flush reasons/times in debug stats. 2022-01-16 11:27:42 -08:00
Unknown W. Brackets cc155ec460 softgpu: Avoid texture/CLUT flush unless overlap. 
Only need to flush here if there's some overlap in the target.
 2022-01-16 08:22:13 -08:00
Unknown W. Brackets 9466dc6397 softgpu: Flush on offset changes. 2022-01-16 08:14:10 -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 b42ebe15d8 softgpu: Fix off-by-one size limit on bin queues. 2022-01-15 21:59:23 -08:00
Unknown W. Brackets 2539fb7c3c softgpu: Tune queue push/pop to reduce overhead. 
These aren't safetly atomic with concurrent pushers or poppers, but as
long as there's only one of each, they're still safe.

Shaves a decent % off Drain time for heavy scenes.
 2022-01-15 20:18:49 -08:00
Unknown W. Brackets 3134bd1ff9 softgpu: Cleanup push/pop atomic handling. 
Two concurrent push/pops would hazard, though we don't do that.
This improves perf a bit by avoiding an atomic read again.
 2022-01-15 00:02:31 -08:00
Unknown W. Brackets 5a35525fd4 softgpu: Enqueue batches of prims when binning. 
This cuts some thread overhead.
 2022-01-14 20:19:32 -08:00
Unknown W. Brackets 970e9c2f51 softgpu: Move threading into BinManager. 
This threads much more effectively, across entire prim call.
 2022-01-13 22:45:23 -08:00
Unknown W. Brackets 48ef4a18b1 softgpu: Handle scissor/range in BinManager. 2022-01-13 19:07:41 -08:00
Unknown W. Brackets a0a9b1e89b softgpu: Add class to manage and enqueue for bins. 
For now, just forwarding.
 2022-01-13 09:26:59 -08:00