preserve-emulation/ppsspp
1
0
Fork 0
mirror of https://github.com/hrydgard/ppsspp.git synced 2026-07-23 15:24:49 +02:00
Code Issues Packages Projects Releases Wiki Activity
14,929 Commits 78 Branches 131 Tags
c41e4ef9055df28362b950b7cf99b2873d8a3dac
Commit Graph
298 Commits
Author SHA1 Message Date
Henrik Rydgard 1fe2af3421 Fix D3D NFSU crash #7030 in a cleaner way than #7033 :) 2014-10-29 23:51:23 +01:00
Unknown W. Brackets 2b16b5b79b Use XXH64 on 64-bit systems. 
I'm not able to reproduce any real performance difference, but I know
there are some games which do more hashing and it may help there.
 2014-10-26 17:49:24 -07:00
Henrik Rydgard 7cb5e7f53f Some very minor optimizations. Remove little-used stat counter. 2014-10-19 21:12:25 +02:00
Henrik Rydgard 6c313385ab Reduce the diff between the two draw engines 2014-09-18 00:40:25 +02:00
Henrik Rydgard 5e2c30c640 Move GetCurrentSimpleVertices to DrawEngineCommon 2014-09-17 21:37:53 +02:00
Unknown W. Brackets 01bdb6e160 Cleanup some minor vertex cache stuff. 2014-09-15 07:08:55 -07:00
Unknown W. Brackets 64d955ea49 d3d: Share the bounding box code. 2014-09-14 14:04:09 -07:00
Unknown W. Brackets f0c37e3f61 d3d: Use minihashing in the vertex cache. 2014-09-14 13:50:57 -07:00
Unknown W. Brackets afd9d864be Clear alpha only when clearing stencil. 
Not sure if this will fix anything.
 2014-09-14 08:00:35 -07:00
Unknown W. Brackets d60fd955d6 Vertex cache a bit more frequently. 
Rely upon the minihashing a bit more.
 2014-09-13 14:03:13 -07:00
Unknown W. Brackets 456b2613f0 Avoid full rehashes when failures are frequent. 2014-09-13 14:03:13 -07:00
Unknown W. Brackets c8e6a809c3 Use a more frequent mini hash in the vertex cache. 2014-09-13 14:03:12 -07:00
Unknown W. Brackets 37b9af75b1 Refactor vertex caching a little. 2014-09-13 14:03:12 -07:00
Unknown W. Brackets bcc6b4027a Limit unreliable vertexes uncached per frame. 
Remember them longer, and avoid slamming too many in the same frame.
 2014-09-13 14:03:11 -07:00
Henrik Rydgard 19a9c4481f Move GLES' SoftwareTransform to Common 2014-09-13 13:27:42 +02:00
Henrik Rydgard 4212bbb51e Cleanup SoftwareTransform a bit more 2014-09-13 13:15:18 +02:00
Henrik Rydgard ca3ed34b44 Move most GL code out of GLES/SoftwareTransform.cpp 2014-09-13 13:03:37 +02:00
Henrik Rydgard 37e3cf362f Move vertexdecoder files into GPU/Common 2014-09-12 02:00:32 +02:00
Henrik Rydgard 7cf05e44a2 Add option to vertexdecoder to expand UVs to floats 2014-09-12 02:00:31 +02:00
Unknown W. Brackets 33ac30218a Bind the alphatest tex after we know vert alpha. 
Before we sometimes thought alpha test was trivially true when it wasn't.
 2014-09-10 23:43:48 -07:00
Unknown W. Brackets 94c1271761 Add an experiment with using a texture for tests. 
Using an option for now so it's easy to test, if it works well we can
maybe remove the option.
 2014-09-07 10:04:57 -07:00
Unknown W. Brackets 4a8d52ab1a Add a config option for the slow framebuf effects. 
Like cluts, blitting, etc.  At higher render resolutions these are
expensive, and on some mobile devices they may be too much.

Of course, they're only used when needed so this is a speedhack.
 2014-07-08 23:57:27 -07:00
Unknown W. Brackets 32a11dc28d Minor optimization for prescale/not, fix switch. 2014-06-29 15:47:00 -07:00
Unknown W. Brackets 596abeca06 Use the same vertex decoders for splines. 
Had the vert type bits at the top.
 2014-06-29 15:46:59 -07:00
Unknown W. Brackets 95f4341838 Allow changing prescale uv at runtime. 2014-06-28 21:51:25 -07:00
Unknown W. Brackets 1c3b60a8ee Allow changing software skinning at runtime. 2014-06-28 21:46:43 -07:00
Henrik Rydgard 8b60fe0239 Ridge Racer hack: Flush after every prim if fb addr == tex addr. 
Fixes the bloom errors. @unknownbrackets discovered this method.

Not sure if I want to merge this.. It does add yet another check
to a critical path.
 2014-06-22 20:37:50 +02:00
Unknown W. Brackets 9b12a4893d Track framebuffer memory dirty more carefully. 
Fixes #6296, Grand Knights History slowdown during battle transition.  May
also more correctly identify block transfer downloads.
 2014-06-11 22:21:08 -07:00
Unknown W. Brackets 7226aad903 Don't allow the buf name cache to get too full. 2014-06-07 16:28:24 -07:00
Unknown W. Brackets 5f521bcfb0 Reuse buffer objects when possible. 
glBufferData() recreates anyway.
 2014-06-07 16:20:12 -07:00
Unknown W. Brackets 4971206e92 Allocate vertexcache buffer names in chunks. 
Partially implements #6045.
 2014-06-07 16:15:37 -07:00
Unknown W. Brackets 8d1b3ae7aa Fix a vertexdec crash from the ge debugger. 2014-06-06 21:06:29 -07:00
Unknown W. Brackets 8f9a2b9882 Unbind the fbo texture when it's not used. 
Just in case there are problems with decimating it or etc.
 2014-05-27 01:45:20 -07:00
Unknown W. Brackets ff8f1fa569 Avoid the GE debugger affecting vertexFullAlpha. 2014-05-04 09:54:54 -07:00
Unknown W. Brackets d91ad87daa Small perf improvements in DecodeVertsStep(). 2014-04-05 12:32:10 -07:00
Unknown W. Brackets bb2fb12edd Inline SetupVertexDecoder() in SubmitPrim(). 
Everything SubmitPrim() calls is expensive...
 2014-04-05 01:25:04 -07:00
Unknown W. Brackets e7639f666c Ensure stuff run on every prim are inlined. 
These both get a ton of calls and show up on profiles.
 2014-03-30 00:42:26 -07:00
Unknown W. Brackets 466a5418d2 Cut time in EstimatePerVertexCost(). 
From profling, this cuts 4% time while lighting is enabled, and 14% while
disabled (on ARM.)  But maybe should just inline this...
 2014-03-30 00:42:26 -07:00
Unknown W. Brackets a4327702f1 Reduce some includes under GPU/. 2014-03-29 16:51:38 -07:00
Unknown W. Brackets 69c2500dc6 Make sure vai->flags are set while hashing. 
Since we decode the verts in this case we have a fresh flag.  Might
be #5718?
 2014-03-25 08:19:38 -07:00
Henrik Rydgård cf3d75975d Merge pull request #5720 from unknownbrackets/texcache 
Optimize xxHash for ARM/NEON devices
 2014-03-25 09:32:15 +01:00
Unknown W. Brackets b800762ceb Add a NEON-optimized version of XXH32. 
This takes at least 40% less time to hash on NEON/ARM devices.
 2014-03-25 00:34:55 -07:00
Henrik Rydgard dc07d3410a More checks for alpha test elimination 2014-03-24 17:33:20 +01:00
Henrik Rydgard 2c76e6d023 Correctly keep track of "full alpha" in vertices (x86 jit only). 2014-03-24 11:19:44 +01:00
Henrik Rydgard f2f0355d94 Split up ApplyShader into ApplyVertexShader and ApplyFragmentShader. 
Will allow the alphatest avoiding optimization later.
 2014-03-24 10:55:07 +01:00
Henrik Rydgard b174996c1c Add a conservative check that prevents alpha testing in a few cases. 
This will become really powerful if we add some code to the vertex decoder
to check for non-full alpha in the vertices, and set gstate_c.vertexFullAlpha if none
is found (probably want to do the reverse, set it to true and clear if any non-255 alpha is found).

Alpha testing is a performance killer on many mobile GPUs so big efforts to
avoid it can be worth it.
 2014-03-23 16:32:38 +01:00
Henrik Rydgard 20d480a374 Minor GPU code cleanups 2014-03-23 16:32:38 +01:00
Henrik Rydgard 8b92dcea47 Transform: Compute the "DCID" (draw call ID) incrementally instead of an extra pass. 2014-03-23 01:51:51 +01:00
Henrik Rydgard b4d99b1981 Revert "Avoid caching when HW T&L with morph enabled." 
This reverts commit 557eae7ca9.
 2014-03-15 10:46:04 +01:00
raven02 557eae7ca9 Avoid caching when HW T&L with morph enabled. 2014-03-14 21:04:32 +08:00