Commit Graph

326 Commits

Author SHA1 Message Date
Gregory Hainaut d7c1faf563 gsdx ogl: add GPU timers to measure time between 2 vsync
So far only basic stuff (min/mean/max frame rendering time)
2016-06-26 15:34:36 +02:00
Gregory Hainaut 79587215bb gsdx ogl: add the option force_texture_clear for test purpose
Might be completely useless.

1 => always clear framebuffers and textures to black (aka 0)
2 => always clear framebuffers and textures to red
2016-06-24 18:41:55 +02:00
Gregory Hainaut 3234c8241b gsdx ogl: massively extend glsl self test
* Support Mesa Nouveau IR (free driver for Nvidia's GPU)
  => Print intermediate representation + final shader
  => Dump GPR usage
* Move dumped shader in /tmp/GSdx_Shader/<sub_dir>
  =>  Avoid the landing of 3 thousands of files in $PWD ^^
* Use function instead of macro
2016-06-11 13:34:37 +02:00
Gregory Hainaut 1c8de02c8d gsdx ogl: trace shader permutation
Too much permutation is bad for driver performances
2016-06-11 13:34:23 +02:00
Gregory Hainaut fca2661e05 gsdx ogl: add a pretty name to various opengl opengl 2016-06-09 18:27:58 +02:00
Gregory Hainaut 7202cac7d0 gsdx ogl: remove the dual source blending workaround 2016-06-01 21:00:29 +02:00
Gregory Hainaut 08a8bfa76c gsdx ogl: plug new program compilation for bad driver 2016-06-01 21:00:29 +02:00
Gregory Hainaut 405f312fe8 gsdx ogl: format GSShader code
Move the Geometry Shader to the call of the function
2016-06-01 21:00:29 +02:00
Gregory Hainaut 5d49a6b685 gsdx ogl: replace 4 VS shader variation by an AND mask
Perf will be roughly the same. However there is a single VS for all
the HW emulation.
2016-06-01 09:29:56 +02:00
Gregory Hainaut 959abe64f8 gsdx ogl: implement wildhack on the CPU
Speed impact is likely small and the plan is only to keep a single Vertex Shader
2016-06-01 09:29:56 +02:00
Gregory Hainaut f7ddd488e1 gsdx ogl: Extend uniform buffer with channel parameter
Instead to use the standard ps ubo which is used every draw call.
I reused a barely used buffer to reduce the extra cost of the upload
2016-05-29 10:13:43 +02:00
Gregory Hainaut 6f19d928f6 gsdx ogl: use the new uniform cache upload method 2016-05-29 10:13:43 +02:00
Gregory Hainaut 179681ef18 gsdx ogl: enable AMD driver blending workaround only on latest legacy driver
Legacy GPU:
Older driver will be broken.

Still supported GPU:
Please upgrade to the latest AMD driver 16.5.2 or 16.5.3 (and prey that future driver will still work)
2016-05-28 18:43:55 +02:00
Gregory Hainaut a9c18c57e4 gsdx option: use the new GetConfig* function
v2:
add PSX stuff
ssakash review
2016-05-28 10:07:27 +02:00
Gregory Hainaut 99476f4b13 gsdx ogl: explain better previous clear vs framebuffer comment 2016-05-16 16:49:18 +02:00
Gregory Hainaut 1522cba5b0 gsdx ogl: performance note of texture clear vs framebuffer clear
In case of render target. FB clear is better
2016-05-16 10:38:47 +02:00
Gregory Hainaut 3ab12cef2f gsdx ogl: accelerate special case of accurate date.
Game often uses date to allow a single pixel pass. If this
use case is detected, stencil buffer will be cleared after first pixels
that pass both depth&stencil test.

It seems to reduce the load on the GPU.

Note: with the help of texture barriere, maybe we could implement the algo
with a single pass.
2016-05-15 17:22:58 +02:00
Gregory Hainaut 5b061e062c gsdx ogl: replace ClearRenderTarget_i by glClearTexSubImage
Avoid state change, avoid potential texture buffer reallocation

Note: require GL_ARB_clear_texture
2016-05-15 15:55:31 +02:00
Gregory Hainaut c054b097e9 gsdx ogl: fix wrong depth clear
If a color buffer is still attached and is smaller than depth buffer,
the latter won't be fully cleared.

As a faster alternative, use GL4.4 clear texture function. Avoid to fiddle with
framebuffer and pixel tests.

Fix #1362x Ar Tolenico 2 map clip
2016-05-15 14:57:37 +02:00
Gregory Hainaut 913e3d65d2 gsdx ogl: use glViewportIndexedf and glScissorIndexed 2016-05-14 17:18:16 +02:00
Gregory Hainaut cd9ee3c468 gsdx ogl: emulate GL_ARB_draw_buffers_blend and GL_ARB_viewport_array
New functions only set the parameter to a single buffer whereas old function
set all buffers
2016-05-14 17:15:30 +02:00
Gregory Hainaut d41613c46a gsdx ogl: add a Tales of Abyss HLE shader
Again fast and efficient but it relies on CRC

v2: forget to update the precompiled shader...
2016-05-07 22:46:41 +02:00
Gregory Hainaut e890ce989c gsdx: throw oom exception + error
Texture can't be NULL anymore, so we can remove a couples of check
2016-05-05 18:53:51 +02:00
Gregory Hainaut d58b71688b gsdx debug: auto GL_POP
Nicer and exception safe

V2: miss definition of vs/ps
2016-05-05 18:53:32 +02:00
Gregory Hainaut 5948c2d8cd gsdx AMD: broken for broken
Until AMD release the driver with a fix, I can't use 2nd blending source with SSO.

So let's use the first source. Blending/Alpha will be wrong. But it is likely better
than an uninitialized alpha value.
2016-05-05 12:24:55 +02:00
Gregory Hainaut 14e1ed06df glsl: add an HLE shader for Urban Chaos
Pro:
* Replace 140 draw calls into a single one
* No complex texture conversion/lookup
* smaller solution than a generic solution
2016-04-30 14:52:53 +02:00
Gregory Hainaut c445a14c46 gsdx ogl: extend shader to lookup a single channel 2016-04-28 22:56:38 +02:00
Gregory Hainaut fda511a949 gsdx glsl: extend hw shader to sample depth texture
Will use integral coordinate to avoid any rescaling.

Bilinear interpolation isn't supported. I don't think it is allowed to
filter a depth texture anyway.
2016-04-24 22:18:26 +02:00
Gregory Hainaut b4b878ac24 gsdx-ogl: add a note for a future small optimization 2016-04-24 22:14:51 +02:00
Gregory Hainaut 4281b8630b gsdx ogl: remove the useless shadeboost Constant Buffer 2016-04-24 11:08:14 +02:00
Gregory Hainaut d027ed2092 gsdx ogl: merge GSTextureFXOGL to GSDeviceOGL
First file is rather small in openGL. And it is linked to the latter.
2016-04-24 10:55:22 +02:00
Gregory Hainaut 821e3ff294 gsdx-ogl: reduce length of debug message 2016-04-21 09:28:44 +02:00
Gregory Hainaut 8217c717f6 gsdx-ogl: require GL_KHR_debug
Will reduce the slowdown on debug build
2016-04-21 09:28:44 +02:00
Gregory Hainaut 9598417f42 gsdx-ogl: GL_ARB_copy_image is now mandatory 2016-04-21 09:28:44 +02:00
refractionpcsx2 dcb676765e GSdx: Clean up some warning under windows (#1289) 2016-04-14 11:00:58 +01:00
Gregory Hainaut 22a40b65ae gsdx-ogl: always use 1 for GL_UNPACK_ALIGNMENT
Avoid a gl function call for each texture uploads
2016-04-11 12:45:11 +02:00
Gregory Hainaut 2941adf364 gsdx ogl: use the new pipeline API
Pre build all SW shader into pipeline.

Directly bind the pipeline instead of all pipeline stages. (less work for the driver)
2016-04-10 17:28:05 +02:00
Gregory Hainaut 95e3dcb448 gsdx-ogl: improve the shader program management
* keep a reference of program/pipeline created to ease the deletion
* extend a bit the API to support multiple pipeline

Final goal will be to use a pre link pipeline for SW shaders. And uses
the default pipeline for HW shaders.
2016-04-10 17:05:33 +02:00
Gregory Hainaut 3f404c8edb gsdx-ogl: update shader pipeline intertace to use vs/gs/ps triplet
Better to have 1 function calls with 3 parameters rather 3 functions call with 1 parameter.
2016-04-10 14:14:30 +02:00
Gregory Hainaut 7cc13ce99c gsdx-ogl: only bind the FB to the draw FB 2016-04-10 11:31:40 +02:00
Gregory Hainaut d533c393b2 gsdx ogl: rely on DSA API 2016-04-10 11:31:40 +02:00
Gregory Hainaut 52e3c3516d gsdx-ogl: GL_ARB_separate_shader_objects is now mandatory 2016-04-07 22:11:35 +02:00
Gregory Hainaut f751f70b1e gsdx ogl: GL_ARB_clip_control is now mandatory 2016-04-07 21:57:54 +02:00
Gregory Hainaut cb279ef321 gsdx ogl: help compiler to devirtualize 2016-04-04 23:12:31 +02:00
Gregory Hainaut 771583c559 gsdx-ogl: help the compiler to propagate stride constant
Optimize various multiplication

Note: potentially we can do it with constexpr but I'm not sure of compiler support
2016-04-04 22:46:31 +02:00
Gregory Hainaut f8c442cf76 gsdx-ogl: make VS more generic
Texture coordinate could be dummy/float/int integral/int normalized.

Old behavior:
* VS was in charge to select the texture coordinate
* int integral format wasn't supported

New behavior:
* Always compute all formats
* FS will be in charge to select the good format

Impact:
* VS will be slightly slower but it reduces shaders permutation from
   little to 0 (won't be bad for CPU)
* FS speed isn't impacted as 2 separate code paths were already required
  to support both format
* Rasterizer will be 33% slower but unlikely to be the limited factor of
  the GPU
* In future we could directly use the integral format in the FS.

V2: remove useless PSin_t
2016-02-18 19:02:05 +01:00
Gregory Hainaut 3451f7e760 gsdx-ogl: handle invalid gl_lengh in debug message
Mesa intel sets it to -1.

Close #1089
2016-01-07 21:49:17 +01:00
Gregory Hainaut ef3aa17025 gsdx-ogl: disable useless Nvidia driver message 2015-11-12 17:04:54 +01:00
Gregory Hainaut 5cbd0cf42a gsdx-debug: add more push/pop debug group on the init 2015-10-24 14:14:37 +02:00
Gregory Hainaut 9ba949c2d9 gsdx-debug: support correct logging on mesa
By default low severity message were disabled. (thanks to be open source)
2015-10-24 14:14:37 +02:00