Commit Graph

1121 Commits

Author SHA1 Message Date
Gregory Hainaut 89d5e5637c glsl: use an explicit cast insead of notEqual function
If the compiler didn't optimize the code, it will be a bit faster
2015-04-24 18:01:25 +02:00
Gregory Hainaut 56836561f4 glsl: replace runtime condition by preprocessor condition
It might be an easier work for the compiler

I didn't replace all occurences to keep readability
2015-04-24 17:51:29 +02:00
Gregory Hainaut 4bb8d15228 gsdx: be more verbose on title and bandwidth debug 2015-04-24 17:13:56 +02:00
Gregory Hainaut 258b73409c gsdx-ogl: update flags for buffer storage
Fix issue with Mesa driver
2015-04-23 21:10:43 +02:00
Gregory Hainaut f6652e9a50 gsdx-ogl: disable slow and buggy code until I found a better solution
ogl_texture_storage 1 creates texture corruption.

Advance date is too slow, code need to be updated (properly) to uses 2 passes only not 3
Maybe one could be enough (sometimes)
2015-04-22 09:33:41 +02:00
Gregory Hainaut b32f808fd4 gsdx-ogl: increase the number of pbo
It would cost 16MB of extra storage on the GPU but it might
reduce conflict of texture upload.
2015-04-22 00:40:38 +02:00
Gregory Hainaut bd6ea17bdc gsdx-ogl: speed improvement for DATE
DATE is implemented in 2 ways.
1/ with stencil
2/ purely in FS (sw)

I kept method 1 to reduce the work on method 2. It sucks for performance.
So it would be either 1 or 2.

Note: DATE has a big impact on higher upscaling
Note2: you can disable the 2nd method with this configuration parameter

override_GL_ARB_shader_image_load_store = 0
2015-04-22 00:36:34 +02:00
Gregory Hainaut 15dcf07b3b revert previous commit
Not better, worst slower.

I'm afraid I will need proper fencing
2015-04-22 00:32:46 +02:00
Gregory Hainaut 8386b427ea gsdx ogl: restore GL_MAP_COHERENT_BIT for texture upload
I hope to fix the texture upload corruption
I hope that impact on perf will remain small
2015-04-21 23:34:26 +02:00
Gregory Hainaut 19eb1f00d1 gsdx ogl: flush vbo range instead of barrier
For testing purpose. I don't know which one is better.

It seems flushing have less fps fluctuation than barrier.
2015-04-21 21:44:50 +02:00
Gregory Hainaut ce98276322 gsdx-ogl: improve speed of vertex streaming
Note yet enabled because I'm afraid of data corruption but feel free to test it

The option:
ogl_vertex_storage = 1

Performance note (warm cache+gs replay on colin3)
60 fps -> 76 fps
2015-04-20 09:38:03 +02:00
Gregory Hainaut 62489f42f1 gsdx-ogl: add an optimization note for later
Only 1 byte of fog is useful
2015-04-20 07:18:09 +02:00
Gregory Hainaut 6d253c0b8f glsl: fix debugging of tex coordinate in apitrace 2015-04-20 07:18:08 +02:00
Gregory Hainaut 31f8c065db gsdx-ogl: implement a new hack UserHacks_UnscaleSprite for opengl
UserHacks_UnscaleSprite = 1 will unscale flat sprites
UserHacks_UnscaleSprite = 2 will unscale all  sprites (don't work well so far)

The idea of the hack is to redo the interpolation of texture coordinate
based on the non-upscaled pixel position.

It avoids various glitches but sprites aren't upscaled anymore (so no
more anti-aliasing, potentially a coefficient can be added).
2015-04-20 07:18:08 +02:00
Gregory Hainaut 6124eb844e gsdx-ogl: only compile useful VS
logz is a constant
wildhack is only compatbile with TME/FST

Compilation goes down from 64 to 20 vertex shaders.
2015-04-20 07:17:58 +02:00
Gregory Hainaut 16e6d0d305 glsl: move shader into a separate directory
Only keep glsl_source.h for clarity
2015-04-19 18:49:02 +02:00
Gregory Hainaut 55fdf26898 glsl: remove the older file tfx.glsl 2015-04-19 18:49:02 +02:00
Gregory Hainaut 15264c6c63 glsl: split the main shader
* separate VS/GS and FS
* separate subroutine part of the FS

It already complex enough without subroutine stuff. Besides I'm not sure
we will keep subroutine on the future.
2015-04-19 18:49:02 +02:00
Gregory Hainaut 6fc9afb175 Merge pull request #507 from PCSX2/stdcall-for-plugin
Stdcall for plugin
2015-04-19 18:48:32 +02:00
Gregory Hainaut e605ed1d09 gsdx-queue: add a comment for the future 2015-04-17 19:12:36 +02:00
Gregory Hainaut fa243afbab gsdx SW: enable new queue && C++11 on linux/MSVC 2012+ 2015-04-17 19:12:36 +02:00
Gregory Hainaut d91e989abb gsdx-queue: pass shared_ptr by reference
It avoids atomic +1/-1 of the reference counter

The counter is still incremented when the ptr is copyed into the queue
2015-04-17 19:12:36 +02:00
Gregory Hainaut 84b33d2ddb gsdx-queue: plug the new queue as a drop-off replacement of previous boost queue 2015-04-17 19:12:36 +02:00
Gregory Hainaut 90794c302a gsdx-queue: import spsc_queue of boost
I remove 80% of the file to only keep the ring buffer core

Same speed as boost but without the boost dependency
2015-04-17 19:12:36 +02:00
Gregory Hainaut c9194301a0 gsdx-queue: (linux) add a GUI option to select the queue 2015-04-17 19:12:33 +02:00
Gregory Hainaut 0aac47ca59 gsdx-queue: add a new option "spin_thread" to select the queue behavior at runtime
If someone has a more elegant solution, feel free to share it

spin_thread = 0
spin_thread = 1 // the faster but GS thread will never stop, very bad for laptop
2015-04-17 19:03:21 +02:00
Gregory Hainaut 9682061472 gsdx-queue:add a new job dispatcher queue based on boost and C++11
It is faster on linux, it requires less code, and it is "portable"

It requires boost (only hpp files) + MSVC 2013 (for atomic) (seem doable by 2012 too)

Actually there are several queues that either use spinlock or full sleep
2015-04-17 19:03:21 +02:00
Gregory Hainaut a75d78bd7e gsdx: use standard lock_guard instead of GSAutoLock 2015-04-17 19:03:21 +02:00
Gregory Hainaut 9ad5933120 gsdx: Use composition insead of inheritance to support lock
To ease update to C++11
2015-04-17 19:03:21 +02:00
Gregory Hainaut 8deee6afbc gsdx: include some C++11 define for later 2015-04-17 19:03:21 +02:00
Gregory Hainaut c456764342 spu2x: declare exported function as stdcall
* Fix a stack buffer overflow on address sanitizer
* Might explain an old clang crash
2015-04-17 18:24:23 +02:00
Gregory Hainaut 545c1d387c gsdx: add an include for MSVC 2015-04-16 22:28:09 +02:00
Gregory Hainaut aff999cf69 gsdx linux: refresh the gui
* Use tooltip for hack
* Update string of previous hack
* Remove unused hack

Note: hack_sprite_check requires 3 states (and potentially others hack too) but
I don't know how to do it. It likely requires a "scale button"
2015-04-16 19:28:55 +02:00
Gregory Hainaut 85cbe285f0 gsdx dialog: move hack info into a separate file
This way it can be reused on linux.
2015-04-16 19:26:33 +02:00
Gregory Hainaut 670bcc1818 gsdx hack: use a dumb implementation for UserHacks_round_sprite_offset = 2
It works as bad as a "clever" implementation.

It seems to be enough for games such as venus/taisho-monoke/FFX

Note: it might creates glitches. Code will never be nice, so it is just
a trade-off
2015-04-16 19:26:33 +02:00
Gregory Hainaut 41e5be0673 gsdx hack: limit range of UserHacks_align_sprite_X
Just an extra check to avoid an overlap (avoid white line in tekken)
2015-04-16 19:26:33 +02:00
Gregory Hainaut e8698a9c0c gsdx hack: hole in vertex position also appears in !FST projection
Fix various black vertical line regression with UserHacks_align_sprite_X
2015-04-15 09:40:06 +02:00
Gregory Hainaut 65a84b9f64 gsdx hack: drop previous filtering hack
superseeded by UserHacks_round_sprite_offset
2015-04-15 09:26:01 +02:00
Gregory Hainaut 2da4a5e6bc gsdx hack: use a 0 offset instead of -1
-1 is annoying because minimum value is 0. Instead to add more logic,
let's try to use 0 which seems to be good enough (fix regressions on DQ8/AT)

Unfortunately it causes a mini regression on taisho-mononoke. Rotation of sprite is done with 2 triangles.
Potentially previous value wrongly recover this section.
2015-04-14 18:39:26 +02:00
Gregory Hainaut cde72300a5 gsdx hack: Correct formulae for reversed texture
It fixes issues on Ace Combat UI
2015-04-13 22:14:36 +02:00
Gregory Hainaut 418f2e69a8 gsdx-ogl: implement the wildhack on the GPU
Likely much faster for opengl and much easier to implement

Note: hopefully UserHacks_round_sprite_offset will replace it
2015-04-13 22:14:36 +02:00
Gregory Hainaut 50bac01d89 gsdx hack: disable upscaling hack at native resolution 2015-04-13 22:14:36 +02:00
Gregory Hainaut 815c99b6e4 gsdx-hack: remove an extra offset + duplicate linear computing
Initially code aligned texture on pixel boundary + 1/16th. It seems to work without it.
2015-04-13 22:14:36 +02:00
Gregory Hainaut 399ddbd74a gsdx hack: reduce complexity of alpha* function
It will help the compiler to make a couple of optimizations (and it will help
a futur porting to GSVector)
2015-04-13 22:14:36 +02:00
Gregory Hainaut 1d0deb5ffe gsdx hack: move round sprite offset into a separate function
Use a boolean template to separate the linear case. I'm afraid it will
cost extra computation but 90% of sprites use nearest.
2015-04-13 22:14:36 +02:00
Gregory Hainaut 1aebca3740 gsdx hack: m_userhacks_round_sprite_offset take precedence over m_userhacks_stretch_sprite
m_userhacks_stretch_sprite is only kept for debugging. Let's avoid bad mixing by users
2015-04-12 11:05:23 +02:00
Gregory Hainaut c4fef0256e gsdx hack: Better rounding fix letter in ace combat and line in tekken
Tekken is broken with setting UserHacks_round_sprite_offset = 2 use 1 instead.
=> game use upscaling internaly so my rounding of coordinate break everything

Yakuza uses float coordinate but this hack only correct the integer coordinate
=> the solution will be to add a dedicated hack for this game.

Colin3 got a regression but I don't know when...
2015-04-12 10:31:04 +02:00
Gregory Hainaut 7e89247bf4 gsdx hack: don't add extra correct for zero length texture
Otherwise t1-t0 is 16 instead of 0
2015-04-11 10:54:30 +02:00
Gregory Hainaut d2156c86eb gsdx: improve previous hack for bilinear sprite
PS2 uses a -0.5 offset before sampling so texels must be rounded to half-pixel boundary

If fixes glitches on Venus and taisho-mononoke

Note: I didn't fixed yet texture render in reverse because I don't have any test for it.
2015-04-10 09:42:39 +02:00
Gregory Hainaut f5eeb2024e gsdx: anti-glitches v3: Tester wanted
UserHacks_round_sprite_offset = 1 <= enable correction of flat sprites
UserHacks_round_sprite_offset = 2 <= enable correction of all sprites (better on a couple of dump but not sure of the consequence)

I completely redo the algorithm. This time I do the projection and
interpolation of the 2 extrem vertex. This way I can compute the min/max
valid texture coordinate.

It gives stronger guarantee that texture sampling will be done inside the texture.

However it might have a performance impact, likely reasonable because it
is limited to sprite vertex.

A big thanks you to all people that provide me GS dump and test reports.
2015-04-09 19:51:50 +02:00