Commit Graph

204 Commits

Author SHA1 Message Date
Gregory Hainaut bef8447447 Merge pull request #1186 from PCSX2/gsdx-unoptimize-vs
gsdx-ogl: make VS more generic
2016-02-20 10:39:19 +01:00
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 0dadc124d3 gsdx-ogl: invert behavior of unsafe user hack
It would be on by default. Unsafe & fast path.

The hack is a security if someone encounters any issue

v2: update Windows gui file
v3: fix typo in tooltip and linux gui
2016-02-18 18:48:13 +01:00
Gregory Hainaut 3f3141c186 GSdx-ogl: limit accurate sprite blending to smaller draw call
Completely restore speed in Zone of Ender

Fix #1093
2016-01-22 09:07:35 +01:00
Jonathan Li 37deeb0d52 gsdx-ogl: Optimise overlap detection algorithm
Vectorise coordinate min/max sorting, and use the ordered runion
instead.
2016-01-16 19:21:56 +00:00
Jonathan Li 2fc3ef8124 gsdx-ogl: Add sprite overlap statistics to debug messages 2016-01-16 19:21:54 +00:00
Jonathan Li 74ace74d50 gsdx-ogl: Group non-overlapped accurate blend sprite draws
Reduces the number of draw calls and barriers when drawing sprites.
2016-01-13 08:07:53 +00:00
Jonathan Li f3b6829c74 gsdx-ogl: Restart collision detection algorithm for remaining sprites
It's useless on its own, but it prepares for the next commit.
2016-01-13 08:02:42 +00:00
Gregory Hainaut 24be4b4969 gsdx-ogl: remove unsafe fbmask of the free SW blending
Initially it was free to do the SW blending because safe fbmask
will already do a sw blending.

Unsafe version uses a fast path with a limited blending. Therefore
SW blending isn't free anymore.

Improve the speed of the previous speed hack (xenosaga 1)
2016-01-03 15:39:45 +01:00
Gregory Hainaut 0d25a0592a gsdx-ogl: fast blending accurate hack
The hack relies on the undefined behavior of the hardware so it can
potentially generate rendering corruption.

This new hack drops the cache flusing when only the alpha channel is masked.
Alpha is a direct copy of the fragment. Normally masked bits will be constant
everywhere (RT, FS output, texture cache) so it would likely work.

Just in case, code is only enabled with the new shiny hack
2016-01-03 15:39:45 +01:00
Gregory Hainaut c0d0e2e89d gsdx-ogl: PrimitiveOverlap requires real rectangle
runion/rempty/rinter requires x < z and y < w

Help issue #762 (accurate blending issue)

If you want to shine, please put better GSVector code (AVX512 is 2 instruction :p)
2015-10-22 12:34:50 +02:00
Gregory Hainaut 6561fbc831 gsdx-ogl: only enable aniso when sampling from the HW texture unit
Potentially help issue #884
2015-10-22 12:21:43 +02:00
Gregory Hainaut 0958b9db8e gsdx-ogl: use the standard openGL name
Function pointer was mangled to avoid any collision. Nowadays all symbols
are hidden so no risk of collision.

Syntax is nicer beside it would allow to put back GLES3.2. I think it
supports most of the used extension.

glActiveTexture & glBlendColor are provided without symbol query.
2015-10-17 17:05:15 +02:00
Gregory Hainaut 03bc304ecf gsdx-ogl: fix colclip regression when accurate blending is enabled
Fix #865
2015-10-16 22:07:50 +02:00
Gregory Hainaut 5fb8c7e65c gsdx: initialize members in constructor of objects
A couple of useless members were removed too.

Also fix wnd initialization

Coverity:
CID 146955 (#1 of 1): Uninitialized pointer read (UNINIT)
18. uninit_use: Using uninitialized value wnd[i].
2015-09-23 09:46:53 +02:00
Gregory Hainaut 9b69afe541 gsdx-ogl: typo that broke TC offset hack 2015-09-20 22:19:45 +02:00
Gregory Hainaut 78569ee833 gsdx-ogl: redo properly the setup of texture format
* add lengthly comment to explain the format
* Likely reduce the number of shader permutation
* Avoid slow AEM (on GPU)

Expect regressions because TC needs some fixes

v2: fix palette mode
2015-09-11 14:14:22 +02:00
Gregory Hainaut ddc03cbccf gsdx: add a nice enum to replace the hardcoded convert shader 2015-09-08 16:51:24 +02:00
Gregory Hainaut a46c11b088 gsdx-ogl: delete various leftovers 2015-08-22 13:33:32 +02:00
Gregory Hainaut 7002ff3ec3 gsdx-ogl: move texture scale from vs_cb to fs_cb
It avoid useless update of the vs_cb.
2015-08-22 13:33:04 +02:00
Gregory Hainaut 499206f0ea gsdx-ogl: disable advance rendering when drivers don't support it 2015-08-22 12:22:35 +02:00
Gregory Hainaut 37f9bcf9cb gsdx-ogl: reduce state change
* don't dirty aref when a fog color is uploaded
* only set clamp mode in clamp mode (region clamp is handled in shader)

v2: fix SSE2/3 compilation
2015-08-20 23:01:43 +02:00
Gregory Hainaut c5a786ed2c gsdx-ogl: remove support WMS/T == 2 in hardware unit
I think behavior was wrong because only first texel coordinate was clamped.

Beside we can't interpolate if AEM isn't yet applied
2015-08-20 23:01:43 +02:00
Gregory Hainaut 6b84a89b6a gsdx-ogl: remove old ati hack for point sampler
Was never used on openGL

In the future absolute coordinate will be use anyway
2015-08-18 19:09:45 +02:00
Gregory Hainaut 37661f3a86 gsdx-ogl: fix blending regression (VP2)
Negative factor must be disabled in negative accumulation "-Cs*As + Cd"
Shader must output Cs*As and blending unit will do the Cd - C subtraction
2015-08-18 09:14:11 +02:00
Gregory Hainaut 260c12756a gsdx-ogl: accelerate Cd - Cs * Alpha in sw blending
It is similar as Cd + Cs * Alpha except the operator

It would help to emulate other color clipping case correctly in basic
mode at full speed
2015-08-16 18:12:34 +02:00
Jonathan Li 6580f4922f gsdx-ogl: correct fbmask conversion on texture shuffle
Fbmask is RGBA8 and must be converted to RGB5A1
=> bit 31 must be moved to bit 15

Signed-off-by: Gregory Hainaut <gregory.hainaut@gmail.com>
2015-08-16 14:06:10 +02:00
Gregory Hainaut c32f1dfc23 gsdx-ogl: fix blend clear regression
Field selector are very dangerous to use.

Fix #764
2015-08-15 20:06:34 +02:00
Gregory Hainaut 0e3491bd58 gsdx-ogl: always do RW RT check
Jak uses it with triangle rendering.
2015-08-15 20:06:34 +02:00
Gregory Hainaut c7b2f9d1d2 gsdx-ogl: setup fbo to always write to first buffer
GL_NONE was kind of useless because nothing was attached anyway
2015-08-15 20:06:34 +02:00
Gregory Hainaut 87f497af0a gsdx-ogl: remove an useless variable 2015-08-15 20:06:34 +02:00
Gregory Hainaut 6046d6c417 Merge pull request #749 from PCSX2/gsdx-micro-optimization
Gsdx micro optimization
2015-08-12 08:54:36 +02:00
Gregory Hainaut 98c74879bf Merge pull request #718 from PCSX2/depth-color-direct-write
Depth color direct write
2015-08-10 15:50:48 +02:00
Gregory Hainaut 9f92f63194 gsdx-ogl: Use GetAlphaMinMax to limit the scope of FULL accurate blending
Provide a massive speed up in this level.
2015-08-09 13:44:31 +02:00
Gregory Hainaut 61694013a5 gsdx-ogl: compact blending parameter structure
Save 656B of data. It is good for the cache.
2015-08-09 13:44:30 +02:00
Gregory Hainaut df3ade896b gsdx-ogl: use integer for blend factor
Integer argument&comparison might be lighter

V2: Forget to change one OMSetBlendState call
2015-08-09 13:44:05 +02:00
Gregory Hainaut 5b57405517 gsdx-ogl: blend management cleanup
* reorder the blend function
* remove OM bsel object
* add a bit to support pabe (miss the glsl part)
2015-08-08 09:18:09 +02:00
Gregory Hainaut 4d12410707 gsdx-ogl: latch constant buffer in rendering object
* Initialization of the object is done once
* Avoid to reupload it when an useless parameter toggle
 => -10% of UBO update
2015-08-08 09:18:09 +02:00
Gregory Hainaut b17803bb34 gsdx-ogl: wipeout of GL_ARB_bindless_texture
Code is completely broken. It doesn't help to improve speed.

Remove 200 lines ;)
2015-08-08 09:16:49 +02:00
Gregory Hainaut b7e16b5989 gsdx-ogl: clean the blending management
Intially GSBlendStateOGL was an alias of the m_blendMapD3D9 array

The object was replaced by an index in the array. Save 2k of memory duplication.
And too much useless code.

v2: push/pop blending state in DATE stuff
v3: remove m_state which is useless now
2015-08-05 22:55:12 +02:00
Gregory Hainaut 717f0fcb4d gsdx-ogl: optimize fbmask setup 2015-08-05 22:55:12 +02:00
Gregory Hainaut 73e2ff6ff6 gsdx-ogl: change PrimitiveOverlap algo from O(n^2) to O(n)
+ only enable this optimization when it is useful (date or sw blending)

Less impact on the perf even for big vertex array
2015-08-05 17:59:55 +02:00
Gregory Hainaut 36554c3375 Merge branch 'hdr-colclip-32bits' 2015-08-04 21:55:40 +02:00
Gregory Hainaut 45bb27d6db gsdx-ogl: extend HDR colclip to 32 bits texture
Unfortunately 16 bits wasn't enough for Castlevania.
2015-08-04 21:54:27 +02:00
Gregory Hainaut c6ff7531fb gsdx-ogl: performance boost on virtuafighter 2015-08-04 21:26:03 +02:00
Gregory Hainaut d80aa0b0bd gsdx-ogl: remove 2 printfs
GL_INS is a better tracing solution
2015-08-04 20:47:02 +02:00
Gregory Hainaut 744f9ebc09 gsdx-ogl: rare corner case when both texture shuffle and date are enabled
In texture shuffle mode the texture data is either RG or BA. It means
that DATE must either checks MSB of G or A.

Close #693
2015-08-04 20:10:44 +02:00
Gregory Hainaut 3784ea768f gsdx: check null pointer when doing a texture clear 2015-08-04 19:26:17 +02:00
Gregory Hainaut 8424c18e9f Merge pull request #688 from PCSX2/hdr-colclip
Hdr colclip
2015-08-02 18:13:28 +02:00
Gregory Hainaut 1f402b1b56 gsdx-ogl: fix bad detection of overlapping
avoid rendering corruption with SW blending
2015-08-01 13:29:25 +02:00