libretro/ppsspp
1
0
Fork 0
mirror of https://github.com/libretro/ppsspp.git synced 2024-11-25 17:19:42 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity
16,024 Commits 19 Branches 54 Tags 247 MiB
39cb74a8d3
Commit Graph

2304 Commits

Author SHA1 Message Date
Unknown W. Brackets
9af40cb14f d3d: Update the texture cache to mostly match. 
Now we can reduce the differences down.
 2014-09-09 00:53:01 -07:00
Unknown W. Brackets
bc6a67babf d3d: Use unsigned Z in throughmode. 2014-09-08 21:06:26 -07:00
Unknown W. Brackets
b45711ba94 d3d: Centralize parts of the texture cache. 
Should merge these.
 2014-09-08 20:55:56 -07:00
Unknown W. Brackets
04147827ee Account for texel offsets in the frag test lookup. 
Darn, thought this wouldn't be necessary but it is is for 1/255 and other
values.

Without cache:
PowerVR: temp, add, mul, fract, mul, compare
Other: mul, add, floor, compare

With cache: mul, add, texture lookup, compare
 2014-09-08 19:04:44 -07:00
Unknown W. Brackets
ac97f1ed96 Avoid a very unlikely fbo leak. 2014-09-08 07:41:23 -07:00
Unknown W. Brackets
7e248376f8 Track the last drawn width/height/format. 
This way we can apply render-to-self effects more correctly.
 2014-09-08 07:40:43 -07:00
Unknown W. Brackets
76d5f67796 Arg, fix color testing from frag test cache. 2014-09-07 23:21:22 -07:00
Unknown W. Brackets
2532639f70 Refactor fbo binding into DrawPixels(). 2014-09-07 21:53:22 -07:00
Henrik Rydgård
1985e6db9a Merge pull request #6853 from unknownbrackets/testcache 
Cache fragment test results using a texture
 2014-09-07 23:55:53 +02:00
Unknown W. Brackets
072e9c3a0d Use a more logical name for the test tex creator. 2014-09-07 14:52:22 -07:00
Unknown W. Brackets
c0908421ba Skip binding when the test won't be used. 2014-09-07 10:06:57 -07:00
Unknown W. Brackets
95a620695b Actually cache the last bound test texture. 2014-09-07 10:06:23 -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
b20cd3cba2 Clear stencil on switch from 565 buffer. 
Fixes #2094, shadows in Kingdom Hearts in some scenes.

Not a complete solution, but should be a relatively safe one.
 2014-09-07 08:47:46 -07:00
Unknown W. Brackets
47e98f972d Try to simplify alpha replace for poor compilers. 2014-09-07 08:25:39 -07:00
Unknown W. Brackets
de2c951665 Detect EXT_gpu_shader4 to use bitwise ops pre-3.3. 
This may allow it to work on more cards, and should fix a (probably driver
bug) error we've had reported in GL 3.0 / GLSL 1.3.
 2014-09-04 00:39:54 -07:00
Unknown W. Brackets
514772e18e Avoid some magic numbers. 2014-08-30 22:06:25 -07:00
Unknown W. Brackets
337b34ef6a Eat cycles during block transfers. 
Tests show this takes a number very close to 1.9 times the number of bytes
in cycles @222.

Using that value caused a tiny fps drop in one area of the God of War
demo, which may be correct but might also cause complaints.  A lot of our
other timing is not accurate, so using a lower value (1.6) to be on the
safer side.

This may cause things to reschedule more accurately when games drawsync,
and improves the gpu/commands/blocktransfer test.
 2014-08-30 22:02:28 -07:00
Unknown W. Brackets
5e07a40a83 Oops, buildfix Android/etc. 
This is not declared there.
 2014-08-30 07:13:36 -07:00
Unknown W. Brackets
361b9a2572 Use 8888, always, for Danganronpa's FBOs. 
This should fix dithering on draw causing problems for Adreno 3xx devices.
 2014-08-30 02:05:51 -07:00
Unknown W. Brackets
5a3a972d95 Perform intra-buffer blits directly. 
As far as I can tell, this doesn't need a temp fbo.
 2014-08-27 23:13:08 -07:00
Unknown W. Brackets
9cab344026 Rebind the fbo after inter-buffer copy. 
Might be needed if the list is stalled, a copy is done, list unstalled,
and then something happens to flush a pending prim.
 2014-08-27 23:10:07 -07:00
Unknown W. Brackets
fc5d62804e Avoid unbinding the draw fbo where possible. 
If we're just going to bind another one, no need to trade around.
 2014-08-27 23:08:17 -07:00
Unknown W. Brackets
77e050d044 Only unbind the read fbo id draw can be skipped. 2014-08-27 23:07:11 -07:00
Unknown W. Brackets
a8b78a46bd Forget the bound texture on fbo_create(). 
Just in case, since it overrides the texture.
 2014-08-27 23:05:21 -07:00
Unknown W. Brackets
b4b81744ee Use dst factor w/ dst in reverse sub blending. 2014-08-26 00:39:14 -07:00
Henrik Rydgård
caa34c4db8 Merge pull request #6777 from unknownbrackets/gpu-blend 
Use blending to approximate incr/decr/invert
 2014-08-25 11:38:08 +02:00
Unknown W. Brackets
42aa494aeb d3d: Get some initial GE previews working. 
Wrong colors.
 2014-08-24 21:52:47 -07:00
Unknown W. Brackets
39ee5bb6b7 Support stencil incr/decr even without blending. 2014-08-24 10:24:30 -07:00
Unknown W. Brackets
17ab856a35 Always zero drawn pixels with zero. 2014-08-24 10:24:29 -07:00
Unknown W. Brackets
757f10d47a Oops, equation not func. 2014-08-24 10:24:29 -07:00
Unknown W. Brackets
90746fb5b2 Support INVERT with stencil alpha replacement. 2014-08-24 10:24:28 -07:00
Unknown W. Brackets
0018638458 Support INCR/DECR with stencil value replacement. 2014-08-24 10:24:28 -07:00
Henrik Rydgard
1d7642fa48 Remove most mentions of the "_XBOX" define 2014-08-24 14:21:35 +02:00
Henrik Rydgard
6011255e6f D3D: Don't display buffered rendering upside down. Support screen scaling filter setting 2014-08-24 14:04:55 +02:00
Ced2911
d4adc3abb8 [dx9] don't do alpha test in pixel shader, use state mapping 
hrydgard: Cleaned up a bit, disabled by default

Conflicts:
	GPU/Directx9/TextureCacheDX9.cpp
 2014-08-24 09:52:34 +02:00
Henrik Rydgård
f8a4236d58 Merge pull request #6679 from unknownbrackets/gpu-blend 
Handle doubled alpha blending using premultiplication
 2014-08-22 10:06:24 +02:00
Unknown W. Brackets
1e057a76b5 Convert all CLUT entries ever loaded. 
Some games reuse previously loaded bytes, like World Neverland.  We
displayed bad colors in these cases when we don't convert all the entries.

Technically, this means we're hashing less than used in these cases, but if
we hash more we will get lots of false cache misses.

Fixes #6752.
 2014-08-19 22:31:09 -07:00
Unknown W. Brackets
7f61da2e5a Add a hack for Silent Hill games. 
Either way has different problems, but it sounds like this will work
better for Silent Hill.  Probably still rendering wrong, unfortunately.
 2014-08-18 23:20:49 -07:00
Unknown W. Brackets
a43c3771a3 More correctly blend when hitting the frame cap. 
If we hit too many blits per frame, fall back correctly in the shader.
 2014-08-18 23:20:48 -07:00
Unknown W. Brackets
422db25277 Optimize fixed color blending that isn't blending. 
Gods Eater Burst uses this, for example.
 2014-08-18 23:20:48 -07:00
Unknown W. Brackets
b79e4d22b8 Improve a possible obscure blend mode. 
Seems like this would be likely to get overlap problems.
 2014-08-18 23:20:47 -07:00
Unknown W. Brackets
0904085eb5 Avoid using fbo copies for the common 2a + 1-2a. 
We won't even get it right that often since the copy doesn't deal with
overlap properly.
 2014-08-18 23:20:47 -07:00
Unknown W. Brackets
5d1b1ab547 Dirty the fixed color when premultiplying. 2014-08-18 23:20:46 -07:00
Unknown W. Brackets
a41333c012 Avoid premultiplying easy fixed colors. 2014-08-18 23:20:46 -07:00
Unknown W. Brackets
24a0dd12ef Avoid recalculating replaceBlend so many times. 2014-08-18 23:20:45 -07:00
Unknown W. Brackets
44620c92ed Premultiply for doubled alpha blending if possible. 
This makes the effect more accurate where the values were being clamped
before.  Some obscure methods of blending may be slower.
 2014-08-18 23:20:45 -07:00
Henrik Rydgård
d234ab03c6 Merge pull request #6696 from LunaMoo/master 
Add Buffer Scalling Filter option
 2014-08-19 07:55:45 +02:00
Unknown W. Brackets
dc09875eca Refactor s8/s16 to float conversion. 
It's pretty much all done the same way.  Note that normals have always
been a little inconsistent between 127/128 and 32767/32768.  But it's
probably not causing a major impact...
 2014-08-18 00:47:50 -07:00
Unknown W. Brackets
385df1c54e Force positions to scale by 128/32768 as psp does. 
This makes everything use floats for positions.  On some hardware/drivers,
this may be faster.  On some it may be slower.  We'll need testing to see
the performance impact.

Fixes Final Fantasy 4's pos misalignments, and probably others (like
Tekken 5 I suspect.)
 2014-08-17 14:31:49 -07:00
Unknown W. Brackets
147eb92c95 Fix USE_BONE_ARRAY (currently disabled.) 2014-08-17 10:56:30 -07:00
Luna
62a1023615 Add "Screen Scaling Filter" option 2014-08-08 20:51:15 +02:00
The Dax
ee9eb00656 One last buildfix (please correct if GL_FUNC_ADD is wrong, I just wanted buildbot to work again). https://github.com/hrydgard/ppsspp/pull/6679 should replace this soon. 2014-08-05 00:34:36 -04:00
Unknown W. Brackets
cde9746dab Buildfix. 2014-08-04 21:09:58 -07:00
Unknown W. Brackets
3008b4f8a8 Also do it here, although may be removed soon. 2014-08-04 19:33:58 -07:00
Unknown W. Brackets
8d6a5fd694 Use correct stencil values with non-add blending. 
Otherwise, we get the min/max/etc. value instead.
 2014-08-04 19:16:10 -07:00
Unknown W. Brackets
0279aa154c Use glStencilMask() to partially simulate amask. 
Alpha won't be correct, though.
 2014-08-03 21:48:18 -07:00
Unknown W. Brackets
3bf88a3c9a Disable tests and etc. when destroying framebufs. 
Fixes #6678.  May fix problems loading savestates as well.
 2014-08-03 18:45:36 -07:00
Unknown W. Brackets
9844c6e8cc Create the stencil texture after the temp fbo. 
Otherwise, we'll bind the temp fbo as a texture while creating it.
Somehow my NVIDIA driver was making this work anyway?

Fixes #6611.
 2014-07-22 22:36:38 -07:00
Henrik Rydgard
c8ac26e502 Apparently a 1 pixel margin works around Assassin's Creed buffer overlap misdetection 
(while still keeping BOF working).

Should take care of #6599.
 2014-07-21 20:00:25 +02:00
Unknown W. Brackets
bf01fd2185 Try to avoid false reporting of depthbuf reuse. 2014-07-13 11:59:45 -07:00
Unknown W. Brackets
d80d5beb71 Stop reporting blits when render mode is swapped. 2014-07-09 23:40:29 -07:00
Unknown W. Brackets
f553b632be Avoid a texture offset if framebuf attach fails. 
But, I'm not sure in what case this could happen.
 2014-07-09 08:22:57 -07:00
The Dax
97b467683a Don't crash when framebuffer is a nullptr. 2014-07-09 06:11:41 -04:00
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
81e6f93be4 Check for access to offset framebuffers better. 
Even if the height of the framebuffer doesn't exactly match the offset.
Fixes #2759 again.
 2014-07-06 11:52:00 -07:00
Unknown W. Brackets
be76846679 When applying the Danganronpa hack, force 5551. 
This prevents dithering from happening during the blit, and since
dithering is off during drawing, allows the colors to be correct on some
devices which always dither during blits.
 2014-07-05 00:49:30 -07:00
Unknown W. Brackets
8ab33795b3 Delete temporary/blitting FBOs on device lost. 
Fixes issues when the device goes to sleep with Danganronpa, block
transfers, etc.
 2014-07-04 23:27:12 -07:00
Unknown W. Brackets
4cb8c414b0 Use the CPU for download on non-NVIDIA cards. 
Fixes #6494.
 2014-07-04 18:04:32 -07:00
Unknown W. Brackets
3a2d1ac28e Correct async packing alignment. 
Only matters if a framebuffer width is not a multiple of 4, which might
not happen anyway.
 2014-07-04 14:10:56 -07:00
Unknown W. Brackets
6a4269b1d0 Disable dithering (which affects pixel download.) 
Fixes #6494.
 2014-07-04 14:10:20 -07:00
Unknown W. Brackets
be0340bb34 Fix crash when FBO_ARB is not available. 
Experienced this on a Mac / SDL with a Radeon X1600.
 2014-07-04 01:54:03 -07:00
Unknown W. Brackets
edd760fe39 Oops, forgot a cast. 2014-07-03 01:30:08 -07:00
Henrik Rydgård
ed9866b212 Merge pull request #6490 from unknownbrackets/framebuffer-clut 
Fix depal shaders on GLES3 devices
 2014-07-03 09:53:53 +02:00
Henrik Rydgård
4972f63dfa Merge pull request #6491 from unknownbrackets/gpu-minor 
Try to fix color test masking on GLES3 devices
 2014-07-03 09:52:45 +02:00
Unknown W. Brackets
739c215d35 Stop reporting memcpy fbo upload/download. 2014-07-03 00:48:46 -07:00
Unknown W. Brackets
61ea20875c Fix shader blending on ARM GLES3 devices. 
We currently ban them from GLES3, and this was interacting with that
incorrectly.  Should we still ban them, though?
 2014-07-03 00:15:15 -07:00
Unknown W. Brackets
e364de42cb Ensure #extension is before non-# tokens. 
As required by Tegra devices from reporting.
 2014-07-02 23:21:52 -07:00
Unknown W. Brackets
c379169efc Try to fix colormasking on GLES3 devices. 
Not tested, though, since I don't have one... from reporting.
 2014-07-02 23:15:02 -07:00
Unknown W. Brackets
5ddb848d09 Fix depal shaders on GLES3 devices. 
This path should have only > GL 3.3 or GLES3 so texture() is okay.
 2014-07-02 22:35:47 -07:00
Unknown W. Brackets
602daf5f8a Add some missing algorithm includes. 2014-07-01 00:26:22 -07:00
Unknown W. Brackets
5db79dcf11 Fix some missing semicolons on log statements. 2014-06-29 19:09:37 -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
Unknown W. Brackets
d2d7d6e34b Destroy all FBOs/textures upon sceKernelLoadExec(). 
It can confuse our heuristics about size, attachment, memcpy, etc.
 2014-06-28 12:29:09 -07:00
Unknown W. Brackets
63bcb65d96 Height can be equal, just not greater. 
Oops.
 2014-06-28 03:41:44 -07:00
Unknown W. Brackets
7d0aeb067a Make sure the conv buffer matches the used height. 
May matter if bufferHeight > height.
 2014-06-28 02:40:40 -07:00
Unknown W. Brackets
217d8d24ea Validate height better when downloading fbs. 2014-06-28 02:33:28 -07:00
Henrik Rydgård
9351fe44ea Merge pull request #6444 from unknownbrackets/jit-minor 
Use more correct rounding in armjit, update x86jit and minor
 2014-06-28 11:20:11 +02:00
Unknown W. Brackets
e57ed7d2f4 Small optimizations to stencil scanning. 2014-06-28 02:00:14 -07:00
Unknown W. Brackets
a33589b3e9 Optimize away bits in the stencil upload. 
Should be faster than sending to the GPU this way, if we can help it.
Should improve delays on FBO creation at high render resolutions.
 2014-06-28 01:39:18 -07:00
Unknown W. Brackets
464b453748 Use a temp FBO at 1x to upload stencil. 
This improves performance at > 1x resolutions.
 2014-06-28 01:19:07 -07:00
Unknown W. Brackets
89ffbd027e vertexjit: Specify round to zero in ARM rounding. 
Only affects color morphs.  NEON always rounds to zero so let's be
consistent.
 2014-06-28 00:38:57 -07:00
Unknown W. Brackets
b829be1bfe Require 25% of texture to overlap framebuffer. 
Because it's probably wrong if only a small amount overlaps.
 2014-06-26 00:48:48 -07:00
Henrik Rydgard
d040b980dd Experiment: Disable clear optimization on PVR (see #6290) 2014-06-24 23:32:40 +02:00
Unknown W. Brackets
ca2d9286e0 Notify the texcache when fb format changes. 
Fixes #2094 again.
 2014-06-24 00:14:56 -07:00
Henrik Rydgard
cc269488b2 Protect against bad spline patch data better. 
Also fix a struct name collision.
 2014-06-24 01:14:32 +02:00
Henrik Rydgård
e27e285624 Merge pull request #6396 from hrydgard/ridge-racer-hack 
Ridge Racer hack: Flush after every prim if fb addr == tex addr.
 2014-06-23 17:52:25 +02:00
Henrik Rydgård
14c0c019db Merge pull request #6407 from unknownbrackets/replace-funcs 
Fix function replacement hooks on arm
 2014-06-23 17:50:32 +02:00
Unknown W. Brackets
8a7b440933 Don't need this flag, it read from memory. 2014-06-23 08:21:20 -07:00
Unknown W. Brackets
ad060b992d Clear buffers right after creating them. 
Seems like some drivers have trouble otherwise.
 2014-06-23 00:17:35 -07:00
Unknown W. Brackets
6d104edb5d Make sure not to leak any enabled vertex arrays. 
When returning to EmuScreen, we need to reset any enabled vertex arrays or
we could crash in EmuScreen.  Also, when starting a frame / dirting the
shader, make sure also to toggle the vertex arrays.
 2014-06-22 21:42:29 -07:00
Unknown W. Brackets
b8ad665b53 Clear vertex attrib arrays before output to screen. 
Doesn't seem to help the mis-enabled arrays, though.
 2014-06-22 21:03:33 -07:00
Unknown W. Brackets
4c62a17f72 Allow exact render-to-tex format mismatches more. 
If the right format was used recently, it's still okay.  Also check the
x/y offset to try to find the best match.
 2014-06-22 21:02:43 -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
73d77d3e8a Warning fix. 2014-06-21 18:04:00 -07:00
Unknown W. Brackets
e0b4aa68d0 Typos. 2014-06-21 12:04:27 -07:00
Henrik Rydgård
5f57e7ebaf Merge pull request #6383 from unknownbrackets/gpu-minor 
Check for closest render-to-tex, ignore high bits of fb addr
 2014-06-21 21:03:17 +02:00
Unknown W. Brackets
0418c8071a Support more shift/masks in fb cluts on GLES2. 2014-06-21 11:10:57 -07:00
Unknown W. Brackets
e9224e9eaa Actually set the offset framebuffer. 
Oops.
 2014-06-21 09:04:39 -07:00
Unknown W. Brackets
1cea361c65 Remove now-unnecessary MaskedEqual() compares. 
Clearer this way I think.
 2014-06-21 08:30:45 -07:00
Unknown W. Brackets
4319d5eb28 Take the closest offset framebuffer. 
There could be multiple, or even an exact match too.  Make sure to take
the closest one.
 2014-06-21 08:23:10 -07:00
Unknown W. Brackets
0c31e551e7 Clear buffers before using them. 
Fixes a reported crash on an AMD card.
 2014-06-19 23:09:15 -07:00
Unknown W. Brackets
ef390c5c40 Double check upload/download are VRAM addresses. 2014-06-19 23:09:15 -07:00
Unknown W. Brackets
5f4a0d8174 Fix a few minor warnings. 2014-06-19 00:47:58 -07:00
Unknown W. Brackets
19f35bbac1 Clear stencil to 0, not 1. 
Memory is generally initially zeros, after all.  Fixes #5205 (or at least
improves it?) and fixes #6226.
 2014-06-18 00:09:01 -07:00
Unknown W. Brackets
b5eb072e64 Clear depth buffers to zero, not 1.0f. 
Memory is generally initially zeros.
 2014-06-17 23:56:33 -07:00
Unknown W. Brackets
a9da3618ba Upload the color/stencil buffers on FBO creation. 
Fixes pausing or loading a savestate in Star Ocean (previously your
character became invisible until the screen panned.)
 2014-06-17 23:55:09 -07:00
Unknown W. Brackets
171a865b37 Support reversing normals in software transform. 2014-06-17 23:29:33 -07:00
Unknown W. Brackets
d4c416a437 Use the pre-world normal in software proj mapping. 
Fixes #6081.
 2014-06-17 23:27:19 -07:00
Unknown W. Brackets
2d63a74aeb Allow DrawPixels (uploads) to handle alpha. 
Doesn't update stencil, but at least it updates alpha.

This was fixed to 1.0 before because we had blending enabled by accident
during DrawPixels().
 2014-06-17 23:10:38 -07:00
Unknown W. Brackets
359f72078d Disable vsync when exiting the game. 
Not sure why, but with some drivers not doing this makes the process burn
cpu usage when back in the menu.  Fixes #6351.
 2014-06-17 01:06:25 -07:00
Henrik Rydgård
565bb98db1 Merge pull request #6350 from unknownbrackets/gpu-blend 
Avoid blitting for a bunch of gpu blending modes
 2014-06-17 09:54:57 +02:00
Unknown W. Brackets
6e751f8092 Use shader blend for min/max when unsupported. 2014-06-17 00:20:06 -07:00
Unknown W. Brackets
640c58a892 Reorganize alphablend state mapping logic. 2014-06-17 00:20:05 -07:00
Unknown W. Brackets
17f5acd8b0 Support more src alpha doubling modes. 
These prevent the need for more blits.
 2014-06-17 00:20:05 -07:00
Unknown W. Brackets
143ad92afd Convert blending to doubling in more cases. 
We're not doubling the alpha, so all these cases are safe regardless of
the alpha value.
 2014-06-17 00:20:04 -07:00
Unknown W. Brackets
256436b594 Support fixed/fixed blending without a blit. 2014-06-17 00:20:04 -07:00
Sacha
7bbdabdfa2 glBindFragDataLocation is not needed. Thanks HdkR. 2014-06-17 15:52:46 +10:00
Sacha
9961ef1ba7 Qt: Buildfix USING_GLES2 and not MOBILE_DEVICE 2014-06-17 15:30:03 +10:00
Unknown W. Brackets
d1357e5bc8 Fix another mistake in color test masking. 2014-06-16 02:12:08 -07:00
Unknown W. Brackets
a28354f072 Fix dumb typo. 2014-06-16 01:42:19 -07:00
Unknown W. Brackets
9b67594a1d Linux buildfix. 2014-06-16 01:04:53 -07:00
Henrik Rydgård
f061cb4c0f Merge pull request #6341 from unknownbrackets/gpu-minor 
Support alpha and color test masks on desktop/gles3
 2014-06-16 10:02:20 +02:00
Unknown W. Brackets
982f0416a2 Pre-mask alphatest value as a small optimization. 
Might cause small behavior difference on gles2 (without bitwise ops.)
 2014-06-16 00:35:30 -07:00
Unknown W. Brackets
2038bc5527 Support alpha / color test masks on desktop/gles3. 
Won't work on GLES 2, though...
 2014-06-16 00:33:48 -07:00
Unknown W. Brackets
3fea3e1293 Respect "nearest" filtering setting for fbs. 2014-06-15 19:42:24 -07:00
Unknown W. Brackets
107df3b838 Reuse existing buffer matching for MotoGP hack. 2014-06-15 19:34:57 -07:00
Unknown W. Brackets
a54cefc138 Only download width bytes, not stride bytes. 
Fixes #6330, MotoGP not rendering properly with simulate block transfers.
 2014-06-15 19:33:32 -07:00
Henrik Rydgård
43bddd8067 Merge pull request #6336 from unknownbrackets/texcache 
Separate framebuffer sampling params from tex
 2014-06-16 00:52:20 +02:00
Unknown W. Brackets
e762689ea8 Update vsync on resize, and before first FBO. 
May take care of #4440.
 2014-06-15 15:19:49 -07:00
Unknown W. Brackets
fdd759909b Auto-disable shader blending after 48 per frame. 
Should prevent possible massive slowdowns from log reporting.
 2014-06-15 13:48:06 -07:00
Unknown W. Brackets
40f13d51e7 Separate framebuffer sampling params from tex. 
This way we don't need to reset params on the texture or worry about
things like that.  We always force on the framebuffer anyway, so this is
simpler.

Also don't enable mipmaps when using a framebuffer, since they don't make
sense currently.
 2014-06-15 13:34:58 -07:00
Unknown W. Brackets
f6649794df Respect max texture level in GE debugger preview. 2014-06-15 10:31:16 -07:00
Unknown W. Brackets
eaf3288304 Add debug interface for non-L0 texture preview. 2014-06-15 09:30:37 -07:00
Henrik Rydgård
9503603f5f Merge pull request #6326 from unknownbrackets/gpu-minor 
Support render-to-offset within a framebuffer
 2014-06-15 09:13:03 +02:00
Unknown W. Brackets
030b562eb9 Support render-to-offset within a framebuffer. 
Fixes #6324.
 2014-06-14 17:08:41 -07:00
Henrik Rydgard
ab61255018 Allow only a small number of "InvalidateAll" operations per frame on the texture cache. 
Reduces stutter in FF:Type0 tutorial end scene a bit more.
 2014-06-15 00:40:08 +02:00
Unknown W. Brackets
651d8a4a7d Support inter-buffer memcpy transfer. 
Seeing some games using it like 3rd Birthday uses block transfers, to copy
a frame to the other buffer.
 2014-06-14 13:22:00 -07:00
Unknown W. Brackets
d06a0520ae Consistently mark vfb dirty after drawing to it. 2014-06-14 13:19:27 -07:00
Henrik Rydgård
59bef47035 Merge pull request #6321 from unknownbrackets/texscaling 
Reduce some texture scaling limits to rely on texels instead
 2014-06-14 20:08:39 +02:00
Unknown W. Brackets
6ebdf2be29 Skip memcpy fb download when already downloaded. 
Fixes Final Fantasy Tactics lag, discussion in #6308.
 2014-06-14 11:04:17 -07:00
Unknown W. Brackets
b1396d51b2 Log more info about skipped subarea renders. 2014-06-14 11:04:17 -07:00
Unknown W. Brackets
ec7942fc8a Report decoding a texture from a VRAM mirror. 
Might mainly be depth textures?
 2014-06-14 11:04:16 -07:00
Unknown W. Brackets
0b9db176e9 Centralize render-to-tex logic, support mirrors. 
Now everything is decided in AttachFramebuffer().
 2014-06-14 11:04:16 -07:00
Unknown W. Brackets
822f936b05 Validate stride when using clut, more reporting. 2014-06-14 11:04:15 -07:00
Unknown W. Brackets
d8eaeab663 Reduce change frequency limit for texscaling. 
Since we have the texels limit now, this should allow more textures to be
scaled safely.  Also, it may detect alpha optimizations in more textures,
since we currently skip them for frequently changed textures.
 2014-06-14 10:49:13 -07:00
Unknown W. Brackets
0620c6edc7 Forgive freq. changed tex if they mend their ways. 2014-06-14 10:43:22 -07:00
Unknown W. Brackets
99c8319f15 Don't count unscaled frequently changing texels. 
May cause more things to be scaled per frame (especially if first textures
on a frame are frequently changing.)
 2014-06-14 10:40:23 -07:00
Henrik Rydgård
5f6c083a3e Merge pull request #6315 from unknownbrackets/framebuf-estimate 
Avoid shrinking buffer w/h in throughmode
 2014-06-14 11:06:37 +02:00
Unknown W. Brackets
119edd34e6 Avoid shrinking buffer w/h in throughmode. 
Since they could overdraw and we already scissor.
 2014-06-14 01:56:34 -07:00
Henrik Rydgård
036cde7768 Merge pull request #6313 from unknownbrackets/gpu-minor 
Implement in-shader blending on gles2 / gl2
 2014-06-14 09:21:30 +02:00
Henrik Rydgård
5fa331bb56 Merge pull request #6310 from hrydgard/deferred-scaling 
Limit the amount of texture scaling done per frame to reduce stutters
 2014-06-14 08:59:40 +02:00
Unknown W. Brackets
444128dcdd Support NEVER alpha/color tests. 2014-06-13 23:44:40 -07:00
Unknown W. Brackets
1d9f7b04fc Implement in-shader blending on gles2 / gl2. 
I'm sure it'll be slow, though...
 2014-06-13 23:42:59 -07:00
Unknown W. Brackets
673e92927a Avoid an infinite loop on small heights. 2014-06-13 23:37:08 -07:00
Henrik Rydgard
4e81eadc88 Minor cleanup in texcache 2014-06-14 00:18:37 +02:00
Henrik Rydgård
2f86e3e021 Clear STATUS_TO_SCALE at a better place 2014-06-13 23:01:40 +02:00
Henrik Rydgård
ea9f8d8bf3 Comment out the deferred scaling logs. 2014-06-13 22:11:25 +02:00
Henrik Rydgård
c0c058937f Limit the amount of texture scaling done per frame, to reduce stutters 
Scaling causes really bad stutters in some games on some hardware.

This should also reduce the impact of pathological cases like the first
boss of FF Type-0 to be more bearable.
 2014-06-13 22:07:26 +02:00
Unknown W. Brackets
f84650b6be Report possible textures at subareas > 32. 
For the ones outside the hack 0x04110000 range.
 2014-06-13 09:19:51 -07:00
Unknown W. Brackets
ac6cbf3524 Take the nearest fbo for a y-offset memcpy(). 2014-06-13 08:45:26 -07:00
Unknown W. Brackets
392897d447 Allow oversized gpu memcpys. 
Seems to happen in Mana Khemia, where it uses h=384, but it looks
legitimate.
 2014-06-13 08:42:58 -07:00
Unknown W. Brackets
00651ceea1 Support gpu memcpy/memset line-by-line. 
Fixes some glitches in Final Fantasy Tactics, which uses memset line by
line (at width, not stride.)

Should still be relatively safe, unless a buffer is detected as too tall.
The memcpy() will still happen anyway.
 2014-06-13 08:36:02 -07:00
Unknown W. Brackets
dad1bf41f7 Avoid making FBO copies in the debugger. 
When it binds the texture preview.  Not needed.
 2014-06-13 08:35:12 -07:00
Henrik Rydgård
3da962f585 Merge pull request #6300 from unknownbrackets/gpu-blocktransfer 
Try to handle more block transfer / tex render cases
 2014-06-12 20:34:26 +02:00
Unknown W. Brackets
025b92593a Just don't use render-to-text format mismatches. 
This already excludes CLUTs.
 2014-06-12 07:46:56 -07:00
Unknown W. Brackets
1a25456e52 Detach a format mismatch that's not rendering. 
Assume we'd be better off reading ram.
 2014-06-12 00:48:21 -07:00
Unknown W. Brackets
89206a8a09 Allow render-to-tex with > 16 y offset. 
Trying to keep it cautious in case our sizes are wrong.
 2014-06-12 00:41:42 -07:00
Henrik Rydgård
89bbc90678 Merge pull request #6299 from unknownbrackets/gpu-blocktransfer 
Track framebuffer memory dirty more carefully
 2014-06-12 09:01:38 +02:00
Henrik Rydgård
65153b55fa Never detach shaders as some drivers may have broken semantics. May help #6294 
(you're supposed to be able to detach them after link)
 2014-06-12 08:58:14 +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
b9f805c71f Fix graphical artifact in Jeanne d'Arc world map. 
We need to use the last set render framebuffer, not the current one at
copy to output.
 2014-06-11 00:54:54 -07:00
Unknown W. Brackets
efc2de41ae Ignore depth buffers with stride = 0. 
They're clearly not being used.
 2014-06-11 00:40:47 -07:00
Unknown W. Brackets
c49ba888b5 Flush draws before doing any block transfers. 
Otherwise, we might flush on top, or we might download the wrong thing.
 2014-06-11 00:28:28 -07:00
Unknown W. Brackets
dd4b191003 Correct Gods Eater Burst tex download size. 
Probably takes care of #6289, although it doesn't crash for me.
 2014-06-10 22:59:08 -07:00
Henrik Rydgard
3e97713d16 Revert "Remove option" (disable alpha test) 
Fixes issue #6286.

This reverts commit d2ce635bbf.
 2014-06-10 23:13:42 +02:00
Henrik Rydgard
34b5a1c2f3 Disable precision qualifiers on desktop GL 2014-06-10 22:44:53 +02:00
Unknown W. Brackets
adee76cd62 Add a matching PerformMemoryUpload(). 
Just so we have both Download and Upload to use later.
 2014-06-10 00:45:03 -07:00
Unknown W. Brackets
5bd1d9d009 Avoid corrupting mem if our framebuf size is wrong. 
Let the size specify the framebuffer height, which we are most likely to
have gotten wrong.
 2014-06-10 00:15:05 -07:00
Unknown W. Brackets
d5aae5a705 Match block transfer address even when uncached. 
Rhythm8 for example uses an uncached address for the dest framebuffer.
 2014-06-09 23:28:54 -07:00
Unknown W. Brackets
ef12694c4b Don't update render tex x/y offset unless matched. 
Sometimes we have false matches, they should be skipped.  Fixes some bloom
that is off in some games.
 2014-06-09 22:45:14 -07:00
Sacha
9d9d135176 Symbian buildfix, x11 linker order 2014-06-10 15:33:30 +10:00
Sacha
87c7b31e77 Qt: Mac arch detection workaround on Qt4. Fix a conflict. 2014-06-10 15:11:25 +10:00
Henrik Rydgard
e01f8232bd Try the quickfix proposed for 6272 at 1080p, works nicely. 
Might want to do something more ambitions, too, allowing minor crops
on more resolutions...
 2014-06-09 21:22:12 +02:00
Henrik Rydgard
339c4d7dc3 Fix a bunch of typos breaking various utility shaders on GLES 
Fixes white shadows in FF-Type 0 on Tegra, as mentioned in #6268
 2014-06-09 19:15:45 +02:00
Sacha
e57ae29e4a Symbian: Switch to GCC 4.7.3. Blackberry: Switch to GCC 4.8.2 2014-06-09 23:49:30 +10:00
Henrik Rydgård
8bef7c64b1 Merge pull request #6262 from unknownbrackets/framebuf-estimate 
Assume 512 tall framebuffers are wrong too
 2014-06-09 13:02:57 +02:00