9751 Commits
Author SHA1 Message Date
Unknown W. Brackets 4dafb163a3 D3D11: Split color test out by component.
Not working correctly together.  See #10629.
2018-09-08 20:03:26 -07:00
Unknown W. Brackets 97773d3dd5 TexCache: Fix texture alignment in GLES.
We must align to 4 bytes, and we aren't always aligned to 16 anymore, so
we must check when dealing with swizzle.
2018-09-08 19:00:30 -07:00
Unknown W. Brackets 2388be544f Debugger: Fix crash in softgpu with no texaddr yet. 2018-09-04 21:54:25 -07:00
Unknown W. Brackets f65edc20a3 TexCache: Optimize DXT5 alpha lerp.
This makes the overall DXT5 decode about 8% faster.
2018-09-02 11:41:27 -07:00
Unknown W. Brackets 3f35221f3b TexCache: Avoid masking out alpha for DXT3/DXT5.
A little faster.  Also refactor colors a bit to be more readable.
2018-09-02 09:53:31 -07:00
Unknown W. Brackets 8ae2b1e6fb TexCache: Optimize DXT3/DXT5 decode to single pass.
This is significantly faster on Vulkan, and in other situations where
we're decoding directly to uncached memory.
2018-09-02 09:30:46 -07:00
Unknown W. Brackets 653129c528 Debugger: Add an event for GE dump recording. 2018-09-01 10:59:37 -07:00
Unknown W. Brackets c10b2035b5 Debugger: Prevent record with no commands.
Should cut down on empty dumps, at least.
2018-09-01 09:51:04 -07:00
Unknown W. Brackets f1afc51994 GE Debugger: Refactor out stepping callback.
In case we have multiple connected debuggers, we don't want just a
callback.  A counter works fine.
2018-09-01 09:41:20 -07:00
Unknown W. Brackets a4c0640f01 GE Debugger: Factor out host calls some.
Moving more of this to cross platform for the web debugger.
2018-09-01 08:32:03 -07:00
Henrik Rydgård 2f060a4c74 Merge pull request #11337 from unknownbrackets/kh-shadow
GPU: Clear alpha more consistently from 565
2018-09-01 09:48:29 +02:00
Unknown W. Brackets 16d7a80980 GPU: Clear alpha more consistently from 565.
Before, the backends all did different things.  Now they are more in sync,
but Vulkan still behaves slightly differently.

Fixes #11326.
2018-08-30 21:00:21 -07:00
Unknown W. Brackets 00daa00ae5 TexCache: Keep maxSeenV on clut variants in sync.
This ensures that we detect changes properly even when they are outside
the max V used with one CLUT.

Fixes #9355, enemy fade out in FF2.
2018-08-29 22:07:27 -07:00
Unknown W. Brackets 22a536fb0c Merge pull request #11307 from hrydgard/discord-mac-linux
Enable Discord integration for Mac and Linux.
2018-08-26 23:04:49 -04:00
Unknown W. Brackets 274b1eaa0e GPU: Fix typo with wrong type.
No behavior change, just fixes a warning.
2018-08-26 11:51:06 -07:00
Henrik Rydgård 3c99d16da7 Merge pull request #11322 from unknownbrackets/texcache
GPU: Fix texture handling on framebuf detach
2018-08-25 21:12:45 +02:00
Henrik Rydgård 569c073497 Merge pull request #11319 from unknownbrackets/logicop
GPU: Apply logic op in more cases
2018-08-25 21:11:48 +02:00
Henrik Rydgård 2aaa013296 Merge pull request #11323 from unknownbrackets/fb-safesize
GPU: Download safe size on next create, too
2018-08-25 21:09:56 +02:00
Unknown W. Brackets 6ee0612a10 Ge: Keep jump/call optim when fast memory on.
This gives a way to validate with fast memory off, and get less crashes.
2018-08-25 10:38:56 -07:00
Unknown W. Brackets d13f76308d Ge: Avoid executing invalid pointers.
May help #3407 and related errors, as well as crashes and execution of
garbage.  This stops immediately on a bad pointer, rather than trying to
push through.
2018-08-25 10:32:32 -07:00
Unknown W. Brackets 7fa20d4734 GPU: Download safe size on next create, too.
In some cases, games will create a series of framebufs.  Before, we
weren't downloading them if they were only used once, as intended.
2018-08-25 10:01:43 -07:00
Unknown W. Brackets 3f48e446b6 GPU: Fix texture handling on framebuf detach.
We were never creating/recreating the texture, so we ended up with null.
Caused all sorts of problems.
2018-08-25 09:57:37 -07:00
Unknown W. Brackets 5c01bf61e6 GPU: Apply logic op in more cases.
It has well defined behavior with alpha blending, so let's apply it even
when we're blending.  Fixes #11316.
2018-08-24 21:19:54 -07:00
Unknown W. Brackets 734db3d050 GPU: Refactor logic op simulation for better reuse. 2018-08-24 21:05:27 -07:00
Unknown W. Brackets 44b90f638d SoftGPU: Apply pixel mask for clears.
Turns out it's respected even in clear mode.

Also, confirm logic op and blending interact.
2018-08-24 20:55:49 -07:00
weihuoya 7383e90c19 handle cullface, help to #10597 2018-08-23 12:46:09 +08:00
Unknown W. Brackets cd98c4a266 GLES: Fix direct shader depal on GLES.
Was missing a cast and only worked on some desktop before.
2018-08-18 10:41:59 -07:00
Henrik Rydgård e34378a1c1 Enable Discord integration for Mac and Linux. 2018-08-18 10:11:15 +02:00
Henrik Rydgård 6f173b9134 Harmonize some VS project settings. 2018-08-14 22:41:39 +02:00
Henrik Rydgård 39eb2b4153 Merge pull request #11298 from unknownbrackets/gles-depal
GLES: Enable shader depal with Vulkan fixes
2018-08-13 07:50:13 +02:00
Unknown W. Brackets 70318fda37 GLES: Enable shader depal with Vulkan fixes. 2018-08-12 19:49:08 -07:00
Unknown W. Brackets c32ffa1db0 GE Debugger: Rename depth clamp more places.
Was still showing clip enable in disasm.
2018-08-11 16:33:39 -07:00
Unknown W. Brackets 06ad9b829f SoftGPU: Oops, correct rounding to make sense.
Was too hasty in that last commit.
2018-08-09 07:13:42 -07:00
Unknown W. Brackets 31d5c39858 SoftGPU: Fix some minor rounding on viewport cull.
Had some tests failing when on the edge due to this.
2018-08-05 20:07:45 -07:00
Unknown W. Brackets e22cc7ef6d SoftGPU: Always clip, without special neg z case.
Depth clamping doesn't change whether it clips.  Also, avoid culling when
a vertex is behind the near plane.
2018-08-05 17:47:05 -07:00
Unknown W. Brackets cd6b1f73c1 SoftGPU: Drop verts only when depth not clamped.
Depth clamping bypasses the 4096x4096 box check.
2018-08-05 17:17:55 -07:00
Unknown W. Brackets 77f0499f7f GPU: Rename clipping flag to depth clamp.
It seems to just to depth clamp.  When depth clamp happens, it affects
clipping a little, but only for vertices that needed clamping.
2018-08-05 17:11:51 -07:00
Unknown W. Brackets 7885a88c0c GPU: Optimize > 0 alpha test using discard rules.
It should be equally unsafe to use src color as dest blend factor, or
use blending with a subtract or min/max equation.
2018-07-29 11:39:06 -07:00
Unknown W. Brackets 78dc07e7a3 Headless: Fix graphics tests for Vulkan/Direct3D.
Otherwise we hang because there's no thread to ever end.
2018-07-29 11:26:04 -07:00
Unknown W. Brackets fba0de59c1 GPU: Restrict alpha test to zero for dest blend.
We could end up with the wrong blending in other cases, because the
exiting color will get multiplied.

Luckily, this is still the common case.
2018-07-29 11:26:04 -07:00
Henrik Rydgård a776dce593 Merge pull request #11269 from unknownbrackets/fragtest
Avoid discard when we can blend
2018-07-29 09:49:16 +02:00
Henrik Rydgård 42f4d7b40f OpenGL: Fix bug where we could end up calling glUniformMatrix without a bound program. Found by GL debug callback on NV.
This adds a bit of extra checking that's only enabled in _DEBUG builds.
2018-07-28 11:09:01 +02:00
Unknown W. Brackets 5ccd3ee4dc GPU: Allow skipping discard when write disabled.
Even if we depth test, if we don't write, we can still force alpha to
zero.
2018-07-27 20:16:51 -07:00
Unknown W. Brackets 8e49fa988c GLES: Avoid discard when we can blend.
May improve performance on PowerVR.
2018-07-27 20:06:41 -07:00
Unknown W. Brackets 5ad948df73 Vulkan: Avoid discard when we can blend.
Where possible, replace alpha and color testing with a zero alpha value.

This allows early fragment tests more often, which may help #11227.  It
may also generally help performance on PowerVR devices.
2018-07-27 20:04:36 -07:00
Henrik Rydgård a589a371e6 Update SPIRV-Cross to a fresher version. 2018-07-16 20:00:45 +02:00
Unknown W. Brackets 82f1b72c4a Merge pull request #11254 from hrydgard/light-chan-cost
Reduce the estimated transform cost of an additional light from 10 to 7 cycles
2018-07-15 07:42:34 -07:00
Henrik Rydgård c404214042 Update glslang to latest as of 2016-06-26 2018-07-15 12:30:40 +02:00
Henrik Rydgård ea1ef9277c Reduce the estimated transform cost of an additional light from 10 to 7 cycles, see #5530. 2018-07-15 12:13:25 +02:00
weihuoya d9d9b14e53 Don't cut below 2 2018-07-05 15:29:40 +08:00