Unknown W. Brackets
6b8f7294b3
Global: Remove ARM64 define.
...
Was barely used anymore, and some users are specifying it when it's not
the case, so better to detect consistently.
2020-08-29 08:45:50 -07:00
Unknown W. Brackets
2be8f86286
GPU: Avoid compile error if ARM64=1 specified.
...
If it's specified, but incorrect, we would compile incorrectly. We're
moving away from it anyway.
2020-08-29 08:18:54 -07:00
Unknown W. Brackets
7426e2a8f8
TexCache: Try to avoid wrong texcache usage.
...
Now that the speed is closer, disable on non-ARMV7 so we can avoid wrong
compilations more often.
2020-08-29 08:16:23 -07:00
Unknown W. Brackets
364aabe84f
TexCache: Improve quicktexhash by ~1-4% on NEON.
...
Improvement is greater on 64-bit (4.3%) than 32-bit (1.6%.)
2020-08-29 08:03:58 -07:00
Unknown W. Brackets
5945642982
Merge pull request #13351 from hrydgard/virtual-framebuffer-format
...
Virtual framebuffer copies: Make a more informed guess on the pixel format of the copy.
2020-08-29 10:06:01 -04:00
Henrik Rydgård
81159770d6
Merge pull request #13350 from shenweip/memInit
...
Actually clear memory when initializating.
2020-08-29 15:57:05 +02:00
shenweip
0f45e17f09
Only clear when initializating kernelMemory and userMemory.
2020-08-29 19:51:11 +08:00
Henrik Rydgård
8618cc7ab5
Virtual framebuffer copies: Make a more informed guess on appropriate pixel format.
...
This is the first part of the Test Drive fix, which I want in
separately.
2020-08-29 12:42:36 +02:00
shenweip
df82673116
Actually clear memory when initializating.
2020-08-29 15:26:44 +08:00
Henrik Rydgård
cb3ed8f4a1
Merge pull request #13343 from unknownbrackets/xxhash
...
Switch to XXH3
2020-08-28 12:20:26 +02:00
Henrik Rydgård
41fe6f9e12
Delete an unused buffer, saving 0.5MB of RAM.
2020-08-28 10:38:00 +02:00
Unknown W. Brackets
e32524c544
GPU: Remove ReliableHash aliases.
...
Let's just use XXH32/XXH64 directly in texture replacements, clearer
anyway.
2020-08-28 01:15:22 -07:00
Unknown W. Brackets
7f812a5d9a
GPU: Hash CLUT using XXH3 with replacements off.
...
It's faster, but if texture replacements are enabled, we want the hash to
stay the same.
2020-08-27 20:46:39 -07:00
Unknown W. Brackets
53d9c10b22
irjit: Switch to XXH3.
2020-08-27 20:40:55 -07:00
Unknown W. Brackets
15c25be42b
GPU: Switch vertex/lookup hashes to XXH3.
...
This is faster pretty much everywhere.
2020-08-27 20:37:49 -07:00
Unknown W. Brackets
30625225b0
GPU: Remove neon xxhash implementation.
...
It's typically around the same speed now with modern compilers, and much
slower than XXH3.
2020-08-27 20:31:09 -07:00
Unknown W. Brackets
184d4a1fc0
Vulkan: Type comparison warning fix.
2020-08-27 20:01:59 -07:00
Unknown W. Brackets
3b0f7e109d
unittest: Fix build on Android.
2020-08-27 20:01:37 -07:00
Unknown W. Brackets
9b27cbd450
Common: Update to latest xxhash.
2020-08-27 20:01:04 -07:00
Unknown W. Brackets
1371ef298b
Net: Reduce PortManager log noise.
...
Shows up when running unit tests, etc.
2020-08-27 19:23:23 -07:00
Henrik Rydgård
8f4fb89d17
Merge pull request #13342 from hrydgard/vulkan-barrier-fix
...
Vulkan: A couple more barrier/layout-related fixes.
2020-08-28 00:27:49 +02:00
Henrik Rydgård
924329321c
Throw in a minor simplification in depal.
2020-08-27 23:02:17 +02:00
Henrik Rydgård
0b8e377872
Address feedback from #13341 .
2020-08-27 22:59:27 +02:00
Henrik Rydgård
f64d36664e
Fix final subpass dependency for depth
2020-08-27 22:59:18 +02:00
Henrik Rydgård
b58f5e46ae
Correctly update image layout to the final one after a render pass.
2020-08-27 22:59:18 +02:00
Henrik Rydgård
611161b872
Merge pull request #13341 from hrydgard/vulkan-improved-logging
...
Vulkan: improved LogSteps logging
2020-08-27 22:41:32 +02:00
Henrik Rydgård
3607f0ab32
Log validation errors normally as well as using OutputDebugString.
2020-08-27 21:50:05 +02:00
Henrik Rydgård
95a4cd37fc
Improve LogSteps further. Break out PreprocessSteps.
2020-08-27 21:50:05 +02:00
Henrik Rydgård
4e841ca3a6
Naming fix, better names for temp fbos
2020-08-27 21:50:05 +02:00
Henrik Rydgård
6e9d5ffbd7
VulkanRenderManager: Improve logging. LogSteps output is much better.
...
More logging improvements
2020-08-27 21:50:05 +02:00
Henrik Rydgård
e06ade10ab
Merge pull request #13340 from hrydgard/vulkan-separate-depth-view
...
Fix longstanding depth/stencil view issue in Vulkan.
2020-08-27 21:49:34 +02:00
Henrik Rydgård
d54e0b3231
Fix longstanding depth/stencil view issue in Vulkan. Cleans up validation.
...
We didn't specify DEPTH|STENCIL as aspects for views we rendered to,
only DEPTH. Who knows how many of the driver bugs we "found" are this.
DEPTH|STENCIL views can't be sampled though, so we create a separate
DEPTH view for that. This keeps Katamari working.
2020-08-27 21:23:41 +02:00
Henrik Rydgård
b323397c54
Vulkan: Minor fix for depth/stencil pretransitions. Used DEPTH_ATTACHMENT_OPTIMAL by mistake in barrier.
2020-08-27 16:46:14 +02:00
Henrik Rydgård
de4778358e
AspectMask fix
2020-08-27 16:39:15 +02:00
Henrik Rydgård
3be729bfc5
Vulkan: Silly typo fix in barrier
2020-08-27 14:46:02 +02:00
Henrik Rydgård
6ef5c3e375
Merge pull request #13337 from unknownbrackets/sigaltstack
...
Exception: Don't trust sigaltstack()
2020-08-27 08:13:18 +02:00
Unknown W. Brackets
90344fdee5
Exception: Don't trust sigaltstack().
...
Let's just track our allocation separately and free it, always.
2020-08-26 21:44:38 -07:00
Henrik Rydgård
b652f62d19
SavedataParam: Be careful with const char* to string. Might help #13187
2020-08-26 22:17:42 +02:00
Henrik Rydgård
3574a352df
Update lang submodule
2020-08-25 12:19:01 +02:00
Henrik Rydgård
13552550bc
Merge pull request #13330 from unknownbrackets/gpu-clear
...
TexCache: Force recreate texture on detach
2020-08-24 07:24:04 +02:00
Henrik Rydgård
f7283ee993
Merge pull request #13331 from unknownbrackets/debugger
...
http: Avoid logging errors for would block
2020-08-24 07:18:07 +02:00
Unknown W. Brackets
c98d96c78f
http: Avoid logging errors for would block.
...
It's not an error, this case is expected.
2020-08-23 21:32:06 -07:00
Unknown W. Brackets
f44717c4f5
TexCache: Force recreate texture on detach.
...
We could have multiple detaches, and we delete the texture on attach, so
we really must make sure we recreate.
Fixes #13320 .
2020-08-23 21:29:53 -07:00
Henrik Rydgård
a8059d53f9
Merge pull request #13322 from hrydgard/refactor-framebuffer-attach
...
Framebuffer manager refactor step 1: Split ApplyFramebuffer into Match and Apply steps.
2020-08-24 00:29:46 +02:00
Henrik Rydgård
055588c1a1
Framebuffer manager refactor step 1: Split ApplyFramebuffer into Match and Apply steps.
...
This should have no change in behavior, but prepares the ground for the
next steps. Want this merged separately.
2020-08-24 00:01:02 +02:00
Henrik Rydgård
edb1d5e612
Merge pull request #13329 from hrydgard/texture-depth-color-matching
...
Framebuffer matching: Don't try to match depth buffers with color if the swizzle bits don't indicate it
2020-08-23 22:51:56 +02:00
Henrik Rydgård
35c5308c30
Framebuffer matching: Don't try to match depth buffers with color if the swizzle bits don't indicate it.
...
Fixes God of War issue reported in a comment of #13317 .
2020-08-23 22:11:18 +02:00
Henrik Rydgård
6d117fb533
Merge pull request #13327 from unknownbrackets/gpu-clear
...
Fix clear detection mistake
2020-08-23 19:30:30 +02:00
Unknown W. Brackets
3b11b09621
Debugger: Check for no texture before debug bind.
...
This means a framebuffer texture that is somehow not attached, but better
to avoid the crash.
2020-08-23 08:24:46 -07:00
Unknown W. Brackets
5ade93a091
GPU: Properly detect clears.
...
The check was reversed before, oops. Detected masked draws.
Must've gotten this backwards debugging Mana Khemia.
2020-08-23 08:23:39 -07:00