Commit Graph

631 Commits

Author SHA1 Message Date
Ben
007fb63e47
Merge pull request #5230 from B3n30/circumvent_apple_intel_hw_shader_issue
Add flag to disable seperable shaders for osx Intel GPUs.
2020-05-09 12:34:21 +02:00
Ben
96ebd53466
Merge pull request #5264 from lioncash/gen
gl_shader_gen: Make use of fmt where applicable
2020-05-09 12:28:57 +02:00
Ben
e959d44f00
Merge pull request #5262 from lioncash/fmt
gl_shader_decompiler: Make use of fmt with the decompiler
2020-05-09 12:28:34 +02:00
Lioncash
cd201cd60f gl_shader_gen: Make use of fmt where applicable
Avoids string churn and makes it significantly easier to read the shader
template strings that are being used as the basis for formatting.
2020-05-08 14:50:48 -04:00
Lioncash
016d43df98 gl_shader_decompiler: Make use of fmt with the decompiler
Allows us to avoid even more string churn by allowing the AddLine
function to make use of fmt formatting so the string is formatted all at
once instead of concatenating multiple strings.

This is similar to how yuzu's decompiler works, which I've made function
the same way in the past.
2020-05-04 22:40:33 -04:00
Pengfei Zhu
e212dc3a1f
Merge pull request #5297 from lioncash/missing
gl_rasterizer: Amend missing return value in branch in SetupGeometryShader()
2020-05-02 08:07:49 +08:00
Pengfei Zhu
4d7487bd34
Merge pull request #5298 from lioncash/unused
gl_rasterizer: Silence various compilation warnings
2020-05-01 23:34:36 +08:00
Pengfei Zhu
6a0636d631
Merge pull request #5296 from lioncash/flatten
gl_rasterizer_cache: Flatten LoadCustomTexture()
2020-05-01 23:01:47 +08:00
Pengfei Zhu
2eacb11c53
Merge pull request #5292 from lioncash/alloc
gl_shader_manager: Eliminate variable shadowing
2020-05-01 22:57:55 +08:00
Pengfei Zhu
1bb9a71a01
Merge pull request #5291 from lioncash/trunc
renderer_opengl: Resolve compiler truncation warnings
2020-05-01 22:56:46 +08:00
Ben
fd1b254635
Merge pull request #5290 from lioncash/shader
gl_shader_gen: Minor interface cleanup
2020-05-01 16:16:25 +02:00
Lioncash
763778e6c0 gl_rasterizer: Make const on references consistent
Applies const to auto& instances that would actually resolve to a const
reference for readability and explicitness.
2020-05-01 06:19:20 -04:00
Lioncash
437453d32b gl_rasterizer: Amend missing return value in branch in SetupGeometryShader()
Previously undefined behavior was being invoked in the case that
geometry shaders weren't supported.
2020-05-01 06:10:32 -04:00
Lioncash
81768bf0c2 gl_rasterizer: Resolve truncation warnings 2020-05-01 06:08:20 -04:00
Lioncash
df97021f55 gl_rasterizer: Make use of std::string_view in IsVendorAmd()
Same behavior, no heap allocation.

strings returned from glGetString() are guaranteed to be static strings,
so this is safe to do. They're also guaranteed to be null-terminated.
2020-05-01 05:57:25 -04:00
Lioncash
de140a5870 gl_rasterizer: Default destructor in the cpp file 2020-05-01 05:53:40 -04:00
Lioncash
ea08178dde gl_rasterizer: Remove unused emu_window variable in RasterizerOpenGL
Silences an unused private variable warning.
2020-05-01 05:52:48 -04:00
Lioncash
59a614a70f gl_rasterizer_cache: Remove unnecessary reference parameter in LoadCustomTexture()
This is only ever used in a self-referential manner, so we can make use
of the texture info member directly.
2020-05-01 05:42:35 -04:00
Lioncash
e342f36e02 gl_rasterizer_cache: Flatten LoadCustomTexture()
Makes the control flow much nicer to follow, as we don't store to a
local before returning anymore.
2020-05-01 05:33:47 -04:00
Mat M
85d37c9994
gl_shader_disk: Make use of std::nullopt where applicable (#5293)
Some implementations can use the std::nullopt_t constructor of
std::optional to avoid needing to completely zero out the internal
buffer of the optional and instead only set the validity byte within it.

e.g. Consider the following function:

std::optional<std::vector<ShaderDiskCacheRaw>> fn() {
    return {};
}

With libc++ this will result in the following code generation on x86-64:

Fn():
  mov     rax, rdi
  vxorps  xmm0, xmm0, xmm0
  vmovups ymmword ptr [rdi], ymm0
  vzeroupper
  ret

With libstdc++, we also get the similar equivalent:

Fn():
  vpxor   xmm0, xmm0, xmm0
  mov     rax, rdi
  vmovdqu XMMWORD PTR [rdi], xmm0
  vmovdqu XMMWORD PTR [rdi+16], xmm0
  ret

If we change this function to return std::nullopt instead, then this
simplifies both the code gen from libc++ and libstdc++ down to:

Fn():
  mov     BYTE PTR [rdi+24], 0
  mov     rax, rdi
  ret

Given how little of a change is necessary to result in better code
generation, this is essentially a "free" very minor optimization.
2020-04-30 22:42:32 -05:00
Lioncash
f20120e5a7 gl_shader_gen: Mark hash implementations as noexcept
These shouldn't throw at all, so we can mark the interface as such.
2020-04-30 22:57:36 -04:00
Lioncash
ca4f7266c0 gl_shader_manager: Remove unused variable within LoadDiskCache() 2020-04-30 22:48:57 -04:00
Lioncash
138539c9a0 gl_shader_manager: Eliminate variable shadowing
Resolves a few -Wshadow warnings.
2020-04-30 22:46:58 -04:00
Lioncash
2d83b39532 gl_shader_manager: Avoid unnecessary std::vector copy in LoadDiskCache()
Same behavior, but without an unnecessary reallocation.
2020-04-30 22:43:03 -04:00
Lioncash
e118cb1e0b renderer_opengl: Resolve compiler truncation warnings
Resolves numerous truncation warnings on MSVC.
2020-04-30 22:26:14 -04:00
Lioncash
d23ffe9b42 gl_shader_gen: Mark PicaVSConfig reference parameter as const
This isn't mutated whatsoever, so this can be marked const.
2020-04-30 22:07:25 -04:00
Marshall Mohror
8a0b0c2fc7
texture_filters: update ScaleForce (#5270)
* texture_filters: update ScaleForce

* texture_filters: optimize scale_force

* texture_filters/scale_force: optimize final offset calculation
2020-04-27 23:50:47 -05:00
Mat M
bc14f485c4
gl_shader_disk_cache: std::move entries in LoadTransferable() (#5249) 2020-04-24 17:49:54 +02:00
Mat M
8b43dff849
gl_shader_gen: Convert file-scope std::string into a std::string_view (#5263)
Same behavior, no heap allocations at program start up
2020-04-21 20:31:58 -05:00
Marshall Mohror
db5b8b9c88
video_core: reduce string allocations in shader decompiler (#5261)
* video_core: reduce string allocations in shader decompiler

* use append for indentation instead of resize

Co-authored-by: Mat M. <mathew1800@gmail.com>
2020-04-20 22:08:58 -05:00
Sebastian Valle
1469ad339c
Merge pull request #5244 from lioncash/move3
gl_shader_manager: std::move std::string where applicable
2020-04-19 21:48:15 -05:00
Sebastian Valle
acf18ce4dd
Merge pull request #5245 from lioncash/guard
video_core: Add missing header guards
2020-04-19 21:47:45 -05:00
Sebastian Valle
df92b37bbf
Merge pull request #5248 from lioncash/view
post_processing_opengl: Make use of std::string_view with GetPostProcessingShaderCode()
2020-04-19 21:46:48 -05:00
Ben
ca722f1bb1
Merge pull request #5242 from lioncash/noexcept
gl_resource_manager: Make use of noexcept on move assignment and move constructors
2020-04-19 16:56:24 +02:00
Lioncash
623cc926ed post_processing_opengl: Make use of std::string_view with GetPostProcessingShaderCode()
Same behavior, but doesn't result in an allocating copy of the passed in
string. Particularly given the string is only compared against other
existing strings.
2020-04-18 22:27:44 -04:00
Lioncash
7a3d4d7642 video_core: Add missing header guards
Ensures errors related to multiple inclusions never happen.
2020-04-18 22:06:37 -04:00
Lioncash
c72d9b806b gl_shader_manager: std::move std::string where applicable
Prevents unnecessary copies from occurring. In the case of ShaderCache,
we can just remove the parameter, given it's unused.
2020-04-18 21:40:21 -04:00
Lioncash
12279b22cc gl_resource_manager: Make use of noexcept on move assignment and move constructors
Several standard constructors generally check if objects can be moved in
a non-throwing manner (usually via std::move_if_noexcept) to preserve
its exception guarantees. This means that if these were used with
certain containers any reallocations internally would cause resource
churn, as copies would be necessary instead of moves.

This way, if they're every used in that manner, the right behavior is
always performed.
2020-04-18 21:12:03 -04:00
Lioncash
dd2a86bed1 video_core: Amend misplaced forward declarations
ShaderProgramManager was placed within the wrong namespace. Backend
simply isn't necessary, so it can be removed.
2020-04-18 20:12:22 -04:00
B3n30
922a5f738a fix typo; updated comment about apples bad driver 2020-04-18 15:55:19 +02:00
B3n30
1e54f27cde Add flag to disable seperable shaders for osx Intel GPUs. Default is false 2020-04-18 13:37:58 +02:00
Hamish Milne
7af4b1a18e Merge branch 'master' into feature/savestates-2 2020-04-17 00:01:43 +01:00
Marshall Mohror
55ecaa0344
video_core: add ScaleForce to texture filters (#5210)
* video_core: add ScaleForce to texture filters

* fix erroneous namespace end comment

Co-Authored-By: Valentin Vanelslande <vvanelslandedev@gmail.com>

* move copyright header to shader file

Co-authored-by: Valentin Vanelslande <vvanelslandedev@gmail.com>
2020-04-16 13:33:14 -05:00
Hamish Milne
9cd669db84 Merge branch 'master' into feature/savestates-2 2020-04-16 19:03:49 +01:00
Marshall Mohror
45dfb3aacc
Merge pull request #5186 from hamish-milne/feature/fix-5067
Revert one change from #4844
2020-04-13 23:57:46 -05:00
Hamish Milne
828f88d20a Merge branch 'master' into feature/savestates-2 2020-04-12 11:24:06 +01:00
Hamish Milne
a210e7e2bd Sync GPU state after loading (fix FE terrain bug) 2020-04-11 10:28:52 +01:00
Hamish Milne
eb78fe0c10
Revert one change from #4844
This fixes #5067 by reverting a speculative change made in a previous PR.

From this one can conclude that, for disabled textures, black (0,0,0,1) is the correct colour and clear (0,0,0,0) is not.
2020-04-07 15:38:24 +01:00
Marshall Mohror
d37b0476ad
video_core/renderer_opengl/gl_rasterizer_cache: Create Format Reinterpretation Framework (#5170)
* video_core/renderer_opengl/gl_rasterizer_cache: Create Format Reinterpretation Framework

Adds RGBA4 -> RGB5A1 reinterpretation commonly used by virtual console
If no matching surface can be found, ValidateSurface checks for a surface in the cache which is reinterpretable to the requested format.
If that fails, the cache is checked for any surface with a matching bit-width. If one is found, the region is flushed.
If not, the region is checked against dirty_regions to see if it was created entirely on the GPU.
If not, then the surface is flushed.

Co-Authored-By: James Rowe <jroweboy@users.noreply.github.com>
Co-Authored-By: Ben <b3n30@users.noreply.github.com>

temporary change to avoid merge conflicts with video dumping

* re-add D24S8->RGBA8 res_scale hack

* adress review comments

* fix dirty region check

* check for surfaces with invalid pixel format, and break logic into separate functions
2020-04-07 09:12:32 -05:00
Marshall Mohror
9c7da35382
Merge pull request #5083 from zhaowenlan1779/video-dumping-update
video_core, citra_qt: Video dumping updates
2020-04-03 21:15:32 -05:00