mirror of
https://git.suyu.dev/suyu/suyu
synced 2024-11-05 06:47:53 +00:00
4e35177e23
Implement VOTE using Nvidia's intrinsics. Documentation about these can be found here https://developer.nvidia.com/reading-between-threads-shader-intrinsics Instead of using portable ARB instructions I opted to use Nvidia intrinsics because these are the closest we have to how Tegra X1 hardware renders. To stub VOTE on non-Nvidia drivers (including nouveau) this commit simulates a GPU with a warp size of one, returning what is meaningful for the instruction being emulated: * anyThreadNV(value) -> value * allThreadsNV(value) -> value * allThreadsEqualNV(value) -> true ballotARB, also known as "uint64_t(activeThreadsNV())", emits VOTE.ANY Rd, PT, PT; on nouveau's compiler. This doesn't match exactly to Nvidia's code VOTE.ALL Rd, PT, PT; Which is emulated with activeThreadsNV() by this commit. In theory this shouldn't really matter since .ANY, .ALL and .EQ affect the predicates (set to PT on those cases) and not the registers.
153 lines
5.2 KiB
CMake
153 lines
5.2 KiB
CMake
# Add a custom command to generate a new shader_cache_version hash when any of the following files change
|
|
# NOTE: This is an approximation of what files affect shader generation, its possible something else
|
|
# could affect the result, but much more unlikely than the following files. Keeping a list of files
|
|
# like this allows for much better caching since it doesn't force the user to recompile binary shaders every update
|
|
set(VIDEO_CORE "${CMAKE_SOURCE_DIR}/src/video_core")
|
|
if (DEFINED ENV{CI})
|
|
if (DEFINED ENV{TRAVIS})
|
|
set(BUILD_REPOSITORY $ENV{TRAVIS_REPO_SLUG})
|
|
set(BUILD_TAG $ENV{TRAVIS_TAG})
|
|
elseif(DEFINED ENV{APPVEYOR})
|
|
set(BUILD_REPOSITORY $ENV{APPVEYOR_REPO_NAME})
|
|
set(BUILD_TAG $ENV{APPVEYOR_REPO_TAG_NAME})
|
|
endif()
|
|
endif()
|
|
add_custom_command(OUTPUT scm_rev.cpp
|
|
COMMAND ${CMAKE_COMMAND}
|
|
-DSRC_DIR="${CMAKE_SOURCE_DIR}"
|
|
-DBUILD_REPOSITORY="${BUILD_REPOSITORY}"
|
|
-DBUILD_TAG="${BUILD_TAG}"
|
|
-P "${CMAKE_SOURCE_DIR}/CMakeModules/GenerateSCMRev.cmake"
|
|
DEPENDS
|
|
# WARNING! It was too much work to try and make a common location for this list,
|
|
# so if you need to change it, please update CMakeModules/GenerateSCMRev.cmake as well
|
|
"${VIDEO_CORE}/renderer_opengl/gl_shader_cache.cpp"
|
|
"${VIDEO_CORE}/renderer_opengl/gl_shader_cache.h"
|
|
"${VIDEO_CORE}/renderer_opengl/gl_shader_decompiler.cpp"
|
|
"${VIDEO_CORE}/renderer_opengl/gl_shader_decompiler.h"
|
|
"${VIDEO_CORE}/renderer_opengl/gl_shader_disk_cache.cpp"
|
|
"${VIDEO_CORE}/renderer_opengl/gl_shader_disk_cache.h"
|
|
"${VIDEO_CORE}/renderer_opengl/gl_shader_gen.cpp"
|
|
"${VIDEO_CORE}/renderer_opengl/gl_shader_gen.h"
|
|
"${VIDEO_CORE}/shader/decode/arithmetic.cpp"
|
|
"${VIDEO_CORE}/shader/decode/arithmetic_half.cpp"
|
|
"${VIDEO_CORE}/shader/decode/arithmetic_half_immediate.cpp"
|
|
"${VIDEO_CORE}/shader/decode/arithmetic_immediate.cpp"
|
|
"${VIDEO_CORE}/shader/decode/arithmetic_integer.cpp"
|
|
"${VIDEO_CORE}/shader/decode/arithmetic_integer_immediate.cpp"
|
|
"${VIDEO_CORE}/shader/decode/bfe.cpp"
|
|
"${VIDEO_CORE}/shader/decode/bfi.cpp"
|
|
"${VIDEO_CORE}/shader/decode/conversion.cpp"
|
|
"${VIDEO_CORE}/shader/decode/ffma.cpp"
|
|
"${VIDEO_CORE}/shader/decode/float_set.cpp"
|
|
"${VIDEO_CORE}/shader/decode/float_set_predicate.cpp"
|
|
"${VIDEO_CORE}/shader/decode/half_set.cpp"
|
|
"${VIDEO_CORE}/shader/decode/half_set_predicate.cpp"
|
|
"${VIDEO_CORE}/shader/decode/hfma2.cpp"
|
|
"${VIDEO_CORE}/shader/decode/image.cpp"
|
|
"${VIDEO_CORE}/shader/decode/integer_set.cpp"
|
|
"${VIDEO_CORE}/shader/decode/integer_set_predicate.cpp"
|
|
"${VIDEO_CORE}/shader/decode/memory.cpp"
|
|
"${VIDEO_CORE}/shader/decode/texture.cpp"
|
|
"${VIDEO_CORE}/shader/decode/other.cpp"
|
|
"${VIDEO_CORE}/shader/decode/predicate_set_predicate.cpp"
|
|
"${VIDEO_CORE}/shader/decode/predicate_set_register.cpp"
|
|
"${VIDEO_CORE}/shader/decode/register_set_predicate.cpp"
|
|
"${VIDEO_CORE}/shader/decode/shift.cpp"
|
|
"${VIDEO_CORE}/shader/decode/video.cpp"
|
|
"${VIDEO_CORE}/shader/decode/warp.cpp"
|
|
"${VIDEO_CORE}/shader/decode/xmad.cpp"
|
|
"${VIDEO_CORE}/shader/control_flow.cpp"
|
|
"${VIDEO_CORE}/shader/control_flow.h"
|
|
"${VIDEO_CORE}/shader/decode.cpp"
|
|
"${VIDEO_CORE}/shader/node.h"
|
|
"${VIDEO_CORE}/shader/node_helper.cpp"
|
|
"${VIDEO_CORE}/shader/node_helper.h"
|
|
"${VIDEO_CORE}/shader/shader_ir.cpp"
|
|
"${VIDEO_CORE}/shader/shader_ir.h"
|
|
"${VIDEO_CORE}/shader/track.cpp"
|
|
# and also check that the scm_rev files haven't changed
|
|
"${CMAKE_CURRENT_SOURCE_DIR}/scm_rev.cpp.in"
|
|
"${CMAKE_CURRENT_SOURCE_DIR}/scm_rev.h"
|
|
# technically we should regenerate if the git version changed, but its not worth the effort imo
|
|
"${CMAKE_SOURCE_DIR}/CMakeModules/GenerateSCMRev.cmake"
|
|
)
|
|
|
|
add_library(common STATIC
|
|
alignment.h
|
|
assert.h
|
|
detached_tasks.cpp
|
|
detached_tasks.h
|
|
binary_find.h
|
|
bit_field.h
|
|
bit_util.h
|
|
cityhash.cpp
|
|
cityhash.h
|
|
color.h
|
|
common_funcs.h
|
|
common_paths.h
|
|
common_types.h
|
|
file_util.cpp
|
|
file_util.h
|
|
hash.h
|
|
hex_util.cpp
|
|
hex_util.h
|
|
logging/backend.cpp
|
|
logging/backend.h
|
|
logging/filter.cpp
|
|
logging/filter.h
|
|
logging/log.h
|
|
logging/text_formatter.cpp
|
|
logging/text_formatter.h
|
|
lz4_compression.cpp
|
|
lz4_compression.h
|
|
math_util.h
|
|
memory_hook.cpp
|
|
memory_hook.h
|
|
microprofile.cpp
|
|
microprofile.h
|
|
microprofileui.h
|
|
misc.cpp
|
|
multi_level_queue.h
|
|
page_table.cpp
|
|
page_table.h
|
|
param_package.cpp
|
|
param_package.h
|
|
quaternion.h
|
|
ring_buffer.h
|
|
scm_rev.cpp
|
|
scm_rev.h
|
|
scope_exit.h
|
|
string_util.cpp
|
|
string_util.h
|
|
swap.h
|
|
telemetry.cpp
|
|
telemetry.h
|
|
thread.cpp
|
|
thread.h
|
|
thread_queue_list.h
|
|
threadsafe_queue.h
|
|
timer.cpp
|
|
timer.h
|
|
uint128.cpp
|
|
uint128.h
|
|
uuid.cpp
|
|
uuid.h
|
|
vector_math.h
|
|
web_result.h
|
|
zstd_compression.cpp
|
|
zstd_compression.h
|
|
)
|
|
|
|
if(ARCHITECTURE_x86_64)
|
|
target_sources(common
|
|
PRIVATE
|
|
x64/cpu_detect.cpp
|
|
x64/cpu_detect.h
|
|
)
|
|
endif()
|
|
|
|
create_target_directory_groups(common)
|
|
|
|
target_link_libraries(common PUBLIC Boost::boost fmt microprofile)
|
|
target_link_libraries(common PRIVATE lz4_static libzstd_static)
|