diff options
Diffstat (limited to 'src/video_core/renderer_opengl/util_shaders.cpp')
-rw-r--r-- | src/video_core/renderer_opengl/util_shaders.cpp | 119 |
1 files changed, 98 insertions, 21 deletions
diff --git a/src/video_core/renderer_opengl/util_shaders.cpp b/src/video_core/renderer_opengl/util_shaders.cpp index eb849cbf2..2fe4799bc 100644 --- a/src/video_core/renderer_opengl/util_shaders.cpp +++ b/src/video_core/renderer_opengl/util_shaders.cpp @@ -14,6 +14,7 @@ #include "video_core/host_shaders/block_linear_unswizzle_2d_comp.h" #include "video_core/host_shaders/block_linear_unswizzle_3d_comp.h" #include "video_core/host_shaders/opengl_copy_bc4_comp.h" +#include "video_core/host_shaders/opengl_copy_bgra_comp.h" #include "video_core/host_shaders/pitch_unswizzle_comp.h" #include "video_core/renderer_opengl/gl_resource_manager.h" #include "video_core/renderer_opengl/gl_shader_manager.h" @@ -48,6 +49,11 @@ OGLProgram MakeProgram(std::string_view source) { return program; } +size_t NumPixelsInCopy(const VideoCommon::ImageCopy& copy) { + return static_cast<size_t>(copy.extent.width * copy.extent.height * + copy.src_subresource.num_layers); +} + } // Anonymous namespace UtilShaders::UtilShaders(ProgramManager& program_manager_) @@ -55,6 +61,7 @@ UtilShaders::UtilShaders(ProgramManager& program_manager_) block_linear_unswizzle_2d_program(MakeProgram(BLOCK_LINEAR_UNSWIZZLE_2D_COMP)), block_linear_unswizzle_3d_program(MakeProgram(BLOCK_LINEAR_UNSWIZZLE_3D_COMP)), pitch_unswizzle_program(MakeProgram(PITCH_UNSWIZZLE_COMP)), + copy_bgra_program(MakeProgram(OPENGL_COPY_BGRA_COMP)), copy_bc4_program(MakeProgram(OPENGL_COPY_BC4_COMP)) { const auto swizzle_table = Tegra::Texture::MakeSwizzleTable(); swizzle_table_buffer.Create(); @@ -63,7 +70,7 @@ UtilShaders::UtilShaders(ProgramManager& program_manager_) UtilShaders::~UtilShaders() = default; -void UtilShaders::BlockLinearUpload2D(Image& image, const ImageBufferMap& map, size_t buffer_offset, +void UtilShaders::BlockLinearUpload2D(Image& image, const ImageBufferMap& map, std::span<const SwizzleParameters> swizzles) { static constexpr Extent3D WORKGROUP_SIZE{32, 32, 1}; static constexpr GLuint BINDING_SWIZZLE_BUFFER = 0; @@ -71,13 +78,13 @@ void UtilShaders::BlockLinearUpload2D(Image& image, const ImageBufferMap& map, s static constexpr GLuint BINDING_OUTPUT_IMAGE = 0; program_manager.BindHostCompute(block_linear_unswizzle_2d_program.handle); - glFlushMappedNamedBufferRange(map.Handle(), buffer_offset, image.guest_size_bytes); + glFlushMappedNamedBufferRange(map.buffer, map.offset, image.guest_size_bytes); glBindBufferBase(GL_SHADER_STORAGE_BUFFER, BINDING_SWIZZLE_BUFFER, swizzle_table_buffer.handle); const GLenum store_format = StoreFormat(BytesPerBlock(image.info.format)); for (const SwizzleParameters& swizzle : swizzles) { const Extent3D num_tiles = swizzle.num_tiles; - const size_t input_offset = swizzle.buffer_offset + buffer_offset; + const size_t input_offset = swizzle.buffer_offset + map.offset; const u32 num_dispatches_x = Common::DivCeil(num_tiles.width, WORKGROUP_SIZE.width); const u32 num_dispatches_y = Common::DivCeil(num_tiles.height, WORKGROUP_SIZE.height); @@ -91,16 +98,16 @@ void UtilShaders::BlockLinearUpload2D(Image& image, const ImageBufferMap& map, s glUniform1ui(5, params.x_shift); glUniform1ui(6, params.block_height); glUniform1ui(7, params.block_height_mask); - glBindBufferRange(GL_SHADER_STORAGE_BUFFER, BINDING_INPUT_BUFFER, map.Handle(), - input_offset, image.guest_size_bytes - swizzle.buffer_offset); - glBindImageTexture(BINDING_OUTPUT_IMAGE, image.Handle(), swizzle.level, GL_TRUE, 0, + glBindBufferRange(GL_SHADER_STORAGE_BUFFER, BINDING_INPUT_BUFFER, map.buffer, input_offset, + image.guest_size_bytes - swizzle.buffer_offset); + glBindImageTexture(BINDING_OUTPUT_IMAGE, image.StorageHandle(), swizzle.level, GL_TRUE, 0, GL_WRITE_ONLY, store_format); glDispatchCompute(num_dispatches_x, num_dispatches_y, image.info.resources.layers); } program_manager.RestoreGuestCompute(); } -void UtilShaders::BlockLinearUpload3D(Image& image, const ImageBufferMap& map, size_t buffer_offset, +void UtilShaders::BlockLinearUpload3D(Image& image, const ImageBufferMap& map, std::span<const SwizzleParameters> swizzles) { static constexpr Extent3D WORKGROUP_SIZE{16, 8, 8}; @@ -108,14 +115,14 @@ void UtilShaders::BlockLinearUpload3D(Image& image, const ImageBufferMap& map, s static constexpr GLuint BINDING_INPUT_BUFFER = 1; static constexpr GLuint BINDING_OUTPUT_IMAGE = 0; - glFlushMappedNamedBufferRange(map.Handle(), buffer_offset, image.guest_size_bytes); + glFlushMappedNamedBufferRange(map.buffer, map.offset, image.guest_size_bytes); program_manager.BindHostCompute(block_linear_unswizzle_3d_program.handle); glBindBufferBase(GL_SHADER_STORAGE_BUFFER, BINDING_SWIZZLE_BUFFER, swizzle_table_buffer.handle); const GLenum store_format = StoreFormat(BytesPerBlock(image.info.format)); for (const SwizzleParameters& swizzle : swizzles) { const Extent3D num_tiles = swizzle.num_tiles; - const size_t input_offset = swizzle.buffer_offset + buffer_offset; + const size_t input_offset = swizzle.buffer_offset + map.offset; const u32 num_dispatches_x = Common::DivCeil(num_tiles.width, WORKGROUP_SIZE.width); const u32 num_dispatches_y = Common::DivCeil(num_tiles.height, WORKGROUP_SIZE.height); @@ -132,16 +139,16 @@ void UtilShaders::BlockLinearUpload3D(Image& image, const ImageBufferMap& map, s glUniform1ui(7, params.block_height_mask); glUniform1ui(8, params.block_depth); glUniform1ui(9, params.block_depth_mask); - glBindBufferRange(GL_SHADER_STORAGE_BUFFER, BINDING_INPUT_BUFFER, map.Handle(), - input_offset, image.guest_size_bytes - swizzle.buffer_offset); - glBindImageTexture(BINDING_OUTPUT_IMAGE, image.Handle(), swizzle.level, GL_TRUE, 0, + glBindBufferRange(GL_SHADER_STORAGE_BUFFER, BINDING_INPUT_BUFFER, map.buffer, input_offset, + image.guest_size_bytes - swizzle.buffer_offset); + glBindImageTexture(BINDING_OUTPUT_IMAGE, image.StorageHandle(), swizzle.level, GL_TRUE, 0, GL_WRITE_ONLY, store_format); glDispatchCompute(num_dispatches_x, num_dispatches_y, num_dispatches_z); } program_manager.RestoreGuestCompute(); } -void UtilShaders::PitchUpload(Image& image, const ImageBufferMap& map, size_t buffer_offset, +void UtilShaders::PitchUpload(Image& image, const ImageBufferMap& map, std::span<const SwizzleParameters> swizzles) { static constexpr Extent3D WORKGROUP_SIZE{32, 32, 1}; static constexpr GLuint BINDING_INPUT_BUFFER = 0; @@ -159,21 +166,22 @@ void UtilShaders::PitchUpload(Image& image, const ImageBufferMap& map, size_t bu "Non-power of two images are not implemented"); program_manager.BindHostCompute(pitch_unswizzle_program.handle); - glFlushMappedNamedBufferRange(map.Handle(), buffer_offset, image.guest_size_bytes); + glFlushMappedNamedBufferRange(map.buffer, map.offset, image.guest_size_bytes); glUniform2ui(LOC_ORIGIN, 0, 0); glUniform2i(LOC_DESTINATION, 0, 0); glUniform1ui(LOC_BYTES_PER_BLOCK, bytes_per_block); glUniform1ui(LOC_PITCH, pitch); - glBindImageTexture(BINDING_OUTPUT_IMAGE, image.Handle(), 0, GL_FALSE, 0, GL_WRITE_ONLY, format); + glBindImageTexture(BINDING_OUTPUT_IMAGE, image.StorageHandle(), 0, GL_FALSE, 0, GL_WRITE_ONLY, + format); for (const SwizzleParameters& swizzle : swizzles) { const Extent3D num_tiles = swizzle.num_tiles; - const size_t input_offset = swizzle.buffer_offset + buffer_offset; + const size_t input_offset = swizzle.buffer_offset + map.offset; const u32 num_dispatches_x = Common::DivCeil(num_tiles.width, WORKGROUP_SIZE.width); const u32 num_dispatches_y = Common::DivCeil(num_tiles.height, WORKGROUP_SIZE.height); - glBindBufferRange(GL_SHADER_STORAGE_BUFFER, BINDING_INPUT_BUFFER, map.Handle(), - input_offset, image.guest_size_bytes - swizzle.buffer_offset); + glBindBufferRange(GL_SHADER_STORAGE_BUFFER, BINDING_INPUT_BUFFER, map.buffer, input_offset, + image.guest_size_bytes - swizzle.buffer_offset); glDispatchCompute(num_dispatches_x, num_dispatches_y, 1); } program_manager.RestoreGuestCompute(); @@ -195,15 +203,52 @@ void UtilShaders::CopyBC4(Image& dst_image, Image& src_image, std::span<const Im glUniform3ui(LOC_SRC_OFFSET, copy.src_offset.x, copy.src_offset.y, copy.src_offset.z); glUniform3ui(LOC_DST_OFFSET, copy.dst_offset.x, copy.dst_offset.y, copy.dst_offset.z); - glBindImageTexture(BINDING_INPUT_IMAGE, src_image.Handle(), copy.src_subresource.base_level, - GL_FALSE, 0, GL_READ_ONLY, GL_RG32UI); - glBindImageTexture(BINDING_OUTPUT_IMAGE, dst_image.Handle(), + glBindImageTexture(BINDING_INPUT_IMAGE, src_image.StorageHandle(), + copy.src_subresource.base_level, GL_FALSE, 0, GL_READ_ONLY, GL_RG32UI); + glBindImageTexture(BINDING_OUTPUT_IMAGE, dst_image.StorageHandle(), copy.dst_subresource.base_level, GL_FALSE, 0, GL_WRITE_ONLY, GL_RGBA8UI); glDispatchCompute(copy.extent.width, copy.extent.height, copy.extent.depth); } program_manager.RestoreGuestCompute(); } +void UtilShaders::CopyBGR(Image& dst_image, Image& src_image, + std::span<const VideoCommon::ImageCopy> copies) { + static constexpr GLuint BINDING_INPUT_IMAGE = 0; + static constexpr GLuint BINDING_OUTPUT_IMAGE = 1; + static constexpr VideoCommon::Offset3D zero_offset{0, 0, 0}; + const u32 bytes_per_block = BytesPerBlock(dst_image.info.format); + switch (bytes_per_block) { + case 2: + // BGR565 copy + for (const ImageCopy& copy : copies) { + ASSERT(copy.src_offset == zero_offset); + ASSERT(copy.dst_offset == zero_offset); + bgr_copy_pass.Execute(dst_image, src_image, copy); + } + break; + case 4: { + // BGRA8 copy + program_manager.BindHostCompute(copy_bgra_program.handle); + constexpr GLenum FORMAT = GL_RGBA8; + for (const ImageCopy& copy : copies) { + ASSERT(copy.src_offset == zero_offset); + ASSERT(copy.dst_offset == zero_offset); + glBindImageTexture(BINDING_INPUT_IMAGE, src_image.StorageHandle(), + copy.src_subresource.base_level, GL_FALSE, 0, GL_READ_ONLY, FORMAT); + glBindImageTexture(BINDING_OUTPUT_IMAGE, dst_image.StorageHandle(), + copy.dst_subresource.base_level, GL_FALSE, 0, GL_WRITE_ONLY, FORMAT); + glDispatchCompute(copy.extent.width, copy.extent.height, copy.extent.depth); + } + program_manager.RestoreGuestCompute(); + break; + } + default: + UNREACHABLE(); + break; + } +} + GLenum StoreFormat(u32 bytes_per_block) { switch (bytes_per_block) { case 1: @@ -221,4 +266,36 @@ GLenum StoreFormat(u32 bytes_per_block) { return GL_R8UI; } +void Bgr565CopyPass::Execute(const Image& dst_image, const Image& src_image, + const ImageCopy& copy) { + if (CopyBufferCreationNeeded(copy)) { + CreateNewCopyBuffer(copy, GL_TEXTURE_2D_ARRAY, GL_RGB565); + } + // Copy from source to PBO + glPixelStorei(GL_PACK_ALIGNMENT, 1); + glPixelStorei(GL_PACK_ROW_LENGTH, copy.extent.width); + glBindBuffer(GL_PIXEL_PACK_BUFFER, bgr16_pbo.handle); + glGetTextureSubImage(src_image.Handle(), 0, 0, 0, 0, copy.extent.width, copy.extent.height, + copy.src_subresource.num_layers, GL_RGB, GL_UNSIGNED_SHORT_5_6_5, + static_cast<GLsizei>(bgr16_pbo_size), nullptr); + + // Copy from PBO to destination in reverse order + glPixelStorei(GL_UNPACK_ALIGNMENT, 1); + glPixelStorei(GL_UNPACK_ROW_LENGTH, copy.extent.width); + glBindBuffer(GL_PIXEL_UNPACK_BUFFER, bgr16_pbo.handle); + glTextureSubImage3D(dst_image.Handle(), 0, 0, 0, 0, copy.extent.width, copy.extent.height, + copy.dst_subresource.num_layers, GL_RGB, GL_UNSIGNED_SHORT_5_6_5_REV, + nullptr); +} + +bool Bgr565CopyPass::CopyBufferCreationNeeded(const ImageCopy& copy) { + return bgr16_pbo_size < NumPixelsInCopy(copy) * sizeof(u16); +} + +void Bgr565CopyPass::CreateNewCopyBuffer(const ImageCopy& copy, GLenum target, GLuint format) { + bgr16_pbo.Create(); + bgr16_pbo_size = NumPixelsInCopy(copy) * sizeof(u16); + glNamedBufferData(bgr16_pbo.handle, bgr16_pbo_size, nullptr, GL_STREAM_COPY); +} + } // namespace OpenGL |