ns-emu/pineapple-src
1
0
Fork 0
forked from etc/pineapple-src
Code Pull requests Activity

early-access version 1520

Browse source
This commit is contained in:
pineappleEA 2021-03-14 09:08:46 +01:00
parent 7c0c4aee76
commit 5f8c5c9a16
4 changed files with 30 additions and 32 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
README.md
View file

@ -1,7 +1,7 @@
yuzu emulator early access
=============
This is the source code for early-access 1519.
This is the source code for early-access 1520.
## Legal Notice

4
src/video_core/renderer_opengl/util_shaders.cpp
View file

@ -111,6 +111,8 @@ void UtilShaders::ASTCDecode(Image& image, const ImageBufferMap& map,
glFlushMappedNamedBufferRange(map.buffer, map.offset, image.guest_size_bytes);
glUniform2ui(1, tile_size.width, tile_size.height);
// Ensure buffer data is valid before dispatching
glFlush();
for (const SwizzleParameters& swizzle : swizzles) {
const size_t input_offset = swizzle.buffer_offset + map.offset;
const u32 num_dispatches_x = Common::DivCeil(swizzle.num_tiles.width, 32U);
@ -133,8 +135,6 @@ void UtilShaders::ASTCDecode(Image& image, const ImageBufferMap& map,
glBindBufferRange(GL_SHADER_STORAGE_BUFFER, BINDING_INPUT_BUFFER, map.buffer, input_offset,
image.guest_size_bytes - swizzle.buffer_offset);
// Ensure buffer data is valid before dispatching compute
glFinish();
glDispatchCompute(num_dispatches_x, num_dispatches_y, image.info.resources.layers);
}
program_manager.RestoreGuestCompute();

53
src/video_core/renderer_vulkan/vk_compute_pass.cpp
View file

@ -472,30 +472,33 @@ void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map,
if (!data_buffer) {
MakeDataBuffer();
}
const VkPipeline vk_pipeline = *pipeline;
const VkImageAspectFlags aspect_mask = image.AspectMask();
const VkImage vk_image = image.Handle();
const bool is_initialized = image.ExchangeInitialization();
scheduler.Record([vk_image, aspect_mask, is_initialized](vk::CommandBuffer cmdbuf) {
const VkImageMemoryBarrier image_barrier{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT,
.dstAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT,
.oldLayout = is_initialized ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_UNDEFINED,
.newLayout = VK_IMAGE_LAYOUT_GENERAL,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = vk_image,
.subresourceRange{
.aspectMask = aspect_mask,
.baseMipLevel = 0,
.levelCount = VK_REMAINING_MIP_LEVELS,
.baseArrayLayer = 0,
.layerCount = VK_REMAINING_ARRAY_LAYERS,
},
};
cmdbuf.PipelineBarrier(0, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, 0, image_barrier);
});
scheduler.Record(
[vk_pipeline, vk_image, aspect_mask, is_initialized](vk::CommandBuffer cmdbuf) {
const VkImageMemoryBarrier image_barrier{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT,
.dstAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT,
.oldLayout = is_initialized ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_UNDEFINED,
.newLayout = VK_IMAGE_LAYOUT_GENERAL,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = vk_image,
.subresourceRange{
.aspectMask = aspect_mask,
.baseMipLevel = 0,
.levelCount = VK_REMAINING_MIP_LEVELS,
.baseArrayLayer = 0,
.layerCount = VK_REMAINING_ARRAY_LAYERS,
},
};
cmdbuf.PipelineBarrier(0, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, 0, image_barrier);
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, vk_pipeline);
});
for (const VideoCommon::SwizzleParameters& swizzle : swizzles) {
const size_t input_offset = swizzle.buffer_offset + map.offset;
const u32 num_dispatches_x = Common::DivCeil(swizzle.num_tiles.width, 32U);
@ -522,14 +525,13 @@ void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map,
const VkDescriptorSet set = CommitDescriptorSet(update_descriptor_queue);
const VkPipelineLayout vk_layout = *layout;
const VkPipeline vk_pipeline = *pipeline;
// To unswizzle the ASTC data
const auto params = MakeBlockLinearSwizzle2DParams(swizzle, image.info);
ASSERT(params.origin == (std::array<u32, 3>{0, 0, 0}));
ASSERT(params.destination == (std::array<s32, 3>{0, 0, 0}));
scheduler.Record([vk_layout, vk_pipeline, num_dispatches_x, num_dispatches_y,
num_dispatches_z, block_dims, params, set](vk::CommandBuffer cmdbuf) {
scheduler.Record([vk_layout, num_dispatches_x, num_dispatches_y, num_dispatches_z,
block_dims, params, set](vk::CommandBuffer cmdbuf) {
const AstcPushConstants uniforms{
.blocks_dims = block_dims,
.bytes_per_block_log2 = params.bytes_per_block_log2,
@ -539,7 +541,6 @@ void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map,
.block_height = params.block_height,
.block_height_mask = params.block_height_mask,
};
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, vk_pipeline);
cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, vk_layout, 0, set, {});
cmdbuf.PushConstants(vk_layout, VK_SHADER_STAGE_COMPUTE_BIT, uniforms);
cmdbuf.Dispatch(num_dispatches_x, num_dispatches_y, num_dispatches_z);

3
src/video_core/textures/astc.h
View file

@ -129,7 +129,4 @@ struct AstcBufferData {
decltype(REPLICATE_BYTE_TO_16_TABLE) replicate_byte_to_16 = REPLICATE_BYTE_TO_16_TABLE;
} constexpr ASTC_BUFFER_DATA;
void Decompress(std::span<const uint8_t> data, uint32_t width, uint32_t height, uint32_t depth,
uint32_t block_width, uint32_t block_height, std::span<uint8_t> output);
} // namespace Tegra::Texture::ASTC