artemist/yuzu
1
0
Fork 0
Code Issues Pull requests Packages Projects Releases Wiki Activity

TextureCache: fix rescaling in aliases and overlap joins.

Browse source
This commit is contained in:
FernandoS27 2021-10-20 18:27:25 +02:00 committed by Fernando Sahmkow
parent 7506ac4118
commit d37d10e7a7
4 changed files with 48 additions and 23 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

19
src/video_core/renderer_vulkan/vk_rasterizer.cpp
View file

@ -63,7 +63,7 @@ VkViewport GetViewportState(const Device& device, const Maxwell& regs, size_t in
const auto conv = [scale](float value) {
float new_value = value * scale;
if (scale < 1.0f) {
bool sign = std::signbit(new_value);
const bool sign = std::signbit(value);
new_value = std::round(std::abs(new_value));
new_value = sign ? -new_value : new_value;
}
@ -96,21 +96,22 @@ VkViewport GetViewportState(const Device& device, const Maxwell& regs, size_t in
VkRect2D GetScissorState(const Maxwell& regs, size_t index, u32 up_scale = 1, u32 down_shift = 0) {
const auto& src = regs.scissor_test[index];
VkRect2D scissor;
const auto scale_up = [&](u32 value) -> u32 {
const auto scale_up = [&](s32 value) -> s32 {
if (value == 0) {
return 0U;
}
const u32 upset = value * up_scale;
u32 acumm = 0;
const s32 upset = value * up_scale;
s32 acumm = 0;
if ((up_scale >> down_shift) == 0) {
acumm = upset & 0x1;
acumm = upset % 2;
}
const u32 converted_value = (value * up_scale) >> down_shift;
return std::max<u32>(converted_value + acumm, 1U);
const s32 converted_value = (value * up_scale) >> down_shift;
return value < 0 ? std::min<s32>(converted_value - acumm, -1)
: std::max<s32>(converted_value + acumm, 1);
};
if (src.enable) {
scissor.offset.x = static_cast<s32>(scale_up(src.min_x));
scissor.offset.y = static_cast<s32>(scale_up(src.min_y));
scissor.offset.x = scale_up(static_cast<s32>(src.min_x));
scissor.offset.y = scale_up(static_cast<s32>(src.min_y));
scissor.extent.width = scale_up(src.max_x - src.min_x);
scissor.extent.height = scale_up(src.max_y - src.min_y);
} else {

33
src/video_core/texture_cache/texture_cache.h
View file

@ -1037,8 +1037,11 @@ ImageId TextureCache<P>::JoinImages(const ImageInfo& info, GPUVAddr gpu_addr, VA
if (overlap.info.num_samples != new_image.info.num_samples) {
LOG_WARNING(HW_GPU, "Copying between images with different samples is not implemented");
} else {
const auto& resolution = Settings::values.resolution_info;
const SubresourceBase base = new_image.TryFindBase(overlap.gpu_addr).value();
auto copies = MakeShrinkImageCopies(new_info, overlap.info, base);
const u32 up_scale = can_rescale ? resolution.up_scale : 1;
const u32 down_shift = can_rescale ? resolution.down_shift : 0;
auto copies = MakeShrinkImageCopies(new_info, overlap.info, base, up_scale, down_shift);
runtime.CopyImage(new_image, overlap, std::move(copies));
}
if (True(overlap.flags & ImageFlagBits::Tracked)) {
@ -1659,19 +1662,35 @@ void TextureCache<P>::SynchronizeAliases(ImageId image_id) {
const ImageBase& rhs_image = slot_images[rhs->id];
return lhs_image.modification_tick < rhs_image.modification_tick;
});
const auto& resolution = Settings::values.resolution_info;
for (const AliasedImage* const aliased : aliased_images) {
if (any_rescaled) {
if (!resolution.active | !any_rescaled) {
CopyImage(image_id, aliased->id, aliased->copies);
continue;
}
Image& aliased_image = slot_images[aliased->id];
if (can_rescale) {
ScaleUp(aliased_image);
} else {
if (!can_rescale) {
ScaleDown(aliased_image);
if (any_blacklisted) {
aliased_image.flags |= ImageFlagBits::Blacklisted;
}
}
}
CopyImage(image_id, aliased->id, aliased->copies);
continue;
}
ScaleUp(aliased_image);
const bool both_2d{image.info.type == ImageType::e2D &&
aliased_image.info.type == ImageType::e2D};
auto copies = aliased->copies;
for (auto copy : copies) {
copy.extent.width = std::max<u32>(
(copy.extent.width * resolution.up_scale) >> resolution.down_shift, 1);
if (both_2d) {
copy.extent.height = std::max<u32>(
(copy.extent.height * resolution.up_scale) >> resolution.down_shift, 1);
}
}
CopyImage(image_id, aliased->id, copies);
}
}

8
src/video_core/texture_cache/util.cpp
View file

@ -723,7 +723,7 @@ ImageViewType RenderTargetImageViewType(const ImageInfo& info) noexcept {
}
std::vector<ImageCopy> MakeShrinkImageCopies(const ImageInfo& dst, const ImageInfo& src,
SubresourceBase base) {
SubresourceBase base, u32 up_scale, u32 down_shift) {
ASSERT(dst.resources.levels >= src.resources.levels);
ASSERT(dst.num_samples == src.num_samples);
@ -732,7 +732,7 @@ std::vector<ImageCopy> MakeShrinkImageCopies(const ImageInfo& dst, const ImageIn
ASSERT(src.type == ImageType::e3D);
ASSERT(src.resources.levels == 1);
}
const bool both_2d{src.type == ImageType::e2D && dst.type == ImageType::e2D};
std::vector<ImageCopy> copies;
copies.reserve(src.resources.levels);
for (s32 level = 0; level < src.resources.levels; ++level) {
@ -762,6 +762,10 @@ std::vector<ImageCopy> MakeShrinkImageCopies(const ImageInfo& dst, const ImageIn
if (is_dst_3d) {
copy.extent.depth = src.size.depth;
}
copy.extent.width = std::max<u32>((copy.extent.width * up_scale) >> down_shift, 1);
if (both_2d) {
copy.extent.height = std::max<u32>((copy.extent.height * up_scale) >> down_shift, 1);
}
}
return copies;
}

3
src/video_core/texture_cache/util.h
View file

@ -55,7 +55,8 @@ struct OverlapResult {
[[nodiscard]] std::vector<ImageCopy> MakeShrinkImageCopies(const ImageInfo& dst,
const ImageInfo& src,
SubresourceBase base);
SubresourceBase base, u32 up_scale = 1,
u32 down_shift = 0);
[[nodiscard]] bool IsValidEntry(const Tegra::MemoryManager& gpu_memory, const TICEntry& config);