rpcsx/rpcs3/Emu/RSX/Common/surface_store.h

1589 lines
49 KiB
C++

#pragma once
#include "Utilities/GSL.h"
#include "Emu/Memory/vm.h"
#include "TextureUtils.h"
#include "../GCM.h"
#include "../rsx_utils.h"
#include <list>
namespace
{
template <typename T>
gsl::span<T> as_const_span(gsl::span<const gsl::byte> unformated_span)
{
return{ (T*)unformated_span.data(), ::narrow<int>(unformated_span.size_bytes() / sizeof(T)) };
}
}
namespace rsx
{
namespace utility
{
std::vector<u8> get_rtt_indexes(surface_target color_target);
size_t get_aligned_pitch(surface_color_format format, u32 width);
size_t get_packed_pitch(surface_color_format format, u32 width);
}
enum surface_state_flags : u32
{
ready = 0,
erase_bkgnd = 1
};
template <typename surface_type>
struct surface_overlap_info_t
{
surface_type surface = nullptr;
u32 base_address = 0;
bool is_depth = false;
bool is_clipped = false;
u16 src_x = 0;
u16 src_y = 0;
u16 dst_x = 0;
u16 dst_y = 0;
u16 width = 0;
u16 height = 0;
areai get_src_area() const
{
return coordi{ {src_x, src_y}, {width, height} };
}
areai get_dst_area() const
{
return coordi{ {dst_x, dst_y}, {width, height} };
}
};
struct surface_format_info
{
u32 surface_width;
u32 surface_height;
u16 native_pitch;
u16 rsx_pitch;
u8 bpp;
};
template <typename surface_type>
struct deferred_clipped_region
{
u16 src_x, src_y, dst_x, dst_y, width, height;
f32 transfer_scale_x, transfer_scale_y;
surface_type target;
surface_type source;
template <typename T>
deferred_clipped_region<T> cast() const
{
deferred_clipped_region<T> ret;
ret.src_x = src_x;
ret.src_y = src_y;
ret.dst_x = dst_x;
ret.dst_y = dst_y;
ret.width = width;
ret.height = height;
ret.transfer_scale_x = transfer_scale_x;
ret.transfer_scale_y = transfer_scale_y;
ret.target = (T)(target);
ret.source = (T)(source);
return ret;
}
operator bool() const
{
return (source != nullptr);
}
template <typename T>
void init_transfer(T target_surface)
{
if (!width)
{
// Perform intersection here
const auto region = rsx::get_transferable_region(target_surface);
width = std::get<0>(region);
height = std::get<1>(region);
transfer_scale_x = f32(std::get<2>(region)) / width;
transfer_scale_y = f32(std::get<3>(region)) / height;
target = target_surface;
}
}
areai src_rect() const
{
verify(HERE), width;
return { src_x, src_y, src_x + width, src_y + height };
}
areai dst_rect() const
{
verify(HERE), width;
return { dst_x, dst_y, dst_x + u16(width * transfer_scale_x + 0.5f), dst_y + u16(height * transfer_scale_y + 0.5f) };
}
};
template <typename image_storage_type>
struct render_target_descriptor
{
u64 last_use_tag = 0; // tag indicating when this block was last confirmed to have been written to
std::array<std::pair<u32, u64>, 5> memory_tag_samples;
deferred_clipped_region<image_storage_type> old_contents{};
rsx::surface_antialiasing read_aa_mode = rsx::surface_antialiasing::center_1_sample;
GcmTileInfo *tile = nullptr;
rsx::surface_antialiasing write_aa_mode = rsx::surface_antialiasing::center_1_sample;
flags32_t memory_usage_flags = surface_usage_flags::unknown;
flags32_t state_flags = surface_state_flags::ready;
union
{
rsx::surface_color_format gcm_color_format;
rsx::surface_depth_format gcm_depth_format;
}
format_info;
render_target_descriptor() {}
virtual ~render_target_descriptor()
{
if (old_contents)
{
// Cascade resource derefs
LOG_ERROR(RSX, "Resource was destroyed whilst holding a resource reference!");
}
}
virtual image_storage_type get_surface() = 0;
virtual u16 get_surface_width() const = 0;
virtual u16 get_surface_height() const = 0;
virtual u16 get_rsx_pitch() const = 0;
virtual u16 get_native_pitch() const = 0;
virtual bool is_depth_surface() const = 0;
virtual void release_ref(image_storage_type) const = 0;
u8 get_bpp() const
{
return u8(get_native_pitch() / get_surface_width());
}
void save_aa_mode()
{
read_aa_mode = write_aa_mode;
write_aa_mode = rsx::surface_antialiasing::center_1_sample;
}
void reset_aa_mode()
{
write_aa_mode = read_aa_mode = rsx::surface_antialiasing::center_1_sample;
}
void set_format(rsx::surface_color_format format)
{
format_info.gcm_color_format = format;
}
void set_format(rsx::surface_depth_format format)
{
format_info.gcm_depth_format = format;
}
rsx::surface_color_format get_surface_color_format()
{
return format_info.gcm_color_format;
}
rsx::surface_depth_format get_surface_depth_format()
{
return format_info.gcm_depth_format;
}
bool dirty() const
{
return (state_flags != rsx::surface_state_flags::ready) || old_contents;
}
bool test() const
{
if (dirty())
{
// TODO
// Should RCB or mem-sync (inherit previous mem) to init memory
LOG_TODO(RSX, "Resource used before memory initialization");
}
// Tags are tested in an X pattern
for (const auto &tag : memory_tag_samples)
{
if (!tag.first)
break;
if (tag.second != *reinterpret_cast<u64*>(vm::g_sudo_addr + tag.first))
return false;
}
return true;
}
void clear_rw_barrier()
{
release_ref(old_contents.source);
old_contents = {};
}
template<typename T>
void set_old_contents(T* other)
{
verify(HERE), !old_contents;
if (!other || other->get_rsx_pitch() != this->get_rsx_pitch())
{
old_contents = {};
return;
}
old_contents = {};
old_contents.source = other;
other->add_ref();
}
template<typename T>
void set_old_contents_region(const T& region, bool normalized)
{
if (old_contents)
{
// This can happen when doing memory splits
auto old_surface = static_cast<decltype(region.source)>(old_contents.source);
if (old_surface->last_use_tag > region.source->last_use_tag)
{
return;
}
clear_rw_barrier();
}
// NOTE: This method will not perform pitch verification!
verify(HERE), !old_contents, region.source, region.source != this;
old_contents = region.template cast<image_storage_type>();
region.source->add_ref();
// Reverse normalization process if needed
if (normalized)
{
const u16 bytes_to_texels_x = region.source->get_bpp() * (region.source->write_aa_mode == rsx::surface_antialiasing::center_1_sample? 1 : 2);
const u16 rows_to_texels_y = (region.source->write_aa_mode > rsx::surface_antialiasing::diagonal_centered_2_samples? 2 : 1);
old_contents.src_x /= bytes_to_texels_x;
old_contents.src_y /= rows_to_texels_y;
old_contents.width /= bytes_to_texels_x;
old_contents.height /= rows_to_texels_y;
const u16 bytes_to_texels_x2 = (get_bpp() * (write_aa_mode == rsx::surface_antialiasing::center_1_sample? 1 : 2));
const u16 rows_to_texels_y2 = (write_aa_mode > rsx::surface_antialiasing::diagonal_centered_2_samples)? 2 : 1;
old_contents.dst_x /= bytes_to_texels_x2;
old_contents.dst_y /= rows_to_texels_y2;
old_contents.transfer_scale_x = f32(bytes_to_texels_x) / bytes_to_texels_x2;
old_contents.transfer_scale_y = f32(rows_to_texels_y) / rows_to_texels_y2;
}
// Apply resolution scale if needed
if (g_cfg.video.resolution_scale_percent != 100)
{
auto src_width = rsx::apply_resolution_scale(old_contents.width, true, old_contents.source->width());
auto src_height = rsx::apply_resolution_scale(old_contents.height, true, old_contents.source->height());
auto dst_width = rsx::apply_resolution_scale(old_contents.width, true, old_contents.target->width());
auto dst_height = rsx::apply_resolution_scale(old_contents.height, true, old_contents.target->height());
old_contents.transfer_scale_x *= f32(dst_width) / src_width;
old_contents.transfer_scale_y *= f32(dst_height) / src_height;
old_contents.width = src_width;
old_contents.height = src_height;
old_contents.src_x = rsx::apply_resolution_scale(old_contents.src_x, false, old_contents.source->width());
old_contents.src_y = rsx::apply_resolution_scale(old_contents.src_y, false, old_contents.source->height());
old_contents.dst_x = rsx::apply_resolution_scale(old_contents.dst_x, false, old_contents.target->width());
old_contents.dst_y = rsx::apply_resolution_scale(old_contents.dst_y, false, old_contents.target->height());
}
}
void queue_tag(u32 address)
{
for (int i = 0; i < memory_tag_samples.size(); ++i)
{
if (LIKELY(i))
memory_tag_samples[i].first = 0;
else
memory_tag_samples[i].first = address; // Top left
}
const u32 pitch = get_native_pitch();
if (UNLIKELY(pitch < 16))
{
// Not enough area to gather samples if pitch is too small
return;
}
// Top right corner
memory_tag_samples[1].first = address + pitch - 8;
if (const u32 h = get_surface_height(); h > 1)
{
// Last row
const u32 pitch2 = get_rsx_pitch();
const u32 last_row_offset = pitch2 * (h - 1);
memory_tag_samples[2].first = address + last_row_offset; // Bottom left corner
memory_tag_samples[3].first = address + last_row_offset + pitch - 8; // Bottom right corner
// Centroid
const u32 center_row_offset = pitch2 * (h / 2);
memory_tag_samples[4].first = address + center_row_offset + pitch / 2;
}
}
void sync_tag()
{
for (auto &tag : memory_tag_samples)
{
if (!tag.first)
break;
tag.second = *reinterpret_cast<u64*>(vm::g_sudo_addr + tag.first);
}
}
void on_write(u64 write_tag = 0)
{
if (write_tag)
{
// Update use tag if requested
last_use_tag = write_tag;
}
// Tag unconditionally without introducing new data
sync_tag();
read_aa_mode = write_aa_mode;
// HACK!! This should be cleared through memory barriers only
state_flags = rsx::surface_state_flags::ready;
if (old_contents.source)
{
clear_rw_barrier();
}
}
// Returns the rect area occupied by this surface expressed as an 8bpp image with no AA
areau get_normalized_memory_area() const
{
const u16 internal_width = get_native_pitch() * (write_aa_mode > rsx::surface_antialiasing::center_1_sample? 2: 1);
const u16 internal_height = get_surface_height() * (write_aa_mode > rsx::surface_antialiasing::diagonal_centered_2_samples? 2: 1);
return { 0, 0, internal_width, internal_height };
}
rsx::address_range get_memory_range() const
{
const u32 internal_height = get_surface_height() * (write_aa_mode > rsx::surface_antialiasing::diagonal_centered_2_samples? 2: 1);
return rsx::address_range::start_length(memory_tag_samples[0].first, internal_height * get_rsx_pitch());
}
};
/**
* Helper for surface (ie color and depth stencil render target) management.
* It handles surface creation and storage. Backend should only retrieve pointer to surface.
* It provides 2 methods get_texture_from_*_if_applicable that should be used when an app
* wants to sample a previous surface.
* Please note that the backend is still responsible for creating framebuffer/descriptors
* and need to inform surface_store everytime surface format/size/addresses change.
*
* Since it's a template it requires a trait with the followings:
* - type surface_storage_type which is a structure containing texture.
* - type surface_type which is a pointer to storage_type or a reference.
* - type command_list_type that can be void for backend without command list
* - type download_buffer_object used by issue_download_command and map_downloaded_buffer functions to handle sync
*
* - a member function static surface_type(const surface_storage_type&) that returns underlying surface pointer from a storage type.
* - 2 member functions static surface_storage_type create_new_surface(u32 address, Surface_color_format/Surface_depth_format format, size_t width, size_t height,...)
* used to create a new surface_storage_type holding surface from passed parameters.
* - a member function static prepare_rtt_for_drawing(command_list, surface_type) that makes a sampleable surface a color render target one.
* - a member function static prepare_rtt_for_drawing(command_list, surface_type) that makes a render target surface a sampleable one.
* - a member function static prepare_ds_for_drawing that does the same for depth stencil surface.
* - a member function static prepare_ds_for_sampling that does the same for depth stencil surface.
* - a member function static bool rtt_has_format_width_height(const surface_storage_type&, Surface_color_format surface_color_format, size_t width, size_t height)
* that checks if the given surface has the given format and size
* - a member function static bool ds_has_format_width_height that does the same for ds
* - a member function static download_buffer_object issue_download_command(surface_type, Surface_color_format color_format, size_t width, size_t height,...)
* that generates command to download the given surface to some mappable buffer.
* - a member function static issue_depth_download_command that does the same for depth surface
* - a member function static issue_stencil_download_command that does the same for stencil surface
* - a member function gsl::span<const gsl::byte> map_downloaded_buffer(download_buffer_object, ...) that maps a download_buffer_object
* - a member function static unmap_downloaded_buffer that unmaps it.
*/
template<typename Traits>
struct surface_store
{
template<typename T, typename U>
void copy_pitched_src_to_dst(gsl::span<T> dest, gsl::span<const U> src, size_t src_pitch_in_bytes, size_t width, size_t height)
{
for (int row = 0; row < height; row++)
{
for (unsigned col = 0; col < width; col++)
dest[col] = src[col];
src = src.subspan(src_pitch_in_bytes / sizeof(U));
dest = dest.subspan(width);
}
}
constexpr u32 get_aa_factor_u(surface_antialiasing aa_mode)
{
return (aa_mode == surface_antialiasing::center_1_sample)? 1 : 2;
}
constexpr u32 get_aa_factor_v(surface_antialiasing aa_mode)
{
switch (aa_mode)
{
case surface_antialiasing::center_1_sample:
case surface_antialiasing::diagonal_centered_2_samples:
return 1;
default:
return 2;
}
}
public:
using surface_storage_type = typename Traits::surface_storage_type;
using surface_type = typename Traits::surface_type;
using command_list_type = typename Traits::command_list_type;
using download_buffer_object = typename Traits::download_buffer_object;
using surface_overlap_info = surface_overlap_info_t<surface_type>;
protected:
std::unordered_map<u32, surface_storage_type> m_render_targets_storage = {};
std::unordered_map<u32, surface_storage_type> m_depth_stencil_storage = {};
rsx::address_range m_render_targets_memory_range;
rsx::address_range m_depth_stencil_memory_range;
public:
std::pair<u8, u8> m_bound_render_targets_config = {};
std::array<std::pair<u32, surface_type>, 4> m_bound_render_targets = {};
std::pair<u32, surface_type> m_bound_depth_stencil = {};
// List of sections derived from a section that has been split and invalidated
std::vector<surface_type> orphaned_surfaces;
std::list<surface_storage_type> invalidated_resources;
u64 cache_tag = 1ull; // Use 1 as the start since 0 is default tag on new surfaces
u64 write_tag = 1ull;
surface_store() = default;
~surface_store() = default;
surface_store(const surface_store&) = delete;
private:
template <bool is_depth_surface>
void split_surface_region(command_list_type cmd, u32 address, surface_type prev_surface, u16 width, u16 height, u8 bpp, rsx::surface_antialiasing aa)
{
#ifndef INCOMPLETE_SURFACE_CACHE_IMPL
auto insert_new_surface = [&](
u32 new_address,
deferred_clipped_region<surface_type>& region,
std::unordered_map<u32, surface_storage_type>& data)
{
verify(HERE), prev_surface;
if (prev_surface->read_barrier(cmd); !prev_surface->test())
{
return;
}
surface_storage_type sink;
if (const auto found = data.find(new_address);
found != data.end())
{
if (Traits::is_compatible_surface(Traits::get(found->second), region.source, region.width, region.height, 1))
{
// There is no need to erase due to the reinsertion below
sink = std::move(found->second);
}
else
{
// TODO: Merge the 2 regions
invalidated_resources.push_back(std::move(found->second));
data.erase(new_address);
auto &old = invalidated_resources.back();
Traits::notify_surface_invalidated(old);
}
}
Traits::clone_surface(cmd, sink, region.source, new_address, region);
verify(HERE), region.target == Traits::get(sink);
orphaned_surfaces.push_back(region.target);
data[new_address] = std::move(sink);
};
// Define incoming region
size2u old, _new;
const auto prev_area = prev_surface->get_normalized_memory_area();
old.width = prev_area.x2;
old.height = prev_area.y2;
_new.width = width * bpp * get_aa_factor_u(aa);
_new.height = height * get_aa_factor_v(aa);
if (old.width > _new.width)
{
// Split in X
const u32 baseaddr = address + _new.width;
const u32 bytes_to_texels_x = (bpp * get_aa_factor_u(prev_surface->write_aa_mode));
deferred_clipped_region<surface_type> copy;
copy.src_x = _new.width / bytes_to_texels_x;
copy.src_y = 0;
copy.dst_x = 0;
copy.dst_y = 0;
copy.width = (old.width - _new.width) / bytes_to_texels_x;
copy.height = prev_surface->get_surface_height();
copy.transfer_scale_x = 1.f;
copy.transfer_scale_y = 1.f;
copy.target = nullptr;
copy.source = prev_surface;
if constexpr (is_depth_surface)
{
insert_new_surface(baseaddr, copy, m_depth_stencil_storage);
}
else
{
insert_new_surface(baseaddr, copy, m_render_targets_storage);
}
}
if (old.height > _new.height)
{
// Split in Y
const u32 baseaddr = address + (_new.height * prev_surface->get_rsx_pitch());
const u32 bytes_to_texels_x = (bpp * get_aa_factor_u(prev_surface->write_aa_mode));
deferred_clipped_region<surface_type> copy;
copy.src_x = 0;
copy.src_y = _new.height / get_aa_factor_v(prev_surface->write_aa_mode);
copy.dst_x = 0;
copy.dst_y = 0;
copy.width = std::min(_new.width, old.width) / bytes_to_texels_x;
copy.height = (old.height - _new.height) / get_aa_factor_v(prev_surface->write_aa_mode);
copy.transfer_scale_x = 1.f;
copy.transfer_scale_y = 1.f;
copy.target = nullptr;
copy.source = prev_surface;
if constexpr (is_depth_surface)
{
insert_new_surface(baseaddr, copy, m_depth_stencil_storage);
}
else
{
insert_new_surface(baseaddr, copy, m_render_targets_storage);
}
}
#endif
}
template <bool is_depth_surface>
void intersect_surface_region(command_list_type cmd, u32 address, surface_type new_surface, surface_type prev_surface)
{
#ifndef INCOMPLETE_SURFACE_CACHE_IMPL
auto scan_list = [&new_surface, address](const rsx::address_range& mem_range, u64 timestamp_check,
std::unordered_map<u32, surface_storage_type>& data) -> std::vector<std::pair<u32, surface_type>>
{
std::vector<std::pair<u32, surface_type>> result;
for (const auto &e : data)
{
auto surface = Traits::get(e.second);
if (e.second->last_use_tag <= timestamp_check ||
new_surface == surface ||
address == e.first ||
e.second->dirty())
{
// Do not bother synchronizing with uninitialized data
continue;
}
// Memory partition check
if (mem_range.start >= 0xc0000000)
{
if (e.first < 0xc0000000) continue;
}
else
{
if (e.first >= 0xc0000000) continue;
}
// Pitch check
if (!rsx::pitch_compatible(surface, new_surface))
{
continue;
}
// Range check
const rsx::address_range this_range = surface->get_memory_range();
if (!this_range.overlaps(mem_range))
{
continue;
}
result.push_back({ e.first, surface });
}
return result;
};
const rsx::address_range mem_range = new_surface->get_memory_range();
const u64 timestamp_check = prev_surface ? prev_surface->last_use_tag : new_surface->last_use_tag;
auto list1 = scan_list(mem_range, timestamp_check, m_render_targets_storage);
auto list2 = scan_list(mem_range, timestamp_check, m_depth_stencil_storage);
if (prev_surface)
{
// Append the previous removed surface to the intersection list
std::pair<u32, surface_type> e = { address, prev_surface };
if constexpr (is_depth_surface)
{
list2.push_back({ address, prev_surface });
}
else
{
list1.push_back({ address, prev_surface });
}
}
else
{
if (list1.empty() && list2.empty())
{
return;
}
}
std::vector<std::pair<u32, surface_type>> surface_info;
if (list1.empty())
{
surface_info = std::move(list2);
}
else if (list2.empty())
{
surface_info = std::move(list1);
}
else
{
surface_info = std::move(list1);
surface_info.reserve(list1.size() + list2.size());
for (const auto& e : list2) surface_info.push_back(e);
}
if (UNLIKELY(surface_info.size() > 1))
{
// Sort with newest first for early exit
std::sort(surface_info.begin(), surface_info.end(), [](const auto& a, const auto& b)
{
return (a.second->last_use_tag > b.second->last_use_tag);
});
}
// TODO: Modify deferred_clip_region::direct_copy() to take a few more things into account!
const areau child_region = new_surface->get_normalized_memory_area();
const auto child_w = child_region.width();
const auto child_h = child_region.height();
const auto pitch = new_surface->get_rsx_pitch();
for (const auto &e: surface_info)
{
const auto parent_region = e.second->get_normalized_memory_area();
const auto parent_w = parent_region.width();
const auto parent_h = parent_region.height();
const auto rect = rsx::intersect_region(e.first, parent_w, parent_h, 1, address, child_w, child_h, 1, pitch);
const auto src_offset = std::get<0>(rect);
const auto dst_offset = std::get<1>(rect);
const auto size = std::get<2>(rect);
if (src_offset.x >= parent_w || src_offset.y >= parent_h)
{
continue;
}
if (dst_offset.x >= child_w || dst_offset.y >= child_h)
{
continue;
}
// TODO: Eventually need to stack all the overlapping regions, but for now just do the latest rect in the space
deferred_clipped_region<surface_type> region;
region.src_x = src_offset.x;
region.src_y = src_offset.y;
region.dst_x = dst_offset.x;
region.dst_y = dst_offset.y;
region.width = size.width;
region.height = size.height;
region.source = e.second;
region.target = new_surface;
new_surface->set_old_contents_region(region, true);
break;
}
#endif
}
protected:
/**
* If render target already exists at address, issue state change operation on cmdList.
* Otherwise create one with width, height, clearColor info.
* returns the corresponding render target resource.
*/
template <typename ...Args>
surface_type bind_address_as_render_targets(
command_list_type command_list,
u32 address,
surface_color_format color_format,
surface_antialiasing antialias,
size_t width, size_t height, size_t pitch,
Args&&... extra_params)
{
// TODO: Fix corner cases
// This doesn't take overlapping surface(s) into account.
surface_storage_type old_surface_storage;
surface_storage_type new_surface_storage;
surface_type old_surface = nullptr;
surface_type new_surface = nullptr;
bool store = true;
// Check if render target already exists
auto It = m_render_targets_storage.find(address);
if (It != m_render_targets_storage.end())
{
surface_storage_type &rtt = It->second;
const bool pitch_compatible = Traits::surface_is_pitch_compatible(rtt, pitch);
if (pitch_compatible)
{
// Preserve memory outside the area to be inherited if needed
const u8 bpp = get_format_block_size_in_bytes(color_format);
split_surface_region<false>(command_list, address, Traits::get(rtt), (u16)width, (u16)height, bpp, antialias);
}
if (Traits::rtt_has_format_width_height(rtt, color_format, width, height))
{
if (pitch_compatible)
Traits::notify_surface_persist(rtt);
else
Traits::invalidate_surface_contents(command_list, Traits::get(rtt), address, pitch);
Traits::prepare_rtt_for_drawing(command_list, Traits::get(rtt));
new_surface = Traits::get(rtt);
store = false;
}
else
{
old_surface = Traits::get(rtt);
old_surface_storage = std::move(rtt);
m_render_targets_storage.erase(address);
}
}
if (!new_surface)
{
// Range test
const auto aa_factor_v = get_aa_factor_v(antialias);
rsx::address_range range = rsx::address_range::start_length(address, u32(pitch * height * aa_factor_v));
m_render_targets_memory_range = range.get_min_max(m_render_targets_memory_range);
// Search invalidated resources for a suitable surface
for (auto It = invalidated_resources.begin(); It != invalidated_resources.end(); It++)
{
auto &rtt = *It;
if (Traits::rtt_has_format_width_height(rtt, color_format, width, height, true))
{
new_surface_storage = std::move(rtt);
#ifndef INCOMPLETE_SURFACE_CACHE_IMPL
Traits::notify_surface_reused(new_surface_storage);
#endif
if (old_surface)
{
// Exchange this surface with the invalidated one
Traits::notify_surface_invalidated(old_surface_storage);
rtt = std::move(old_surface_storage);
}
else
{
// rtt is now empty - erase it
invalidated_resources.erase(It);
}
new_surface = Traits::get(new_surface_storage);
Traits::invalidate_surface_contents(command_list, new_surface, address, pitch);
Traits::prepare_rtt_for_drawing(command_list, new_surface);
break;
}
}
}
// Check for stale storage
if (old_surface != nullptr && new_surface == nullptr)
{
// This was already determined to be invalid and is excluded from testing above
Traits::notify_surface_invalidated(old_surface_storage);
invalidated_resources.push_back(std::move(old_surface_storage));
}
if (!new_surface)
{
verify(HERE), store;
new_surface_storage = Traits::create_new_surface(address, color_format, width, height, pitch, std::forward<Args>(extra_params)...);
new_surface = Traits::get(new_surface_storage);
}
if (!old_surface)
{
// Remove and preserve if possible any overlapping/replaced depth surface
auto aliased_depth_surface = m_depth_stencil_storage.find(address);
if (aliased_depth_surface != m_depth_stencil_storage.end())
{
old_surface = Traits::get(aliased_depth_surface->second);
Traits::notify_surface_invalidated(aliased_depth_surface->second);
invalidated_resources.push_back(std::move(aliased_depth_surface->second));
m_depth_stencil_storage.erase(aliased_depth_surface);
}
}
#ifndef INCOMPLETE_SURFACE_CACHE_IMPL
// TODO: This can be done better after refactoring
new_surface->write_aa_mode = antialias;
// Check if old_surface is 'new' and avoid intersection
if (old_surface && old_surface->last_use_tag >= write_tag)
{
new_surface->set_old_contents(old_surface);
}
else
#endif
{
intersect_surface_region<false>(command_list, address, new_surface, old_surface);
}
if (store)
{
// New surface was found among invalidated surfaces
m_render_targets_storage[address] = std::move(new_surface_storage);
}
return new_surface;
}
template <typename ...Args>
surface_type bind_address_as_depth_stencil(
command_list_type command_list,
u32 address,
surface_depth_format depth_format,
surface_antialiasing antialias,
size_t width, size_t height, size_t pitch,
Args&&... extra_params)
{
surface_storage_type old_surface_storage;
surface_storage_type new_surface_storage;
surface_type old_surface = nullptr;
surface_type new_surface = nullptr;
bool store = true;
auto It = m_depth_stencil_storage.find(address);
if (It != m_depth_stencil_storage.end())
{
surface_storage_type &ds = It->second;
const bool pitch_compatible = Traits::surface_is_pitch_compatible(ds, pitch);
if (pitch_compatible)
{
const u8 bpp = (depth_format == rsx::surface_depth_format::z16)? 2 : 4;
split_surface_region<true>(command_list, address, Traits::get(ds), (u16)width, (u16)height, bpp, antialias);
}
if (Traits::ds_has_format_width_height(ds, depth_format, width, height))
{
if (pitch_compatible)
Traits::notify_surface_persist(ds);
else
Traits::invalidate_surface_contents(command_list, Traits::get(ds), address, pitch);
Traits::prepare_ds_for_drawing(command_list, Traits::get(ds));
new_surface = Traits::get(ds);
store = false;
}
else
{
old_surface = Traits::get(ds);
old_surface_storage = std::move(ds);
m_depth_stencil_storage.erase(address);
}
}
if (!new_surface)
{
// Range test
const auto aa_factor_v = get_aa_factor_v(antialias);
rsx::address_range range = rsx::address_range::start_length(address, u32(pitch * height * aa_factor_v));
m_depth_stencil_memory_range = range.get_min_max(m_depth_stencil_memory_range);
//Search invalidated resources for a suitable surface
for (auto It = invalidated_resources.begin(); It != invalidated_resources.end(); It++)
{
auto &ds = *It;
if (Traits::ds_has_format_width_height(ds, depth_format, width, height, true))
{
new_surface_storage = std::move(ds);
#ifndef INCOMPLETE_SURFACE_CACHE_IMPL
Traits::notify_surface_reused(new_surface_storage);
#endif
if (old_surface)
{
//Exchange this surface with the invalidated one
Traits::notify_surface_invalidated(old_surface_storage);
ds = std::move(old_surface_storage);
}
else
invalidated_resources.erase(It);
new_surface = Traits::get(new_surface_storage);
Traits::prepare_ds_for_drawing(command_list, new_surface);
Traits::invalidate_surface_contents(command_list, new_surface, address, pitch);
break;
}
}
}
if (old_surface != nullptr && new_surface == nullptr)
{
// This was already determined to be invalid and is excluded from testing above
Traits::notify_surface_invalidated(old_surface_storage);
invalidated_resources.push_back(std::move(old_surface_storage));
}
if (!new_surface)
{
verify(HERE), store;
new_surface_storage = Traits::create_new_surface(address, depth_format, width, height, pitch, std::forward<Args>(extra_params)...);
new_surface = Traits::get(new_surface_storage);
}
if (!old_surface)
{
// Remove and preserve if possible any overlapping/replaced color surface
auto aliased_rtt_surface = m_render_targets_storage.find(address);
if (aliased_rtt_surface != m_render_targets_storage.end())
{
old_surface = Traits::get(aliased_rtt_surface->second);
Traits::notify_surface_invalidated(aliased_rtt_surface->second);
invalidated_resources.push_back(std::move(aliased_rtt_surface->second));
m_render_targets_storage.erase(aliased_rtt_surface);
}
}
#ifndef INCOMPLETE_SURFACE_CACHE_IMPL
// TODO: Forward this to the management functions
new_surface->write_aa_mode = antialias;
// Check if old_surface is 'new' and avoid intersection
if (old_surface && old_surface->last_use_tag >= write_tag)
{
new_surface->set_old_contents(old_surface);
}
else
#endif
{
intersect_surface_region<true>(command_list, address, new_surface, old_surface);
}
if (store)
{
// New surface was found among invalidated surfaces
m_depth_stencil_storage[address] = std::move(new_surface_storage);
}
return new_surface;
}
public:
/**
* Update bound color and depth surface.
* Must be called everytime surface format, clip, or addresses changes.
*/
template <typename ...Args>
void prepare_render_target(
command_list_type command_list,
surface_color_format color_format, surface_depth_format depth_format,
u32 clip_horizontal_reg, u32 clip_vertical_reg,
surface_target set_surface_target,
surface_antialiasing antialias,
const std::array<u32, 4> &surface_addresses, u32 address_z,
const std::array<u32, 4> &surface_pitch, u32 zeta_pitch,
Args&&... extra_params)
{
u32 clip_width = clip_horizontal_reg;
u32 clip_height = clip_vertical_reg;
// u32 clip_x = clip_horizontal_reg;
// u32 clip_y = clip_vertical_reg;
cache_tag = rsx::get_shared_tag();
// Make previous RTTs sampleable
for (int i = m_bound_render_targets_config.first, count = 0;
count < m_bound_render_targets_config.second;
++i, ++count)
{
auto &rtt = m_bound_render_targets[i];
Traits::prepare_rtt_for_sampling(command_list, std::get<1>(rtt));
rtt = std::make_pair(0, nullptr);
}
const auto rtt_indices = utility::get_rtt_indexes(set_surface_target);
if (LIKELY(!rtt_indices.empty()))
{
m_bound_render_targets_config = { rtt_indices.front(), 0 };
// Create/Reuse requested rtts
for (u8 surface_index : rtt_indices)
{
if (surface_addresses[surface_index] == 0)
continue;
m_bound_render_targets[surface_index] = std::make_pair(surface_addresses[surface_index],
bind_address_as_render_targets(command_list, surface_addresses[surface_index], color_format, antialias,
clip_width, clip_height, surface_pitch[surface_index], std::forward<Args>(extra_params)...));
m_bound_render_targets_config.second++;
}
}
else
{
m_bound_render_targets_config = { 0, 0 };
}
// Same for depth buffer
if (std::get<1>(m_bound_depth_stencil) != nullptr)
Traits::prepare_ds_for_sampling(command_list, std::get<1>(m_bound_depth_stencil));
m_bound_depth_stencil = std::make_pair(0, nullptr);
if (!address_z)
return;
m_bound_depth_stencil = std::make_pair(address_z,
bind_address_as_depth_stencil(command_list, address_z, depth_format, antialias,
clip_width, clip_height, zeta_pitch, std::forward<Args>(extra_params)...));
}
/**
* Search for given address in stored color surface
* Return an empty surface_type otherwise.
*/
surface_type get_texture_from_render_target_if_applicable(u32 address)
{
auto It = m_render_targets_storage.find(address);
if (It != m_render_targets_storage.end())
return Traits::get(It->second);
return surface_type();
}
/**
* Search for given address in stored depth stencil surface
* Return an empty surface_type otherwise.
*/
surface_type get_texture_from_depth_stencil_if_applicable(u32 address)
{
auto It = m_depth_stencil_storage.find(address);
if (It != m_depth_stencil_storage.end())
return Traits::get(It->second);
return surface_type();
}
surface_type get_surface_at(u32 address)
{
auto It = m_render_targets_storage.find(address);
if (It != m_render_targets_storage.end())
return Traits::get(It->second);
auto _It = m_depth_stencil_storage.find(address);
if (_It != m_depth_stencil_storage.end())
return Traits::get(_It->second);
fmt::throw_exception("Unreachable" HERE);
}
/**
* Get bound color surface raw data.
*/
template <typename... Args>
std::array<std::vector<gsl::byte>, 4> get_render_targets_data(
surface_color_format color_format, size_t width, size_t height,
Args&& ...args
)
{
std::array<download_buffer_object, 4> download_data = {};
// Issue download commands
for (int i = 0; i < 4; i++)
{
if (std::get<0>(m_bound_render_targets[i]) == 0)
continue;
surface_type surface_resource = std::get<1>(m_bound_render_targets[i]);
download_data[i] = std::move(
Traits::issue_download_command(surface_resource, color_format, width, height, std::forward<Args&&>(args)...)
);
}
std::array<std::vector<gsl::byte>, 4> result = {};
// Sync and copy data
for (int i = 0; i < 4; i++)
{
if (std::get<0>(m_bound_render_targets[i]) == 0)
continue;
gsl::span<const gsl::byte> raw_src = Traits::map_downloaded_buffer(download_data[i], std::forward<Args&&>(args)...);
size_t src_pitch = utility::get_aligned_pitch(color_format, ::narrow<u32>(width));
size_t dst_pitch = utility::get_packed_pitch(color_format, ::narrow<u32>(width));
result[i].resize(dst_pitch * height);
// Note: MSVC + GSL doesn't support span<byte> -> span<T> for non const span atm
// thus manual conversion
switch (color_format)
{
case surface_color_format::a8b8g8r8:
case surface_color_format::x8b8g8r8_o8b8g8r8:
case surface_color_format::x8b8g8r8_z8b8g8r8:
case surface_color_format::a8r8g8b8:
case surface_color_format::x8r8g8b8_o8r8g8b8:
case surface_color_format::x8r8g8b8_z8r8g8b8:
case surface_color_format::x32:
{
gsl::span<be_t<u32>> dst_span{ (be_t<u32>*)result[i].data(), ::narrow<int>(dst_pitch * height / sizeof(be_t<u32>)) };
copy_pitched_src_to_dst(dst_span, as_const_span<const u32>(raw_src), src_pitch, width, height);
break;
}
case surface_color_format::b8:
{
gsl::span<u8> dst_span{ (u8*)result[i].data(), ::narrow<int>(dst_pitch * height / sizeof(u8)) };
copy_pitched_src_to_dst(dst_span, as_const_span<const u8>(raw_src), src_pitch, width, height);
break;
}
case surface_color_format::g8b8:
case surface_color_format::r5g6b5:
case surface_color_format::x1r5g5b5_o1r5g5b5:
case surface_color_format::x1r5g5b5_z1r5g5b5:
{
gsl::span<be_t<u16>> dst_span{ (be_t<u16>*)result[i].data(), ::narrow<int>(dst_pitch * height / sizeof(be_t<u16>)) };
copy_pitched_src_to_dst(dst_span, as_const_span<const u16>(raw_src), src_pitch, width, height);
break;
}
// Note : may require some big endian swap
case surface_color_format::w32z32y32x32:
{
gsl::span<u128> dst_span{ (u128*)result[i].data(), ::narrow<int>(dst_pitch * height / sizeof(u128)) };
copy_pitched_src_to_dst(dst_span, as_const_span<const u128>(raw_src), src_pitch, width, height);
break;
}
case surface_color_format::w16z16y16x16:
{
gsl::span<u64> dst_span{ (u64*)result[i].data(), ::narrow<int>(dst_pitch * height / sizeof(u64)) };
copy_pitched_src_to_dst(dst_span, as_const_span<const u64>(raw_src), src_pitch, width, height);
break;
}
}
Traits::unmap_downloaded_buffer(download_data[i], std::forward<Args&&>(args)...);
}
return result;
}
/**
* Get bound color surface raw data.
*/
template <typename... Args>
std::array<std::vector<gsl::byte>, 2> get_depth_stencil_data(
surface_depth_format depth_format, size_t width, size_t height,
Args&& ...args
)
{
std::array<std::vector<gsl::byte>, 2> result = {};
if (std::get<0>(m_bound_depth_stencil) == 0)
return result;
size_t row_pitch = align(width * 4, 256);
download_buffer_object stencil_data = {};
download_buffer_object depth_data = Traits::issue_depth_download_command(std::get<1>(m_bound_depth_stencil), depth_format, width, height, std::forward<Args&&>(args)...);
if (depth_format == surface_depth_format::z24s8)
stencil_data = std::move(Traits::issue_stencil_download_command(std::get<1>(m_bound_depth_stencil), width, height, std::forward<Args&&>(args)...));
gsl::span<const gsl::byte> depth_buffer_raw_src = Traits::map_downloaded_buffer(depth_data, std::forward<Args&&>(args)...);
if (depth_format == surface_depth_format::z16)
{
result[0].resize(width * height * 2);
gsl::span<u16> dest{ (u16*)result[0].data(), ::narrow<int>(width * height) };
copy_pitched_src_to_dst(dest, as_const_span<const u16>(depth_buffer_raw_src), row_pitch, width, height);
}
if (depth_format == surface_depth_format::z24s8)
{
result[0].resize(width * height * 4);
gsl::span<u32> dest{ (u32*)result[0].data(), ::narrow<int>(width * height) };
copy_pitched_src_to_dst(dest, as_const_span<const u32>(depth_buffer_raw_src), row_pitch, width, height);
}
Traits::unmap_downloaded_buffer(depth_data, std::forward<Args&&>(args)...);
if (depth_format == surface_depth_format::z16)
return result;
gsl::span<const gsl::byte> stencil_buffer_raw_src = Traits::map_downloaded_buffer(stencil_data, std::forward<Args&&>(args)...);
result[1].resize(width * height);
gsl::span<u8> dest{ (u8*)result[1].data(), ::narrow<int>(width * height) };
copy_pitched_src_to_dst(dest, as_const_span<const u8>(stencil_buffer_raw_src), align(width, 256), width, height);
Traits::unmap_downloaded_buffer(stencil_data, std::forward<Args&&>(args)...);
return result;
}
/**
* Moves a single surface from surface storage to invalidated surface store.
* Can be triggered by the texture cache's blit functionality when formats do not match
*/
void invalidate_single_surface(surface_type surface, bool depth)
{
if (!depth)
{
for (auto It = m_render_targets_storage.begin(); It != m_render_targets_storage.end(); It++)
{
const auto address = It->first;
const auto ref = Traits::get(It->second);
if (surface == ref)
{
Traits::notify_surface_invalidated(It->second);
invalidated_resources.push_back(std::move(It->second));
m_render_targets_storage.erase(It);
cache_tag = rsx::get_shared_tag();
return;
}
}
}
else
{
for (auto It = m_depth_stencil_storage.begin(); It != m_depth_stencil_storage.end(); It++)
{
const auto address = It->first;
const auto ref = Traits::get(It->second);
if (surface == ref)
{
Traits::notify_surface_invalidated(It->second);
invalidated_resources.push_back(std::move(It->second));
m_depth_stencil_storage.erase(It);
cache_tag = rsx::get_shared_tag();
return;
}
}
}
}
/**
* Invalidates surface that exists at an address
*/
void invalidate_surface_address(u32 addr, bool depth)
{
if (address_is_bound(addr))
{
LOG_ERROR(RSX, "Cannot invalidate a currently bound render target!");
return;
}
if (!depth)
{
auto It = m_render_targets_storage.find(addr);
if (It != m_render_targets_storage.end())
{
Traits::notify_surface_invalidated(It->second);
invalidated_resources.push_back(std::move(It->second));
m_render_targets_storage.erase(It);
cache_tag = rsx::get_shared_tag();
return;
}
}
else
{
auto It = m_depth_stencil_storage.find(addr);
if (It != m_depth_stencil_storage.end())
{
Traits::notify_surface_invalidated(It->second);
invalidated_resources.push_back(std::move(It->second));
m_depth_stencil_storage.erase(It);
cache_tag = rsx::get_shared_tag();
return;
}
}
}
bool address_is_bound(u32 address) const
{
for (int i = m_bound_render_targets_config.first, count = 0;
count < m_bound_render_targets_config.second;
++i, ++count)
{
if (m_bound_render_targets[i].first == address)
return true;
}
return (m_bound_depth_stencil.first == address);
}
template <typename commandbuffer_type>
std::vector<surface_overlap_info> get_merged_texture_memory_region(commandbuffer_type& cmd, u32 texaddr, u32 required_width, u32 required_height, u32 required_pitch, u8 required_bpp)
{
std::vector<surface_overlap_info> result;
std::vector<std::pair<u32, bool>> dirty;
const u32 limit = texaddr + (required_pitch * required_height);
auto process_list_function = [&](std::unordered_map<u32, surface_storage_type>& data, bool is_depth)
{
for (auto &tex_info : data)
{
auto this_address = std::get<0>(tex_info);
if (this_address >= limit)
continue;
auto surface = std::get<1>(tex_info).get();
const auto pitch = surface->get_rsx_pitch();
if (!rsx::pitch_compatible(surface, required_pitch, required_height))
continue;
const u16 scale_x = surface->read_aa_mode > rsx::surface_antialiasing::center_1_sample? 2 : 1;
const u16 scale_y = surface->read_aa_mode > rsx::surface_antialiasing::diagonal_centered_2_samples? 2 : 1;
const auto texture_size = pitch * surface->get_surface_height() * scale_y;
if ((this_address + texture_size) <= texaddr)
continue;
if (surface->read_barrier(cmd); !surface->test())
{
dirty.emplace_back(this_address, is_depth);
continue;
}
surface_overlap_info info;
info.surface = surface;
info.base_address = this_address;
info.is_depth = is_depth;
surface_format_info surface_info{};
Traits::get_surface_info(surface, &surface_info);
const auto normalized_surface_width = (surface_info.surface_width * scale_x * surface_info.bpp) / required_bpp;
const auto normalized_surface_height = surface_info.surface_height * scale_y;
if (LIKELY(this_address >= texaddr))
{
const auto offset = this_address - texaddr;
info.dst_y = (offset / required_pitch);
info.dst_x = (offset % required_pitch) / required_bpp;
if (UNLIKELY(info.dst_x >= required_width || info.dst_y >= required_height))
{
// Out of bounds
continue;
}
info.src_x = 0;
info.src_y = 0;
info.width = std::min<u32>(normalized_surface_width, required_width - info.dst_x);
info.height = std::min<u32>(normalized_surface_height, required_height - info.dst_y);
}
else
{
const auto offset = texaddr - this_address;
info.src_y = (offset / required_pitch);
info.src_x = (offset % required_pitch) / required_bpp;
if (UNLIKELY(info.src_x >= normalized_surface_width || info.src_y >= normalized_surface_height))
{
// Region lies outside the actual texture area, but inside the 'tile'
// In this case, a small region lies to the top-left corner, partially occupying the target
continue;
}
info.dst_x = 0;
info.dst_y = 0;
info.width = std::min<u32>(required_width, normalized_surface_width - info.src_x);
info.height = std::min<u32>(required_height, normalized_surface_height - info.src_y);
}
info.is_clipped = (info.width < required_width || info.height < required_height);
if (UNLIKELY(surface_info.bpp != required_bpp))
{
// Width is calculated in the coordinate-space of the requester; normalize
info.src_x = (info.src_x * required_bpp) / surface_info.bpp;
info.width = (info.width * required_bpp) / surface_info.bpp;
}
if (UNLIKELY(scale_x > 1))
{
info.src_x /= scale_x;
info.dst_x /= scale_x;
info.width /= scale_x;
info.src_y /= scale_y;
info.dst_y /= scale_y;
info.height /= scale_y;
}
result.push_back(info);
}
};
// Range test helper to quickly discard blocks
// Fortunately, render targets tend to be clustered anyway
rsx::address_range test = rsx::address_range::start_end(texaddr, limit-1);
if (test.overlaps(m_render_targets_memory_range))
{
process_list_function(m_render_targets_storage, false);
}
if (test.overlaps(m_depth_stencil_memory_range))
{
process_list_function(m_depth_stencil_storage, true);
}
if (!dirty.empty())
{
for (const auto& p : dirty)
{
invalidate_surface_address(p.first, p.second);
}
}
if (result.size() > 1)
{
std::sort(result.begin(), result.end(), [](const auto &a, const auto &b)
{
if (a.surface->last_use_tag == b.surface->last_use_tag)
{
const auto area_a = a.width * a.height;
const auto area_b = b.width * b.height;
return area_a < area_b;
}
return a.surface->last_use_tag < b.surface->last_use_tag;
});
}
return result;
}
void on_write(u32 address = 0)
{
if (!address)
{
if (write_tag == cache_tag)
{
// Nothing to do
return;
}
else
{
write_tag = cache_tag;
}
// Tag all available surfaces
for (int i = m_bound_render_targets_config.first, count = 0;
count < m_bound_render_targets_config.second;
++i, ++count)
{
m_bound_render_targets[i].second->on_write(write_tag);
}
if (m_bound_depth_stencil.first)
{
m_bound_depth_stencil.second->on_write(write_tag);
}
}
else
{
for (int i = m_bound_render_targets_config.first, count = 0;
count < m_bound_render_targets_config.second;
++i, ++count)
{
if (m_bound_render_targets[i].first != address)
{
continue;
}
m_bound_render_targets[i].second->on_write(write_tag);
}
if (m_bound_depth_stencil.first == address)
{
m_bound_depth_stencil.second->on_write(write_tag);
}
}
}
void notify_memory_structure_changed()
{
cache_tag = rsx::get_shared_tag();
}
void invalidate_all()
{
// Unbind and invalidate all resources
auto free_resource_list = [&](auto &data)
{
for (auto &e : data)
{
Traits::notify_surface_invalidated(e.second);
invalidated_resources.push_back(std::move(e.second));
}
data.clear();
};
free_resource_list(m_render_targets_storage);
free_resource_list(m_depth_stencil_storage);
m_bound_depth_stencil = std::make_pair(0, nullptr);
m_bound_render_targets_config = { 0, 0 };
for (auto &rtt : m_bound_render_targets)
{
rtt = std::make_pair(0, nullptr);
}
}
};
}