void LoadCPReg(u32 sub_cmd, u32 value, bool is_preprocess)
{
bool update_global_state = !is_preprocess;
CPState* state = is_preprocess ? &g_preprocess_cp_state : &g_main_cp_state;
switch (sub_cmd & 0xF0)
{
case 0x30:
if (update_global_state)
VertexShaderManager::SetTexMatrixChangedA(value);
break;
case 0x40:
if (update_global_state)
VertexShaderManager::SetTexMatrixChangedB(value);
break;
case 0x50:
state->vtx_desc.Hex &= ~0x1FFFF; // keep the Upper bits
state->vtx_desc.Hex |= value;
state->attr_dirty = BitSet32::AllTrue(8);
state->bases_dirty = true;
break;
case 0x60:
state->vtx_desc.Hex &= 0x1FFFF; // keep the lower 17Bits
state->vtx_desc.Hex |= (u64)value << 17;
state->attr_dirty = BitSet32::AllTrue(8);
state->bases_dirty = true;
break;
case 0x70:
_assert_((sub_cmd & 0x0F) < 8);
state->vtx_attr[sub_cmd & 7].g0.Hex = value;
state->attr_dirty[sub_cmd & 7] = true;
break;
case 0x80:
_assert_((sub_cmd & 0x0F) < 8);
state->vtx_attr[sub_cmd & 7].g1.Hex = value;
state->attr_dirty[sub_cmd & 7] = true;
break;
case 0x90:
_assert_((sub_cmd & 0x0F) < 8);
state->vtx_attr[sub_cmd & 7].g2.Hex = value;
state->attr_dirty[sub_cmd & 7] = true;
break;
// Pointers to vertex arrays in GC RAM
case 0xA0:
state->array_bases[sub_cmd & 0xF] = value;
state->bases_dirty = true;
break;
case 0xB0:
state->array_strides[sub_cmd & 0xF] = value & 0xFF;
break;
}
}
void LoadCPReg(u32 sub_cmd, u32 value)
{
switch (sub_cmd & 0xF0)
{
case 0x30:
VertexShaderManager::SetTexMatrixChangedA(value);
break;
case 0x40:
VertexShaderManager::SetTexMatrixChangedB(value);
break;
case 0x50:
g_main_cp_state.vtx_desc.Hex &= ~0x1FFFF; // keep the Upper bits
g_main_cp_state.vtx_desc.Hex |= value;
g_main_cp_state.attr_dirty = 0xFF;
g_main_cp_state.bases_dirty = true;
break;
case 0x60:
g_main_cp_state.vtx_desc.Hex &= 0x1FFFF; // keep the lower 17Bits
g_main_cp_state.vtx_desc.Hex |= (u64)value << 17;
g_main_cp_state.attr_dirty = 0xFF;
g_main_cp_state.bases_dirty = true;
break;
case 0x70:
_assert_((sub_cmd & 0x0F) < 8);
g_main_cp_state.vtx_attr[sub_cmd & 7].g0.Hex = value;
g_main_cp_state.attr_dirty |= 1 << (sub_cmd & 7);
break;
case 0x80:
_assert_((sub_cmd & 0x0F) < 8);
g_main_cp_state.vtx_attr[sub_cmd & 7].g1.Hex = value;
g_main_cp_state.attr_dirty |= 1 << (sub_cmd & 7);
break;
case 0x90:
_assert_((sub_cmd & 0x0F) < 8);
g_main_cp_state.vtx_attr[sub_cmd & 7].g2.Hex = value;
g_main_cp_state.attr_dirty |= 1 << (sub_cmd & 7);
break;
// Pointers to vertex arrays in GC RAM
case 0xA0:
g_main_cp_state.array_bases[sub_cmd & 0xF] = value;
g_main_cp_state.bases_dirty = true;
break;
case 0xB0:
g_main_cp_state.array_strides[sub_cmd & 0xF] = value & 0xFF;
break;
}
}
void StagingTexture2D::WriteTexel(u32 x, u32 y, const void* data, size_t data_size)
{
_assert_(data_size >= m_texel_size);
VkDeviceSize offset = y * m_row_stride + x * m_texel_size;
VkDeviceSize map_offset = offset - m_map_offset;
_assert_(offset >= m_map_offset && (map_offset + m_texel_size) <= (m_map_offset + m_map_size));
char* ptr = m_map_pointer + map_offset;
memcpy(ptr, data, data_size);
}
void TextureConverter::ConvertTexture(TextureCacheBase::TCacheEntry* dst_entry,
TextureCacheBase::TCacheEntry* src_entry,
VkRenderPass render_pass, const void* palette,
TlutFormat palette_format)
{
struct PSUniformBlock
{
float multiplier;
int texel_buffer_offset;
int pad[2];
};
VKTexture* source_texture = static_cast<VKTexture*>(src_entry->texture.get());
VKTexture* destination_texture = static_cast<VKTexture*>(dst_entry->texture.get());
_assert_(static_cast<size_t>(palette_format) < NUM_PALETTE_CONVERSION_SHADERS);
_assert_(destination_texture->GetConfig().rendertarget);
// We want to align to 2 bytes (R16) or the device's texel buffer alignment, whichever is greater.
size_t palette_size = (src_entry->format & 0xF) == GX_TF_I4 ? 32 : 512;
if (!ReserveTexelBufferStorage(palette_size, sizeof(u16)))
return;
// Copy in palette to texel buffer.
u32 palette_offset = static_cast<u32>(m_texel_buffer->GetCurrentOffset());
memcpy(m_texel_buffer->GetCurrentHostPointer(), palette, palette_size);
m_texel_buffer->CommitMemory(palette_size);
VkCommandBuffer command_buffer = GetCommandBufferForTextureConversion(src_entry);
source_texture->GetRawTexIdentifier()->TransitionToLayout(
command_buffer, VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL);
destination_texture->GetRawTexIdentifier()->TransitionToLayout(
command_buffer, VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL);
// Bind and draw to the destination.
UtilityShaderDraw draw(command_buffer,
g_object_cache->GetPipelineLayout(PIPELINE_LAYOUT_TEXTURE_CONVERSION),
render_pass, g_object_cache->GetScreenQuadVertexShader(), VK_NULL_HANDLE,
m_palette_conversion_shaders[palette_format]);
VkRect2D region = {{0, 0}, {dst_entry->GetWidth(), dst_entry->GetHeight()}};
draw.BeginRenderPass(destination_texture->GetFramebuffer(), region);
PSUniformBlock uniforms = {};
uniforms.multiplier = (src_entry->format & 0xF) == GX_TF_I4 ? 15.0f : 255.0f;
uniforms.texel_buffer_offset = static_cast<int>(palette_offset / sizeof(u16));
draw.SetPushConstants(&uniforms, sizeof(uniforms));
draw.SetPSSampler(0, source_texture->GetRawTexIdentifier()->GetView(),
g_object_cache->GetPointSampler());
draw.SetPSTexelBuffer(m_texel_buffer_view_r16_uint);
draw.SetViewportAndScissor(0, 0, dst_entry->GetWidth(), dst_entry->GetHeight());
draw.DrawWithoutVertexBuffer(VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP, 4);
draw.EndRenderPass();
}
bool SwapChain::SetupSwapChainImages()
{
_assert_(m_swap_chain_images.empty());
uint32_t image_count;
VkResult res =
vkGetSwapchainImagesKHR(g_vulkan_context->GetDevice(), m_swap_chain, &image_count, nullptr);
if (res != VK_SUCCESS)
{
LOG_VULKAN_ERROR(res, "vkGetSwapchainImagesKHR failed: ");
return false;
}
std::vector<VkImage> images(image_count);
res = vkGetSwapchainImagesKHR(g_vulkan_context->GetDevice(), m_swap_chain, &image_count,
images.data());
_assert_(res == VK_SUCCESS);
m_swap_chain_images.reserve(image_count);
for (uint32_t i = 0; i < image_count; i++)
{
SwapChainImage image;
image.image = images[i];
// Create texture object, which creates a view of the backbuffer
image.texture = Texture2D::CreateFromExistingImage(
m_width, m_height, 1, 1, m_surface_format.format, VK_SAMPLE_COUNT_1_BIT,
VK_IMAGE_VIEW_TYPE_2D, image.image);
VkImageView view = image.texture->GetView();
VkFramebufferCreateInfo framebuffer_info = {VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO,
nullptr,
0,
m_render_pass,
1,
&view,
m_width,
m_height,
1};
res = vkCreateFramebuffer(g_vulkan_context->GetDevice(), &framebuffer_info, nullptr,
&image.framebuffer);
if (res != VK_SUCCESS)
{
LOG_VULKAN_ERROR(res, "vkCreateFramebuffer failed: ");
return false;
}
m_swap_chain_images.emplace_back(std::move(image));
}
return true;
}
bool VulkanRenderManager::CopyFramebufferToMemorySync(VKRFramebuffer *src, int aspectBits, int x, int y, int w, int h, Draw::DataFormat destFormat, uint8_t *pixels, int pixelStride) {
VKRStep *step = new VKRStep{ VKRStepType::READBACK };
step->readback.aspectMask = aspectBits;
step->readback.src = src;
step->readback.srcRect.offset = { x, y };
step->readback.srcRect.extent = { (uint32_t)w, (uint32_t)h };
steps_.push_back(step);
curRenderStep_ = nullptr;
FlushSync();
Draw::DataFormat srcFormat;
if (aspectBits & VK_IMAGE_ASPECT_COLOR_BIT) {
if (src) {
switch (src->color.format) {
case VK_FORMAT_R8G8B8A8_UNORM: srcFormat = Draw::DataFormat::R8G8B8A8_UNORM; break;
default: _assert_(false);
}
}
else {
// Backbuffer.
if (!(vulkan_->GetSurfaceCapabilities().supportedUsageFlags & VK_IMAGE_USAGE_TRANSFER_SRC_BIT)) {
ELOG("Copying from backbuffer not supported, can't take screenshots");
return false;
}
switch (vulkan_->GetSwapchainFormat()) {
case VK_FORMAT_B8G8R8A8_UNORM: srcFormat = Draw::DataFormat::B8G8R8A8_UNORM; break;
case VK_FORMAT_R8G8B8A8_UNORM: srcFormat = Draw::DataFormat::R8G8B8A8_UNORM; break;
// NOTE: If you add supported formats here, make sure to also support them in VulkanQueueRunner::CopyReadbackBuffer.
default:
ELOG("Unsupported backbuffer format for screenshots");
return false;
}
}
} else if (aspectBits & VK_IMAGE_ASPECT_STENCIL_BIT) {
// Copies from stencil are always S8.
srcFormat = Draw::DataFormat::S8;
} else if (aspectBits & VK_IMAGE_ASPECT_DEPTH_BIT) {
switch (src->depth.format) {
case VK_FORMAT_D24_UNORM_S8_UINT: srcFormat = Draw::DataFormat::D24_S8; break;
case VK_FORMAT_D32_SFLOAT_S8_UINT: srcFormat = Draw::DataFormat::D32F; break;
case VK_FORMAT_D16_UNORM_S8_UINT: srcFormat = Draw::DataFormat::D16; break;
default: _assert_(false);
}
} else {
_assert_(false);
}
// Need to call this after FlushSync so the pixels are guaranteed to be ready in CPU-accessible VRAM.
queueRunner_.CopyReadbackBuffer(w, h, srcFormat, destFormat, pixelStride, pixels);
return true;
}
void VulkanRenderManager::StopThread() {
if (useThread_ && run_) {
run_ = false;
// Stop the thread.
for (int i = 0; i < vulkan_->GetInflightFrames(); i++) {
auto &frameData = frameData_[i];
{
std::unique_lock<std::mutex> lock(frameData.push_mutex);
frameData.push_condVar.notify_all();
}
{
std::unique_lock<std::mutex> lock(frameData.pull_mutex);
frameData.pull_condVar.notify_all();
}
}
thread_.join();
ILOG("Vulkan submission thread joined. Frame=%d", vulkan_->GetCurFrame());
// Eat whatever has been queued up for this frame if anything.
Wipe();
// Wait for any fences to finish and be resignaled, so we don't have sync issues.
// Also clean out any queued data, which might refer to things that might not be valid
// when we restart...
for (int i = 0; i < vulkan_->GetInflightFrames(); i++) {
auto &frameData = frameData_[i];
_assert_(!frameData.readyForRun);
_assert_(frameData.steps.empty());
if (frameData.hasInitCommands) {
// Clear 'em out. This can happen on restart sometimes.
vkEndCommandBuffer(frameData.initCmd);
frameData.hasInitCommands = false;
}
frameData.readyForRun = false;
for (size_t i = 0; i < frameData.steps.size(); i++) {
delete frameData.steps[i];
}
frameData.steps.clear();
std::unique_lock<std::mutex> lock(frameData.push_mutex);
while (!frameData.readyForFence) {
VLOG("PUSH: Waiting for frame[%d].readyForFence = 1 (stop)", i);
frameData.push_condVar.wait(lock);
}
}
} else {
ILOG("Vulkan submission thread was already stopped.");
}
}
void StreamBuffer::CommitMemory(size_t final_num_bytes)
{
_assert_((m_current_offset + final_num_bytes) <= m_current_size);
_assert_(final_num_bytes <= m_last_allocation_size);
// For non-coherent mappings, flush the memory range
if (!m_coherent_mapping)
{
VkMappedMemoryRange range = {VK_STRUCTURE_TYPE_MAPPED_MEMORY_RANGE, nullptr, m_memory,
m_current_offset, final_num_bytes};
vkFlushMappedMemoryRanges(g_vulkan_context->GetDevice(), 1, &range);
}
m_current_offset += final_num_bytes;
}
void SWLoadCPReg(u32 sub_cmd, u32 value)
{
switch (sub_cmd & 0xF0)
{
case 0x30:
g_main_cp_state.matrix_index_a.Hex = value;
break;
case 0x40:
g_main_cp_state.matrix_index_b.Hex = value;
break;
case 0x50:
g_main_cp_state.vtx_desc.Hex &= ~0x1FFFF; // keep the Upper bits
g_main_cp_state.vtx_desc.Hex |= value;
break;
case 0x60:
g_main_cp_state.vtx_desc.Hex &= 0x1FFFF; // keep the lower 17Bits
g_main_cp_state.vtx_desc.Hex |= (u64)value << 17;
break;
case 0x70:
_assert_((sub_cmd & 0x0F) < 8);
g_main_cp_state.vtx_attr[sub_cmd & 7].g0.Hex = value;
break;
case 0x80:
_assert_((sub_cmd & 0x0F) < 8);
g_main_cp_state.vtx_attr[sub_cmd & 7].g1.Hex = value;
break;
case 0x90:
_assert_((sub_cmd & 0x0F) < 8);
g_main_cp_state.vtx_attr[sub_cmd & 7].g2.Hex = value;
break;
// Pointers to vertex arrays in GC RAM
case 0xA0:
g_main_cp_state.array_bases[sub_cmd & 0xF] = value;
cached_arraybases[sub_cmd & 0xF] = Memory::GetPointer(value);
break;
case 0xB0:
g_main_cp_state.array_strides[sub_cmd & 0xF] = value & 0xFF;
break;
}
}
static VkFormat VarToVkFormat(VarType t, uint32_t components, bool integer)
{
static const VkFormat float_type_lookup[][4] = {
{VK_FORMAT_R8_UNORM, VK_FORMAT_R8G8_UNORM, VK_FORMAT_R8G8B8_UNORM,
VK_FORMAT_R8G8B8A8_UNORM}, // VAR_UNSIGNED_BYTE
{VK_FORMAT_R8_SNORM, VK_FORMAT_R8G8_SNORM, VK_FORMAT_R8G8B8_SNORM,
VK_FORMAT_R8G8B8A8_SNORM}, // VAR_BYTE
{VK_FORMAT_R16_UNORM, VK_FORMAT_R16G16_UNORM, VK_FORMAT_R16G16B16_UNORM,
VK_FORMAT_R16G16B16A16_UNORM}, // VAR_UNSIGNED_SHORT
{VK_FORMAT_R16_SNORM, VK_FORMAT_R16G16_SNORM, VK_FORMAT_R16G16B16_SNORM,
VK_FORMAT_R16G16B16A16_SNORM}, // VAR_SHORT
{VK_FORMAT_R32_SFLOAT, VK_FORMAT_R32G32_SFLOAT, VK_FORMAT_R32G32B32_SFLOAT,
VK_FORMAT_R32G32B32A32_SFLOAT} // VAR_FLOAT
};
static const VkFormat integer_type_lookup[][4] = {
{VK_FORMAT_R8_UINT, VK_FORMAT_R8G8_UINT, VK_FORMAT_R8G8B8_UINT,
VK_FORMAT_R8G8B8A8_UINT}, // VAR_UNSIGNED_BYTE
{VK_FORMAT_R8_SINT, VK_FORMAT_R8G8_SINT, VK_FORMAT_R8G8B8_SINT,
VK_FORMAT_R8G8B8A8_SINT}, // VAR_BYTE
{VK_FORMAT_R16_UINT, VK_FORMAT_R16G16_UINT, VK_FORMAT_R16G16B16_UINT,
VK_FORMAT_R16G16B16A16_UINT}, // VAR_UNSIGNED_SHORT
{VK_FORMAT_R16_SINT, VK_FORMAT_R16G16_SINT, VK_FORMAT_R16G16B16_SINT,
VK_FORMAT_R16G16B16A16_SINT}, // VAR_SHORT
{VK_FORMAT_R32_SFLOAT, VK_FORMAT_R32G32_SFLOAT, VK_FORMAT_R32G32B32_SFLOAT,
VK_FORMAT_R32G32B32A32_SFLOAT} // VAR_FLOAT
};
_assert_(components > 0 && components <= 4);
return integer ? integer_type_lookup[t][components - 1] : float_type_lookup[t][components - 1];
}
VolumeWAD::VolumeWAD(std::unique_ptr<BlobReader> reader) : m_reader(std::move(reader))
{
_assert_(m_reader);
// Source: http://wiibrew.org/wiki/WAD_files
m_hdr_size = m_reader->ReadSwapped<u32>(0x00).value_or(0);
m_cert_size = m_reader->ReadSwapped<u32>(0x08).value_or(0);
m_tick_size = m_reader->ReadSwapped<u32>(0x10).value_or(0);
m_tmd_size = m_reader->ReadSwapped<u32>(0x14).value_or(0);
m_data_size = m_reader->ReadSwapped<u32>(0x18).value_or(0);
m_offset = Common::AlignUp(m_hdr_size, 0x40) + Common::AlignUp(m_cert_size, 0x40);
m_tmd_offset = Common::AlignUp(m_hdr_size, 0x40) + Common::AlignUp(m_cert_size, 0x40) +
Common::AlignUp(m_tick_size, 0x40);
m_opening_bnr_offset =
m_tmd_offset + Common::AlignUp(m_tmd_size, 0x40) + Common::AlignUp(m_data_size, 0x40);
if (!IOS::ES::IsValidTMDSize(m_tmd_size))
{
ERROR_LOG(DISCIO, "TMD is too large: %u bytes", m_tmd_size);
return;
}
std::vector<u8> tmd_buffer(m_tmd_size);
Read(m_tmd_offset, m_tmd_size, tmd_buffer.data());
m_tmd.SetBytes(std::move(tmd_buffer));
}
float Renderer::CalculateDrawAspectRatio(int target_width, int target_height)
{
// The dimensions are the sizes that are used to create the EFB/backbuffer textures, so
// they should always be greater than zero.
_assert_(target_width > 0 && target_height > 0);
if (g_ActiveConfig.iAspectRatio == ASPECT_STRETCH)
{
// If stretch is enabled, we prefer the aspect ratio of the window.
return (static_cast<float>(target_width) / static_cast<float>(target_height)) /
(static_cast<float>(s_backbuffer_width) / static_cast<float>(s_backbuffer_height));
}
float Ratio = static_cast<float>(target_width) / static_cast<float>(target_height);
if (g_ActiveConfig.iAspectRatio == ASPECT_ANALOG_WIDE || (g_ActiveConfig.iAspectRatio != ASPECT_ANALOG && g_ActiveConfig.iAspectRatio < ASPECT_ANALOG_WIDE && Core::g_aspect_wide))
{
Ratio /= AspectToWidescreen(VideoInterface::GetAspectRatio());
}
else if (g_ActiveConfig.iAspectRatio == ASPECT_4_3)
{
Ratio /= (4.0f / 3.0f);
}
else if (g_ActiveConfig.iAspectRatio == ASPECT_16_9)
{
Ratio /= (16.0f / 9.0f);
}
else if (g_ActiveConfig.iAspectRatio == ASPECT_16_10)
{
Ratio /= (16.0f / 10.0f);
}
else
{
Ratio /= VideoInterface::GetAspectRatio();
}
return Ratio;
}
float Renderer::CalculateDrawAspectRatio(int target_width, int target_height)
{
// The dimensions are the sizes that are used to create the EFB/backbuffer textures, so
// they should always be greater than zero.
_assert_(target_width > 0 && target_height > 0);
if (g_ActiveConfig.iAspectRatio == ASPECT_STRETCH)
{
// If stretch is enabled, we prefer the aspect ratio of the window.
return (static_cast<float>(target_width) / static_cast<float>(target_height)) /
(static_cast<float>(s_backbuffer_width) / static_cast<float>(s_backbuffer_height));
}
// The rendering window aspect ratio as a proportion of the 4:3 or 16:9 ratio
if (g_ActiveConfig.iAspectRatio == ASPECT_ANALOG_WIDE ||
(g_ActiveConfig.iAspectRatio != ASPECT_ANALOG && Core::g_aspect_wide))
{
return (static_cast<float>(target_width) / static_cast<float>(target_height)) /
AspectToWidescreen(VideoInterface::GetAspectRatio());
}
else
{
return (static_cast<float>(target_width) / static_cast<float>(target_height)) /
VideoInterface::GetAspectRatio();
}
}
static void StartReadInternal(bool copy_to_ram, u32 output_address, u64 dvd_offset, u32 length,
const DiscIO::Partition& partition,
DVDInterface::ReplyType reply_type, s64 ticks_until_completion)
{
_assert_(Core::IsCPUThread());
ReadRequest request;
request.copy_to_ram = copy_to_ram;
request.output_address = output_address;
request.dvd_offset = dvd_offset;
request.length = length;
request.partition = partition;
request.reply_type = reply_type;
u64 id = s_next_id++;
request.id = id;
request.time_started_ticks = CoreTiming::GetTicks();
request.realtime_started_us = Common::Timer::GetTimeUs();
s_request_queue.Push(std::move(request));
s_request_queue_expanded.Set();
CoreTiming::ScheduleEvent(ticks_until_completion, s_finish_read, id);
}
void StateTracker::SetFramebuffer(VkFramebuffer framebuffer, const VkRect2D& render_area)
{
// Should not be changed within a render pass.
_assert_(!InRenderPass());
m_framebuffer = framebuffer;
m_framebuffer_size = render_area;
}
void Jit::ApplyPrefixD(const u8 *vregs, VectorSize sz) {
_assert_(js.prefixDFlag & ArmJitState::PREFIX_KNOWN);
if (!js.prefixD) return;
int n = GetNumVectorElements(sz);
for (int i = 0; i < n; i++) {
if (js.VfpuWriteMask(i))
continue;
int sat = (js.prefixD >> (i * 2)) & 3;
if (sat == 1) {
// clamped = fabs(x) - fabs(x-0.5f) + 0.5f; // [ 0, 1]
fpr.MapRegV(vregs[i], MAP_DIRTY);
MOVI2F(S0, 0.5, R0);
VABS(S1, fpr.V(vregs[i])); // S1 = fabs(x)
VSUB(S2, fpr.V(vregs[i]), S0); // S2 = fabs(x-0.5f) {VABD}
VABS(S2, S2);
VSUB(fpr.V(vregs[i]), S1, S2); // v[i] = S1 - S2 + 0.5f
VADD(fpr.V(vregs[i]), fpr.V(vregs[i]), S0);
} else if (sat == 3) {
// clamped = fabs(x) - fabs(x-1.0f); // [-1, 1]
fpr.MapRegV(vregs[i], MAP_DIRTY);
MOVI2F(S0, 1.0, R0);
VABS(S1, fpr.V(vregs[i])); // S1 = fabs(x)
VSUB(S2, fpr.V(vregs[i]), S0); // S2 = fabs(x-1.0f) {VABD}
VABS(S2, S2);
VSUB(fpr.V(vregs[i]), S1, S2); // v[i] = S1 - S2
}
}
}
void VKTexture::Bind(unsigned int stage)
{
// Texture should always be in SHADER_READ_ONLY layout prior to use.
// This is so we don't need to transition during render passes.
_assert_(m_texture->GetLayout() == VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL);
StateTracker::GetInstance()->SetTexture(stage, m_texture->GetView());
}
bool SwapChain::SelectSurfaceFormat()
{
u32 format_count;
VkResult res = vkGetPhysicalDeviceSurfaceFormatsKHR(g_vulkan_context->GetPhysicalDevice(),
m_surface, &format_count, nullptr);
if (res != VK_SUCCESS || format_count == 0)
{
LOG_VULKAN_ERROR(res, "vkGetPhysicalDeviceSurfaceFormatsKHR failed: ");
return false;
}
std::vector<VkSurfaceFormatKHR> surface_formats(format_count);
res = vkGetPhysicalDeviceSurfaceFormatsKHR(g_vulkan_context->GetPhysicalDevice(), m_surface,
&format_count, surface_formats.data());
_assert_(res == VK_SUCCESS);
// If there is a single undefined surface format, the device doesn't care, so we'll just use RGBA
if (surface_formats[0].format == VK_FORMAT_UNDEFINED)
{
m_surface_format.format = VK_FORMAT_R8G8B8A8_UNORM;
m_surface_format.colorSpace = VK_COLOR_SPACE_SRGB_NONLINEAR_KHR;
return true;
}
// Use the first surface format, just use what it prefers.
// Some drivers seem to return a SRGB format here (Intel Mesa).
// This results in gamma correction when presenting to the screen, which we don't want.
// Use a linear format instead, if this is the case.
m_surface_format.format = Util::GetLinearFormat(surface_formats[0].format);
m_surface_format.colorSpace = surface_formats[0].colorSpace;
return true;
}
void CommandBufferManager::WaitForFence(VkFence fence)
{
// Find the command buffer that this fence corresponds to.
size_t command_buffer_index = 0;
for (; command_buffer_index < m_frame_resources.size(); command_buffer_index++)
{
if (m_frame_resources[command_buffer_index].fence == fence)
break;
}
_assert_(command_buffer_index < m_frame_resources.size());
// Has this command buffer already been waited for?
if (!m_frame_resources[command_buffer_index].needs_fence_wait)
return;
// Wait for this command buffer to be completed.
VkResult res =
vkWaitForFences(g_vulkan_context->GetDevice(), 1,
&m_frame_resources[command_buffer_index].fence, VK_TRUE, UINT64_MAX);
if (res != VK_SUCCESS)
LOG_VULKAN_ERROR(res, "vkWaitForFences failed: ");
// Immediately fire callbacks and cleanups, since the commands has been completed.
m_frame_resources[command_buffer_index].needs_fence_wait = false;
OnCommandBufferExecuted(command_buffer_index);
}
static void GenericStreamBuffer(DebuggerRequest &req, std::function<bool(const GPUDebugBuffer *&)> func) {
if (!currentDebugMIPS->isAlive()) {
return req.Fail("CPU not started");
}
if (coreState != CORE_STEPPING && !GPUStepping::IsStepping()) {
return req.Fail("Neither CPU or GPU is stepping");
}
bool includeAlpha = false;
if (!req.ParamBool("alpha", &includeAlpha, DebuggerParamType::OPTIONAL))
return;
u32 stackWidth = 0;
if (!req.ParamU32("stackWidth", &stackWidth, false, DebuggerParamType::OPTIONAL))
return;
std::string type = "uri";
if (!req.ParamString("type", &type, DebuggerParamType::OPTIONAL))
return;
if (type != "uri" && type != "base64")
return req.Fail("Parameter 'type' must be either 'uri' or 'base64'");
const GPUDebugBuffer *buf = nullptr;
if (!func(buf)) {
return req.Fail("Could not download output");
}
assert(buf != nullptr);
if (type == "base64") {
StreamBufferToBase64(req, *buf);
} else if (type == "uri") {
StreamBufferToDataURI(req, *buf, includeAlpha, stackWidth);
} else {
_assert_(false);
}
}
void DXTexture::ScaleRectangleFromTexture(const AbstractTexture* source,
const MathUtil::Rectangle<int>& srcrect,
const MathUtil::Rectangle<int>& dstrect)
{
const DXTexture* srcentry = static_cast<const DXTexture*>(source);
_assert_(m_config.rendertarget);
g_renderer->ResetAPIState(); // reset any game specific settings
const D3D11_VIEWPORT vp = CD3D11_VIEWPORT(float(dstrect.left), float(dstrect.top),
float(dstrect.GetWidth()), float(dstrect.GetHeight()));
D3D::stateman->UnsetTexture(m_texture->GetSRV());
D3D::stateman->Apply();
D3D::context->OMSetRenderTargets(1, &m_texture->GetRTV(), nullptr);
D3D::context->RSSetViewports(1, &vp);
D3D::SetLinearCopySampler();
D3D11_RECT srcRC;
srcRC.left = srcrect.left;
srcRC.right = srcrect.right;
srcRC.top = srcrect.top;
srcRC.bottom = srcrect.bottom;
D3D::drawShadedTexQuad(srcentry->m_texture->GetSRV(), &srcRC, srcentry->m_config.width,
srcentry->m_config.height, PixelShaderCache::GetColorCopyProgram(false),
VertexShaderCache::GetSimpleVertexShader(),
VertexShaderCache::GetSimpleInputLayout(),
GeometryShaderCache::GetCopyGeometryShader(), 1.0, 0);
FramebufferManager::BindEFBRenderTarget();
g_renderer->RestoreAPIState();
}
void ComputeShaderDispatcher::SetPushConstants(const void* data, size_t data_size)
{
_assert_(static_cast<u32>(data_size) < PUSH_CONSTANT_BUFFER_SIZE);
vkCmdPushConstants(m_command_buffer, m_pipeline_info.pipeline_layout, VK_SHADER_STAGE_COMPUTE_BIT,
0, static_cast<u32>(data_size), data);
}
void CommandBufferManager::AddFencePointCallback(
const void* key, const CommandBufferQueuedCallback& queued_callback,
const CommandBufferExecutedCallback& executed_callback)
{
// Shouldn't be adding twice.
_assert_(m_fence_point_callbacks.find(key) == m_fence_point_callbacks.end());
m_fence_point_callbacks.emplace(key, std::make_pair(queued_callback, executed_callback));
}
void UtilityShaderDraw::SetPSTexelBuffer(VkBufferView view)
{
// Should only be used with the texture conversion pipeline layout.
_assert_(m_pipeline_info.pipeline_layout ==
g_object_cache->GetPipelineLayout(PIPELINE_LAYOUT_TEXTURE_CONVERSION));
m_ps_texel_buffer = view;
}
void UtilityShaderDraw::SetPushConstants(const void* data, size_t data_size)
{
_assert_(static_cast<u32>(data_size) < PUSH_CONSTANT_BUFFER_SIZE);
vkCmdPushConstants(m_command_buffer, m_pipeline_info.pipeline_layout,
VK_SHADER_STAGE_VERTEX_BIT | VK_SHADER_STAGE_FRAGMENT_BIT, 0,
static_cast<u32>(data_size), data);
}
// This is used by several of the FileIO and /dev/fs functions
std::string HLE_IPC_BuildFilename(const std::string& wii_path)
{
std::string nand_path = File::GetUserPath(D_SESSION_WIIROOT_IDX);
if (wii_path.compare(0, 1, "/") == 0)
return nand_path + Common::EscapePath(wii_path);
_assert_(false);
return nand_path;
}
void StagingTexture2DBuffer::Unmap()
{
_assert_(m_map_pointer);
vkUnmapMemory(g_vulkan_context->GetDevice(), m_memory);
m_map_pointer = nullptr;
m_map_offset = 0;
m_map_size = 0;
}
bool TextureCache::Palettize(TCacheEntryBase* _entry, const TCacheEntryBase* base_entry)
{
TCacheEntry* entry = static_cast<TCacheEntry*>(_entry);
const TCacheEntry* unconverted = static_cast<const TCacheEntry*>(base_entry);
_assert_(entry->config.rendertarget);
m_texture_converter->ConvertTexture(entry, unconverted, m_render_pass, m_pallette, m_pallette_format, m_pallette_size);
return true;
}
void WaitUntilIdle()
{
_assert_(Core::IsCPUThread());
while (!s_request_queue.Empty())
s_result_queue_expanded.Wait();
StopDVDThread();
StartDVDThread();
}
bool StateTracker::CreateInstance()
{
_assert_(!s_state_tracker);
s_state_tracker = std::make_unique<StateTracker>();
if (!s_state_tracker->Initialize())
{
s_state_tracker.reset();
return false;
}
return true;
}
