raphiaura/KytyPS5
1
0
Fork 0
mirror of https://github.com/KytyPS5/KytyPS5.git synced 2026-10-20 02:47:02 +02:00
Code Issues Packages Projects Releases Wiki Activity

fix eop/flipping

Browse Source
This commit is contained in:
nmzik
2026-07-17 01:50:43 +02:00
parent a96b5111e6
commit bf1ea80bd9
14 changed files with 557 additions and 224 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/graphics/host_gpu/graphicContext.h
-1
View File
@@ -29,7 +29,6 @@ struct VulkanSwapchain {
std::unique_ptr<VkSemaphore[]> render_complete_semaphores;
uint32_t current_index = 0;
uint32_t present_frame = 0;
std::vector<std::unique_ptr<CommandBuffer>> present_command_buffers;
};
struct VulkanCommandPool {
src/graphics/host_gpu/renderer/context.cpp
+30 -41
View File
@@ -330,7 +330,8 @@ void CommandBuffer::Execute() {
queue.mutex->Unlock();
}
m_execute = true;
m_execute = true;
m_fence_waited = false;
m_submit_seq = g_command_buffer_submit_seq.fetch_add(1, std::memory_order_relaxed) + 1;
if (result != VK_SUCCESS) {
@@ -393,7 +394,8 @@ void CommandBuffer::ExecuteWithSemaphore(VkSemaphore signal_semaphore) {
queue.mutex->Unlock();
}
m_execute = true;
m_execute = true;
m_fence_waited = false;
m_submit_seq = g_command_buffer_submit_seq.fetch_add(1, std::memory_order_relaxed) + 1;
if (result != VK_SUCCESS) {
@@ -460,7 +462,8 @@ void CommandBuffer::ExecuteWithSemaphore(VkSemaphore wait_semaphore,
queue.mutex->Unlock();
}
m_execute = true;
m_execute = true;
m_fence_waited = false;
m_submit_seq = g_command_buffer_submit_seq.fetch_add(1, std::memory_order_relaxed) + 1;
if (result != VK_SUCCESS) {
@@ -474,27 +477,11 @@ void CommandBuffer::ExecuteWithSemaphore(VkSemaphore wait_semaphore,
}
void CommandBuffer::WaitForFence() {
EXIT_IF(IsInvalid());
const bool was_executed = m_execute;
if (m_execute) {
auto* device = g_render_ctx->GetGraphicCtx()->device;
auto result = vkWaitForFences(device, 1, &m_pool->fences[m_index], VK_TRUE, UINT64_MAX);
if (result != VK_SUCCESS) {
LOGF("vkWaitForFences failed: %s (%d), wait=WaitForFence queue=%d index=%u "
"submit_seq=%" PRIu64 " debug_op=%u debug_submit=%" PRIu64
" args=%u,%u,%u,%u,0x%016" PRIx64 "\n",
string_VkResult(result), static_cast<int>(result), m_queue, m_index, m_submit_seq,
m_debug_op, m_debug_submit_id, m_debug_arg0, m_debug_arg1, m_debug_arg2,
m_debug_arg3, m_debug_arg4);
}
EXIT_NOT_IMPLEMENTED(result != VK_SUCCESS);
m_execute = false;
}
WaitForFenceOnly();
if (was_executed) {
m_execute = false;
m_fence_waited = false;
RecycleDescriptorsAfterFence();
m_fence_resources.ReleaseAfterFence();
}
@@ -505,32 +492,34 @@ void CommandBuffer::WaitForFence() {
m_delete_after_fence.clear();
}
void CommandBuffer::WaitForFenceAndReset() {
void CommandBuffer::WaitForFenceOnly() {
EXIT_IF(IsInvalid());
if (!m_execute || m_fence_waited) {
return;
}
auto* device = g_render_ctx->GetGraphicCtx()->device;
auto result = vkWaitForFences(device, 1, &m_pool->fences[m_index], VK_TRUE, UINT64_MAX);
if (result != VK_SUCCESS) {
LOGF("vkWaitForFences failed: %s (%d), queue=%d index=%u submit_seq=%" PRIu64
" debug_op=%u debug_submit=%" PRIu64 " args=%u,%u,%u,%u,0x%016" PRIx64 "\n",
string_VkResult(result), static_cast<int>(result), m_queue, m_index, m_submit_seq,
m_debug_op, m_debug_submit_id, m_debug_arg0, m_debug_arg1, m_debug_arg2, m_debug_arg3,
m_debug_arg4);
}
EXIT_NOT_IMPLEMENTED(result != VK_SUCCESS);
m_fence_waited = true;
}
void CommandBuffer::WaitForFenceAndReset() {
const bool was_executed = m_execute;
if (m_execute) {
auto* device = g_render_ctx->GetGraphicCtx()->device;
auto result = vkWaitForFences(device, 1, &m_pool->fences[m_index], VK_TRUE, UINT64_MAX);
if (result != VK_SUCCESS) {
LOGF("vkWaitForFences failed: %s (%d), wait=WaitForFenceAndReset queue=%d index=%u "
"submit_seq=%" PRIu64 " debug_op=%u debug_submit=%" PRIu64
" args=%u,%u,%u,%u,0x%016" PRIx64 "\n",
string_VkResult(result), static_cast<int>(result), m_queue, m_index, m_submit_seq,
m_debug_op, m_debug_submit_id, m_debug_arg0, m_debug_arg1, m_debug_arg2,
m_debug_arg3, m_debug_arg4);
}
EXIT_NOT_IMPLEMENTED(result != VK_SUCCESS);
m_execute = false;
WaitForFenceOnly();
if (was_executed) {
m_execute = false;
m_fence_waited = false;
vkResetCommandBuffer(m_pool->buffers[m_index],
VK_COMMAND_BUFFER_RESET_RELEASE_RESOURCES_BIT);
}
m_host_stream.Reset();
if (was_executed) {
RecycleDescriptorsAfterFence();
m_fence_resources.ReleaseAfterFence();
src/graphics/host_gpu/renderer/render.h
+9 -3
View File
@@ -76,6 +76,7 @@ public:
void BeginRenderPass(VulkanFramebuffer* framebuffer, RenderColorInfo* colors,
uint32_t color_count, RenderDepthInfo* depth) const;
void EndRenderPass() const;
void WaitForFenceOnly();
void WaitForFence();
void WaitForFenceAndReset();
void DeleteAfterFence(VulkanBuffer* buffer);
@@ -96,6 +97,7 @@ private:
uint32_t m_index = static_cast<uint32_t>(-1);
int m_queue = -1;
bool m_execute = false;
bool m_fence_waited = false;
uint64_t m_submit_seq = 0;
uint32_t m_debug_op = 0;
uint64_t m_debug_submit_id = 0;
@@ -150,14 +152,18 @@ void GraphicsRenderWriteAtEndOfPipe32(uint64_t submit_id, CommandBuffer* buffer,
void GraphicsRenderWriteAtEndOfPipeGds32(uint64_t submit_id, CommandBuffer* buffer,
uint32_t* dst_gpu_addr, uint32_t dw_offset,
uint32_t dw_num);
uint64_t GraphicsRenderPrepareDisplayBufferFlip(CommandBuffer* buffer, int handle, int index,
int flip_mode, int64_t flip_arg);
void GraphicsRenderWriteAtEndOfPipeWithInterruptWriteBackFlip32(
uint64_t submit_id, CommandBuffer* buffer, uint32_t* dst_gpu_addr, uint32_t value, int handle,
int index, int flip_mode, int64_t flip_arg);
int index, int flip_mode, int64_t flip_arg, uint64_t request_id);
void GraphicsRenderWriteAtEndOfPipeWithFlip32(uint64_t submit_id, CommandBuffer* buffer,
uint32_t* dst_gpu_addr, uint32_t value, int handle,
int index, int flip_mode, int64_t flip_arg);
int index, int flip_mode, int64_t flip_arg,
uint64_t request_id);
void GraphicsRenderWriteAtEndOfPipeOnlyFlip(uint64_t submit_id, CommandBuffer* buffer, int handle,
int index, int flip_mode, int64_t flip_arg);
int index, int flip_mode, int64_t flip_arg,
uint64_t request_id);
void GraphicsRenderWriteAtEndOfPipeWithWriteBack64(uint64_t submit_id, CommandBuffer* buffer,
uint64_t* dst_gpu_addr, uint64_t value);
void GraphicsRenderWriteAtEndOfPipeWithWriteBack32(uint64_t submit_id, CommandBuffer* buffer,
src/graphics/host_gpu/renderer/sync.cpp
+28 -37
View File
@@ -84,30 +84,12 @@ static void PublishImmediateFence(T* dst, T value) {
std::memcpy(dst, &value, sizeof(value));
}
static void SubmitDisplayBufferFlip(const uint64_t* args) {
if (g_render_ctx == nullptr || args == nullptr) {
EXIT("GPU flip submission has invalid state, render_ctx=%p args=%p\n",
static_cast<const void*>(g_render_ctx), static_cast<const void*>(args));
}
const auto handle = static_cast<int>(args[0]);
const auto index = static_cast<int>(args[1]);
const auto flip_mode = static_cast<int>(args[2]);
const auto flip_arg = static_cast<int64_t>(args[3]);
const auto result =
Presentation::DisplayBufferSubmitFlipFromGpu(handle, index, flip_mode, flip_arg);
if (result != 0) {
EXIT("GPU flip submission failed, result=%d handle=%d index=%d mode=%d arg=%" PRId64 "\n",
result, handle, index, flip_mode, flip_arg);
}
}
static bool CompleteDisplayBufferFlip(const uint64_t* args) {
if (g_render_ctx == nullptr || args == nullptr) {
EXIT("GPU flip completion has invalid state, render_ctx=%p args=%p\n",
static_cast<const void*>(g_render_ctx), static_cast<const void*>(args));
}
Presentation::DisplayBufferCompleteFlipFromGpu(static_cast<int>(args[0]),
static_cast<int>(args[1]));
Presentation::DisplayBufferCompleteFlipFromGpu(args[0]);
return true;
}
@@ -348,9 +330,28 @@ void GraphicsRenderWriteAtEndOfPipeWithInterrupt32(uint64_t submit_id, CommandBu
args);
}
uint64_t GraphicsRenderPrepareDisplayBufferFlip(CommandBuffer* buffer, int handle, int index,
int flip_mode, int64_t flip_arg) {
for (;;) {
uint64_t request_id = 0;
const auto result = Presentation::DisplayBufferSubmitFlipFromGpu(
buffer, handle, index, flip_mode, flip_arg, &request_id);
if (result == OK) {
EXIT_IF(request_id == 0);
return request_id;
}
if (result != VideoOut::VIDEO_OUT_ERROR_FLIP_QUEUE_FULL) {
EXIT("GPU flip submission failed, result=%d handle=%d index=%d mode=%d arg=%" PRId64
"\n",
result, handle, index, flip_mode, flip_arg);
}
Presentation::DisplayBufferWaitForFlipQueueSlot();
}
}
void GraphicsRenderWriteAtEndOfPipeWithInterruptWriteBackFlip32(
uint64_t submit_id, CommandBuffer* buffer, uint32_t* dst_gpu_addr, uint32_t value, int handle,
int index, int flip_mode, int64_t flip_arg) {
int index, int flip_mode, int64_t flip_arg, uint64_t request_id) {
EXIT_IF(g_render_ctx == nullptr);
EXIT_IF(dst_gpu_addr == nullptr);
EXIT_IF(buffer == nullptr);
@@ -361,11 +362,7 @@ void GraphicsRenderWriteAtEndOfPipeWithInterruptWriteBackFlip32(
static_cast<uint32_t>(flip_mode), value, static_cast<uint64_t>(flip_arg));
PublishImmediateFence(dst_gpu_addr, value);
uint64_t args[LABEL_ARGS_MAX] = {static_cast<uint64_t>(handle), static_cast<uint64_t>(index),
static_cast<uint64_t>(flip_mode),
static_cast<uint64_t>(flip_arg)};
SubmitDisplayBufferFlip(args);
const uint64_t args[LABEL_ARGS_MAX] = {request_id};
SubmitLabel32(
buffer, nullptr, 0, CompleteDisplayBufferFlip,
[](const uint64_t* /*args*/) {
@@ -378,7 +375,8 @@ void GraphicsRenderWriteAtEndOfPipeWithInterruptWriteBackFlip32(
void GraphicsRenderWriteAtEndOfPipeWithFlip32(uint64_t submit_id, CommandBuffer* buffer,
uint32_t* dst_gpu_addr, uint32_t value, int handle,
int index, int flip_mode, int64_t flip_arg) {
int index, int flip_mode, int64_t flip_arg,
uint64_t request_id) {
EXIT_IF(g_render_ctx == nullptr);
EXIT_IF(dst_gpu_addr == nullptr);
EXIT_IF(buffer == nullptr);
@@ -389,16 +387,13 @@ void GraphicsRenderWriteAtEndOfPipeWithFlip32(uint64_t submit_id, CommandBuffer*
static_cast<uint32_t>(flip_mode), value, static_cast<uint64_t>(flip_arg));
PublishImmediateFence(dst_gpu_addr, value);
uint64_t args[LABEL_ARGS_MAX] = {static_cast<uint64_t>(handle), static_cast<uint64_t>(index),
static_cast<uint64_t>(flip_mode),
static_cast<uint64_t>(flip_arg)};
SubmitDisplayBufferFlip(args);
const uint64_t args[LABEL_ARGS_MAX] = {request_id};
SubmitLabel32(buffer, nullptr, 0, CompleteDisplayBufferFlip, nullptr, args);
}
void GraphicsRenderWriteAtEndOfPipeOnlyFlip(uint64_t submit_id, CommandBuffer* buffer, int handle,
int index, int flip_mode, int64_t flip_arg) {
int index, int flip_mode, int64_t flip_arg,
uint64_t request_id) {
EXIT_IF(g_render_ctx == nullptr);
EXIT_IF(buffer == nullptr);
EXIT_IF(buffer->IsInvalid());
@@ -407,11 +402,7 @@ void GraphicsRenderWriteAtEndOfPipeOnlyFlip(uint64_t submit_id, CommandBuffer* b
static_cast<uint32_t>(handle), static_cast<uint32_t>(index),
static_cast<uint32_t>(flip_mode), 0, static_cast<uint64_t>(flip_arg));
uint64_t args[LABEL_ARGS_MAX] = {static_cast<uint64_t>(handle), static_cast<uint64_t>(index),
static_cast<uint64_t>(flip_mode),
static_cast<uint64_t>(flip_arg)};
SubmitDisplayBufferFlip(args);
const uint64_t args[LABEL_ARGS_MAX] = {request_id};
SubmitLabel32(buffer, nullptr, 0, CompleteDisplayBufferFlip, nullptr, args);
}
src/graphics/host_gpu/utils.cpp
+19 -6
View File
@@ -863,17 +863,21 @@ void UtilCopyImageWithBuffer(CommandBuffer* buffer, GraphicContext* ctx, VulkanI
final_layout);
}
void UtilBlitImage(CommandBuffer* buffer, VulkanImage* src_image, VulkanSwapchain* dst_swapchain) {
void UtilBlitPreparedImage(CommandBuffer* buffer, VulkanImage* src_image,
VulkanSwapchain* dst_swapchain) {
auto* vk_buffer = buffer->GetPool()->buffers[buffer->GetIndex()];
if (src_image->layout != VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL) {
EXIT("invalid prepared presentation image, image=%p vk_image=%p layout=%d\n",
static_cast<const void*>(src_image),
src_image != nullptr ? static_cast<void*>(src_image->image) : nullptr,
src_image != nullptr ? static_cast<int>(src_image->layout) : -1);
}
VulkanImage swapchain_image(VulkanImageType::Unknown);
swapchain_image.image = dst_swapchain->swapchain_images[dst_swapchain->current_index];
swapchain_image.layout = VK_IMAGE_LAYOUT_UNDEFINED;
auto src_layout = src_image->layout;
SetImageLayout(vk_buffer, src_image, 0, 1, VK_IMAGE_ASPECT_COLOR_BIT, src_layout,
VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL);
SetImageLayout(vk_buffer, &swapchain_image, 0, 1, VK_IMAGE_ASPECT_COLOR_BIT,
VK_IMAGE_LAYOUT_UNDEFINED, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL);
@@ -902,9 +906,18 @@ void UtilBlitImage(CommandBuffer* buffer, VulkanImage* src_image, VulkanSwapchai
vkCmdBlitImage(vk_buffer, src_image->image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,
swapchain_image.image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 1, &region,
VK_FILTER_LINEAR);
}
SetImageLayout(vk_buffer, src_image, 0, 1, VK_IMAGE_ASPECT_COLOR_BIT,
VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, VK_IMAGE_LAYOUT_GENERAL);
void UtilClearColorImage(CommandBuffer* buffer, VulkanImage* image,
const VkClearColorValue& color) {
auto* vk_buffer = buffer->GetPool()->buffers[buffer->GetIndex()];
SetImageLayout(vk_buffer, image, 0, 1, VK_IMAGE_ASPECT_COLOR_BIT, image->layout,
VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL);
const VkImageSubresourceRange range {VK_IMAGE_ASPECT_COLOR_BIT, 0, 1, 0, 1};
vkCmdClearColorImage(vk_buffer, image->image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, &color, 1,
&range);
SetImageLayout(vk_buffer, image, 0, 1, VK_IMAGE_ASPECT_COLOR_BIT,
VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL);
}
void VulkanCreateBuffer(GraphicContext* gctx, uint64_t size, VulkanBuffer* buffer) {
src/graphics/host_gpu/utils.h
+4 -1
View File
@@ -90,7 +90,10 @@ void UtilCopyImageWithBuffer(CommandBuffer* buffer, GraphicContext* ctx, VulkanI
VkImageAspectFlags src_aspect, VulkanImage* dst_image,
VkImageAspectFlags dst_aspect, uint32_t bytes_per_element,
uint64_t dst_layout);
void UtilBlitImage(CommandBuffer* buffer, VulkanImage* src_image, VulkanSwapchain* dst_swapchain);
void UtilBlitPreparedImage(CommandBuffer* buffer, VulkanImage* src_image,
VulkanSwapchain* dst_swapchain);
void UtilClearColorImage(CommandBuffer* buffer, VulkanImage* image,
const VkClearColorValue& color);
void UtilFillImage(GraphicContext* ctx, VulkanImage* dst_image, const void* src_data, uint64_t size,
uint32_t src_pitch, uint64_t dst_layout);
void UtilFillImage(GraphicContext* ctx, DepthStencilVulkanImage* dst_image, const void* src_data,