lee/renderdoc
1
0
Fork 0
mirror of https://github.com/baldurk/renderdoc.git synced 2026-09-16 02:35:49 +00:00
Code Issues Packages Projects Releases Wiki Activity
Files
24f4bd7b4263185916331fcd68a2519287ff61c2
renderdoc/renderdoc/driver/vulkan/vk_replay.cpp
T

4857 lines
167 KiB
C++
Raw Blame History

/******************************************************************************
* The MIT License (MIT)
*
* Copyright (c) 2019-2024 Baldur Karlsson
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to deal
* in the Software without restriction, including without limitation the rights
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
* copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
* THE SOFTWARE.
******************************************************************************/
#include "vk_replay.h"
#include <ctype.h>
#include <float.h>
#include <math.h>
#include <algorithm>
#include "core/settings.h"
#include "data/glsl_shaders.h"
#include "driver/ihv/amd/amd_rgp.h"
#include "driver/shaders/spirv/glslang_compile.h"
#include "maths/formatpacking.h"
#include "maths/matrix.h"
#include "replay/dummy_driver.h"
#include "serialise/rdcfile.h"
#include "strings/string_utils.h"
#include "vk_core.h"
#include "vk_debug.h"
#include "vk_resources.h"
#include "vk_shader_cache.h"
#define VULKAN 1
#include "data/glsl/glsl_ubos_cpp.h"
RDOC_EXTERN_CONFIG(bool, Vulkan_Debug_SingleSubmitFlushing);
static const char *SPIRVDisassemblyTarget = "SPIR-V (RenderDoc)";
static const char *AMDShaderInfoTarget = "AMD_shader_info";
static const char *KHRExecutablePropertiesTarget = "KHR_pipeline_executable_properties";
VulkanReplay::VulkanReplay(WrappedVulkan *d)
{
RenderDoc::Inst().RegisterMemoryRegion(this, sizeof(VulkanReplay));
m_pDriver = d;
m_Proxy = false;
m_HighlightCache.driver = this;
m_OutputWinID = 1;
m_ActiveWinID = 0;
m_BindDepth = false;
m_DebugWidth = m_DebugHeight = 1;
RDCEraseEl(m_DriverInfo);
}
VulkanDebugManager *VulkanReplay::GetDebugManager()
{
return m_pDriver->GetDebugManager();
}
VulkanResourceManager *VulkanReplay::GetResourceManager()
{
return m_pDriver->GetResourceManager();
}
void VulkanReplay::Shutdown()
{
SAFE_DELETE(m_RGP);
m_pDriver->Shutdown();
delete m_pDriver;
}
RDResult VulkanReplay::FatalErrorCheck()
{
return m_pDriver->FatalErrorCheck();
}
IReplayDriver *VulkanReplay::MakeDummyDriver()
{
// gather up the shaders we've allocated to pass to the dummy driver
rdcarray<ShaderReflection *> shaders;
for(auto it = m_pDriver->m_CreationInfo.m_ShaderModule.begin();
it != m_pDriver->m_CreationInfo.m_ShaderModule.end(); it++)
{
for(auto reflit = it->second.m_Reflections.begin(); reflit != it->second.m_Reflections.end();
++reflit)
{
shaders.push_back(reflit->second.refl);
reflit->second.refl = NULL;
}
}
IReplayDriver *dummy = new DummyDriver(this, shaders, m_pDriver->DetachStructuredFile());
return dummy;
}
rdcarray<GPUDevice> VulkanReplay::GetAvailableGPUs()
{
rdcarray<GPUDevice> ret;
// do a manual enumerate to avoid any possible remapping
VkInstance instance = m_pDriver->GetInstance();
uint32_t count = 0;
VkResult vkr = ObjDisp(instance)->EnumeratePhysicalDevices(Unwrap(instance), &count, NULL);
if(vkr != VK_SUCCESS)
return ret;
VkPhysicalDevice *devices = new VkPhysicalDevice[count];
vkr = ObjDisp(instance)->EnumeratePhysicalDevices(Unwrap(instance), &count, devices);
CheckVkResult(vkr);
for(uint32_t p = 0; p < count; p++)
{
VkPhysicalDeviceProperties props = {};
ObjDisp(instance)->GetPhysicalDeviceProperties(devices[p], &props);
VkPhysicalDeviceDriverProperties driverProps = {};
GetPhysicalDeviceDriverProperties(ObjDisp(instance), devices[p], driverProps);
VkDriverInfo driverInfo(props, driverProps);
GPUDevice dev;
dev.vendor = driverInfo.Vendor();
dev.deviceID = props.deviceID;
dev.name = props.deviceName;
dev.apis = {GraphicsAPI::Vulkan};
// only set the driver name when it's useful to disambiguate
dev.driver = HumanDriverName(driverProps.driverID);
// don't add duplicate devices even if they get enumerated.
if(ret.indexOf(dev) == -1)
ret.push_back(dev);
}
// loop over devices and remove the driver string unless it's needed to disambiguate from another
// identical device.
for(size_t i = 0; i < ret.size(); i++)
{
bool needDriver = false;
for(size_t j = 0; j < ret.size(); j++)
{
if(i == j)
continue;
if(ret[i].vendor == ret[j].vendor && ret[i].deviceID == ret[j].deviceID)
{
RDCASSERT(ret[i].driver != ret[j].driver);
needDriver = true;
break;
}
}
if(!needDriver)
ret[i].driver = rdcstr();
}
SAFE_DELETE_ARRAY(devices);
return ret;
}
APIProperties VulkanReplay::GetAPIProperties()
{
APIProperties ret = m_pDriver->APIProps;
ret.pipelineType = GraphicsAPI::Vulkan;
ret.localRenderer = GraphicsAPI::Vulkan;
ret.degraded = false;
ret.shadersMutable = false;
ret.rgpCapture =
(m_DriverInfo.vendor == GPUVendor::AMD || m_DriverInfo.vendor == GPUVendor::Samsung) &&
m_RGP != NULL && m_RGP->DriverSupportsInterop();
ret.shaderDebugging = true;
ret.pixelHistory = true;
return ret;
}
RDResult VulkanReplay::ReadLogInitialisation(RDCFile *rdc, bool storeStructuredBuffers)
{
return m_pDriver->ReadLogInitialisation(rdc, storeStructuredBuffers);
}
void VulkanReplay::ReplayLog(uint32_t endEventID, ReplayLogType replayType)
{
if(replayType == eReplay_OnlyDraw)
{
bool replayed = FetchShaderFeedback(endEventID);
if(replayed)
return;
}
m_pDriver->ReplayLog(0, endEventID, replayType);
}
SDFile *VulkanReplay::GetStructuredFile()
{
return m_pDriver->GetStructuredFile();
}
rdcarray<uint32_t> VulkanReplay::GetPassEvents(uint32_t eventId)
{
rdcarray<uint32_t> passEvents;
const ActionDescription *action = m_pDriver->GetAction(eventId);
if(!action)
return passEvents;
// for vulkan a pass == a renderpass, if we're not inside a
// renderpass then there are no pass events.
const ActionDescription *start = action;
while(start)
{
// if we've come to the beginning of a pass, break out of the loop, we've
// found the start.
// Note that vkCmdNextSubPass has both Begin and End flags set, so it will
// break out here before we hit the terminating case looking for ActionFlags::EndPass
if(start->flags & ActionFlags::BeginPass)
break;
// if we come to the END of a pass, since we were iterating backwards that
// means we started outside of a pass, so return empty set.
// Note that vkCmdNextSubPass has both Begin and End flags set, so it will
// break out above before we hit this terminating case
if(start->flags & ActionFlags::EndPass)
return passEvents;
// if we've come to the start of the log we were outside of a render pass
// to start with
if(start->previous == NULL)
return passEvents;
// step back
start = start->previous;
}
// store all the action eventIDs up to the one specified at the start
while(start)
{
if(start->eventId >= action->eventId)
break;
// include pass boundaries, these will be filtered out later
// so we don't actually do anything (init postvs/action overlay)
// but it's useful to have the first part of the pass as part
// of the list
if(start->flags & (ActionFlags::MeshDispatch | ActionFlags::Drawcall | ActionFlags::PassBoundary))
passEvents.push_back(start->eventId);
start = start->next;
}
return passEvents;
}
ResourceId VulkanReplay::GetLiveID(ResourceId id)
{
if(!m_pDriver->GetResourceManager()->HasLiveResource(id))
return ResourceId();
return m_pDriver->GetResourceManager()->GetLiveID(id);
}
rdcarray<DebugMessage> VulkanReplay::GetDebugMessages()
{
return m_pDriver->GetDebugMessages();
}
ResourceDescription &VulkanReplay::GetResourceDesc(ResourceId id)
{
auto it = m_ResourceIdx.find(id);
if(it == m_ResourceIdx.end())
{
m_ResourceIdx[id] = m_Resources.size();
m_Resources.push_back(ResourceDescription());
m_Resources.back().resourceId = id;
return m_Resources.back();
}
return m_Resources[it->second];
}
rdcarray<ResourceDescription> VulkanReplay::GetResources()
{
return m_Resources;
}
rdcarray<TextureDescription> VulkanReplay::GetTextures()
{
rdcarray<TextureDescription> texs;
for(auto it = m_pDriver->m_ImageStates.begin(); it != m_pDriver->m_ImageStates.end(); ++it)
{
// skip textures that aren't from the capture
if(m_pDriver->GetResourceManager()->GetOriginalID(it->first) == it->first)
continue;
texs.push_back(GetTexture(it->first));
}
return texs;
}
rdcarray<BufferDescription> VulkanReplay::GetBuffers()
{
rdcarray<BufferDescription> bufs;
for(auto it = m_pDriver->m_CreationInfo.m_Buffer.begin();
it != m_pDriver->m_CreationInfo.m_Buffer.end(); ++it)
{
// skip textures that aren't from the capture
if(m_pDriver->GetResourceManager()->GetOriginalID(it->first) == it->first)
continue;
bufs.push_back(GetBuffer(it->first));
}
// sort the buffers by ID
std::sort(bufs.begin(), bufs.end());
return bufs;
}
TextureDescription VulkanReplay::GetTexture(ResourceId id)
{
VulkanCreationInfo::Image &iminfo = m_pDriver->m_CreationInfo.m_Image[id];
TextureDescription ret = {};
ret.resourceId = m_pDriver->GetResourceManager()->GetOriginalID(id);
ret.arraysize = iminfo.arrayLayers;
ret.creationFlags = iminfo.creationFlags;
ret.cubemap = iminfo.cube;
ret.width = iminfo.extent.width;
ret.height = iminfo.extent.height;
ret.depth = iminfo.extent.depth;
ret.mips = iminfo.mipLevels;
ret.byteSize = 0;
for(uint32_t s = 0; s < ret.mips; s++)
ret.byteSize += GetByteSize(ret.width, ret.height, ret.depth, iminfo.format, s);
ret.byteSize *= ret.arraysize;
ret.msQual = 0;
ret.msSamp = RDCMAX(1U, (uint32_t)iminfo.samples);
ret.byteSize *= ret.msSamp;
ret.format = MakeResourceFormat(iminfo.format);
switch(iminfo.type)
{
case VK_IMAGE_TYPE_1D:
ret.type = iminfo.arrayLayers > 1 ? TextureType::Texture1DArray : TextureType::Texture1D;
ret.dimension = 1;
break;
case VK_IMAGE_TYPE_2D:
if(ret.msSamp > 1)
ret.type = iminfo.arrayLayers > 1 ? TextureType::Texture2DMSArray : TextureType::Texture2DMS;
else if(ret.cubemap)
ret.type = iminfo.arrayLayers > 6 ? TextureType::TextureCubeArray : TextureType::TextureCube;
else
ret.type = iminfo.arrayLayers > 1 ? TextureType::Texture2DArray : TextureType::Texture2D;
ret.dimension = 2;
break;
case VK_IMAGE_TYPE_3D:
ret.type = TextureType::Texture3D;
ret.dimension = 3;
break;
default:
ret.dimension = 2;
RDCERR("Unexpected image type");
break;
}
return ret;
}
BufferDescription VulkanReplay::GetBuffer(ResourceId id)
{
VulkanCreationInfo::Buffer &bufinfo = m_pDriver->m_CreationInfo.m_Buffer[id];
BufferDescription ret;
ret.resourceId = m_pDriver->GetResourceManager()->GetOriginalID(id);
ret.length = bufinfo.size;
ret.creationFlags = BufferCategory::NoFlags;
ret.gpuAddress = bufinfo.gpuAddress;
if(bufinfo.usage & (VK_BUFFER_USAGE_STORAGE_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_TEXEL_BUFFER_BIT))
ret.creationFlags |= BufferCategory::ReadWrite;
if(bufinfo.usage & (VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT))
ret.creationFlags |= BufferCategory::Constants;
if(bufinfo.usage & (VK_BUFFER_USAGE_INDIRECT_BUFFER_BIT))
ret.creationFlags |= BufferCategory::Indirect;
if(bufinfo.usage & (VK_BUFFER_USAGE_INDEX_BUFFER_BIT))
ret.creationFlags |= BufferCategory::Index;
if(bufinfo.usage & (VK_BUFFER_USAGE_VERTEX_BUFFER_BIT))
ret.creationFlags |= BufferCategory::Vertex;
return ret;
}
rdcarray<ShaderEntryPoint> VulkanReplay::GetShaderEntryPoints(ResourceId shader)
{
auto shad = m_pDriver->m_CreationInfo.m_ShaderModule.find(shader);
if(shad == m_pDriver->m_CreationInfo.m_ShaderModule.end())
return {};
return shad->second.spirv.EntryPoints();
}
ShaderReflection *VulkanReplay::GetShader(ResourceId pipeline, ResourceId shader,
ShaderEntryPoint entry)
{
auto shad = m_pDriver->m_CreationInfo.m_ShaderModule.find(shader);
if(shad == m_pDriver->m_CreationInfo.m_ShaderModule.end())
{
RDCERR("Can't get shader details");
return NULL;
}
// if this shader was never used in a pipeline the reflection won't be prepared. Do that now -
// this will be ignored if it was already prepared.
shad->second.GetReflection(entry.stage, entry.name, pipeline)
.Init(GetResourceManager(), shader, shad->second.spirv, entry.name,
VkShaderStageFlagBits(1 << uint32_t(entry.stage)), {});
return shad->second.GetReflection(entry.stage, entry.name, pipeline).refl;
}
rdcarray<rdcstr> VulkanReplay::GetDisassemblyTargets(bool withPipeline)
{
rdcarray<rdcstr> ret;
if(withPipeline && m_pDriver->GetExtensions(NULL).ext_AMD_shader_info)
ret.push_back(AMDShaderInfoTarget);
if(withPipeline && m_pDriver->GetExtensions(NULL).ext_KHR_pipeline_executable_properties)
ret.push_back(KHRExecutablePropertiesTarget);
// default is always first
ret.insert(0, SPIRVDisassemblyTarget);
// could add canonical disassembly here if spirv-dis is available
// Ditto for SPIRV-cross (to glsl/hlsl)
return ret;
}
void VulkanReplay::CachePipelineExecutables(ResourceId pipeline)
{
auto it = m_PipelineExecutables.insert({pipeline, rdcarray<PipelineExecutables>()});
if(!it.second)
return;
rdcarray<PipelineExecutables> &data = it.first->second;
VkPipeline pipe = m_pDriver->GetResourceManager()->GetCurrentHandle<VkPipeline>(pipeline);
VkPipelineInfoKHR pipeInfo = {
VK_STRUCTURE_TYPE_PIPELINE_INFO_KHR,
NULL,
Unwrap(pipe),
};
VkDevice dev = m_pDriver->GetDev();
const VkDevDispatchTable *vt = ObjDisp(dev);
uint32_t execCount = 0;
vt->GetPipelineExecutablePropertiesKHR(Unwrap(dev), &pipeInfo, &execCount, NULL);
rdcarray<VkPipelineExecutablePropertiesKHR> executables;
executables.resize(execCount);
for(uint32_t i = 0; i < execCount; i++)
executables[i].sType = VK_STRUCTURE_TYPE_PIPELINE_EXECUTABLE_PROPERTIES_KHR;
data.resize(execCount);
vt->GetPipelineExecutablePropertiesKHR(Unwrap(dev), &pipeInfo, &execCount, executables.data());
for(uint32_t i = 0; i < execCount; i++)
{
const VkPipelineExecutablePropertiesKHR &exec = executables[i];
PipelineExecutables &out = data[i];
out.name = exec.name;
out.description = exec.description;
out.stages = exec.stages;
out.subgroupSize = exec.subgroupSize;
rdcarray<VkPipelineExecutableStatisticKHR> &stats = out.statistics;
rdcarray<VkPipelineExecutableInternalRepresentationKHR> &irs = out.representations;
VkPipelineExecutableInfoKHR pipeExecInfo = {
VK_STRUCTURE_TYPE_PIPELINE_EXECUTABLE_INFO_KHR,
NULL,
Unwrap(pipe),
i,
};
// enumerate statistics
uint32_t statCount = 0;
vt->GetPipelineExecutableStatisticsKHR(Unwrap(dev), &pipeExecInfo, &statCount, NULL);
stats.resize(statCount);
for(uint32_t s = 0; s < statCount; s++)
stats[s].sType = VK_STRUCTURE_TYPE_PIPELINE_EXECUTABLE_STATISTIC_KHR;
vt->GetPipelineExecutableStatisticsKHR(Unwrap(dev), &pipeExecInfo, &statCount, stats.data());
// enumerate internal representations
uint32_t irCount = 0;
vt->GetPipelineExecutableInternalRepresentationsKHR(Unwrap(dev), &pipeExecInfo, &irCount, NULL);
irs.resize(irCount);
for(uint32_t ir = 0; ir < irCount; ir++)
irs[ir].sType = VK_STRUCTURE_TYPE_PIPELINE_EXECUTABLE_INTERNAL_REPRESENTATION_KHR;
vt->GetPipelineExecutableInternalRepresentationsKHR(Unwrap(dev), &pipeExecInfo, &irCount,
irs.data());
// need to now allocate space, and try again
out.irbytes.resize(irCount);
for(uint32_t ir = 0; ir < irCount; ir++)
{
out.irbytes[ir].resize(irs[ir].dataSize);
irs[ir].pData = out.irbytes[ir].data();
}
vt->GetPipelineExecutableInternalRepresentationsKHR(Unwrap(dev), &pipeExecInfo, &irCount,
irs.data());
}
}
rdcstr VulkanReplay::DisassembleShader(ResourceId pipeline, const ShaderReflection *refl,
const rdcstr &target)
{
auto it = m_pDriver->m_CreationInfo.m_ShaderModule.find(
GetResourceManager()->GetLiveID(refl->resourceId));
if(it == m_pDriver->m_CreationInfo.m_ShaderModule.end())
return "; Invalid Shader Specified";
if(target == SPIRVDisassemblyTarget || target.empty())
{
VulkanCreationInfo::ShaderModuleReflection &moduleRefl =
it->second.GetReflection(refl->stage, refl->entryPoint, pipeline);
moduleRefl.PopulateDisassembly(it->second.spirv);
return moduleRefl.disassembly;
}
VkDevice dev = m_pDriver->GetDev();
const VkDevDispatchTable *vt = ObjDisp(dev);
if(target == AMDShaderInfoTarget && vt->GetShaderInfoAMD)
{
if(pipeline == ResourceId())
{
return "; No pipeline specified, VK_AMD_shader_info disassembly is not available\n"
"; Shader must be disassembled with a specific pipeline.";
}
VkPipeline pipe = m_pDriver->GetResourceManager()->GetCurrentHandle<VkPipeline>(pipeline);
VkShaderStageFlagBits stageBit = VkShaderStageFlagBits(
1 << it->second.GetReflection(refl->stage, refl->entryPoint, pipeline).stageIndex);
size_t size;
vt->GetShaderInfoAMD(Unwrap(dev), Unwrap(pipe), stageBit, VK_SHADER_INFO_TYPE_DISASSEMBLY_AMD,
&size, NULL);
rdcstr disasm;
disasm.resize(size);
vt->GetShaderInfoAMD(Unwrap(dev), Unwrap(pipe), stageBit, VK_SHADER_INFO_TYPE_DISASSEMBLY_AMD,
&size, (void *)disasm.data());
return disasm;
}
if(target == KHRExecutablePropertiesTarget && vt->GetPipelineExecutablePropertiesKHR)
{
if(pipeline == ResourceId())
{
return "; No pipeline specified, VK_KHR_pipeline_executable_properties disassembly is not "
"available\n"
"; Shader must be disassembled with a specific pipeline.";
}
CachePipelineExecutables(pipeline);
VkShaderStageFlagBits stageBit = VkShaderStageFlagBits(
1 << it->second.GetReflection(refl->stage, refl->entryPoint, pipeline).stageIndex);
const rdcarray<PipelineExecutables> &executables = m_PipelineExecutables[pipeline];
rdcstr disasm;
for(const PipelineExecutables &exec : executables)
{
// if this executable is associated with our stage, definitely include it. If this executable
// is associated with *no* stages, then also include it (since we don't know what it
// corresponds to)
if((exec.stages & stageBit) || exec.stages == 0)
{
disasm += "======== " + exec.name + " ========\n\n";
disasm += exec.description + "\n\n";
// statistics first
disasm += StringFormat::Fmt(
"==== Statistics ====\n\n"
"Subgroup Size: %u"
" // the subgroup size with which this executable is dispatched.\n",
exec.subgroupSize);
for(const VkPipelineExecutableStatisticKHR &stat : exec.statistics)
{
rdcstr value;
switch(stat.format)
{
case VK_PIPELINE_EXECUTABLE_STATISTIC_FORMAT_BOOL32_KHR:
value = stat.value.b32 ? "true" : "false";
break;
case VK_PIPELINE_EXECUTABLE_STATISTIC_FORMAT_INT64_KHR:
value = ToStr(stat.value.i64);
break;
case VK_PIPELINE_EXECUTABLE_STATISTIC_FORMAT_UINT64_KHR:
value = ToStr(stat.value.u64);
break;
case VK_PIPELINE_EXECUTABLE_STATISTIC_FORMAT_FLOAT64_KHR:
value = ToStr(stat.value.f64);
break;
case VK_PIPELINE_EXECUTABLE_STATISTIC_FORMAT_MAX_ENUM_KHR: value = "???"; break;
}
disasm +=
StringFormat::Fmt("%s: %s // %s\n", stat.name, value.c_str(), stat.description);
}
// then IRs
if(!exec.representations.empty())
disasm += "\n\n==== Internal Representations ====\n\n";
for(const VkPipelineExecutableInternalRepresentationKHR &ir : exec.representations)
{
disasm += "---- " + rdcstr(ir.name) + " ----\n\n";
disasm += "; " + rdcstr(ir.description) + "\n\n";
if(ir.isText)
{
char *str = (char *)ir.pData;
// should already be NULL terminated but let's be sure
str[ir.dataSize - 1] = 0;
disasm += str;
}
else
{
// canonical hexdump display
size_t bytesRemaining = ir.dataSize;
size_t offset = 0;
const byte *src = (const byte *)ir.pData;
while(bytesRemaining > 0)
{
uint8_t row[16];
const size_t copySize = RDCMIN(sizeof(row), bytesRemaining);
memcpy(row, src + offset, copySize);
disasm += StringFormat::Fmt("%08zx ", offset);
for(size_t b = 0; b < 16; b++)
{
if(b < bytesRemaining)
disasm += StringFormat::Fmt("%02hhx ", row[b]);
else
disasm += " ";
if(b == 7 || b == 15)
disasm += " ";
}
disasm += "|";
for(size_t b = 0; b < 16; b++)
{
if(b < bytesRemaining)
{
char c = (char)row[b];
if(isprint(c))
disasm.push_back(c);
else
disasm.push_back('.');
}
}
disasm += "|";
disasm += "\n";
offset += copySize;
bytesRemaining -= copySize;
}
disasm += StringFormat::Fmt("%08zx", offset);
}
}
}
}
return disasm;
}
return StringFormat::Fmt("; Invalid disassembly target %s", target.c_str());
}
void VulkanReplay::RenderCheckerboard(FloatVector dark, FloatVector light)
{
auto it = m_OutputWindows.find(m_ActiveWinID);
if(m_ActiveWinID == 0 || it == m_OutputWindows.end())
return;
OutputWindow &outw = it->second;
// if the swapchain failed to create, do nothing. We will try to recreate it
// again in CheckResizeOutputWindow (once per render 'frame')
if(outw.m_WindowSystem != WindowingSystem::Headless && outw.swap == VK_NULL_HANDLE)
return;
VkDevice dev = m_pDriver->GetDev();
VkCommandBuffer cmd = m_pDriver->GetNextCmd();
const VkDevDispatchTable *vt = ObjDisp(dev);
if(cmd == VK_NULL_HANDLE)
return;
VkCommandBufferBeginInfo beginInfo = {VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO, NULL,
VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT};
VkResult vkr = vt->BeginCommandBuffer(Unwrap(cmd), &beginInfo);
CheckVkResult(vkr);
uint32_t uboOffs = 0;
VkRenderPassBeginInfo rpbegin = {
VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
NULL,
Unwrap(outw.rp),
Unwrap(outw.fb),
{{
0,
0,
},
{m_DebugWidth, m_DebugHeight}},
0,
NULL,
};
vt->CmdBeginRenderPass(Unwrap(cmd), &rpbegin, VK_SUBPASS_CONTENTS_INLINE);
if(m_Overlay.m_CheckerPipeline != VK_NULL_HANDLE)
{
CheckerboardUBOData *data = (CheckerboardUBOData *)m_Overlay.m_CheckerUBO.Map(&uboOffs);
if(!data)
return;
data->BorderWidth = 0.0f;
data->RectPosition = Vec2f();
data->RectSize = Vec2f();
data->CheckerSquareDimension = 64.0f;
data->InnerColor = Vec4f();
data->PrimaryColor = ConvertSRGBToLinear(light);
data->SecondaryColor = ConvertSRGBToLinear(dark);
m_Overlay.m_CheckerUBO.Unmap();
vt->CmdBindPipeline(Unwrap(cmd), VK_PIPELINE_BIND_POINT_GRAPHICS,
outw.dsimg == VK_NULL_HANDLE ? Unwrap(m_Overlay.m_CheckerPipeline)
: Unwrap(m_Overlay.m_CheckerMSAAPipeline));
vt->CmdBindDescriptorSets(Unwrap(cmd), VK_PIPELINE_BIND_POINT_GRAPHICS,
Unwrap(m_Overlay.m_CheckerPipeLayout), 0, 1,
UnwrapPtr(m_Overlay.m_CheckerDescSet), 1, &uboOffs);
VkViewport viewport = {0.0f, 0.0f, (float)m_DebugWidth, (float)m_DebugHeight, 0.0f, 1.0f};
vt->CmdSetViewport(Unwrap(cmd), 0, 1, &viewport);
vt->CmdDraw(Unwrap(cmd), 4, 1, 0, 0);
if(m_pDriver->GetDriverInfo().QualcommLeakingUBOOffsets())
{
uboOffs = 0;
vt->CmdBindDescriptorSets(Unwrap(cmd), VK_PIPELINE_BIND_POINT_GRAPHICS,
Unwrap(m_Overlay.m_CheckerPipeLayout), 0, 1,
UnwrapPtr(m_Overlay.m_CheckerDescSet), 1, &uboOffs);
}
}
else
{
// some mobile chips fail to create the checkerboard pipeline. Use an alternate approach with
// CmdClearAttachment and many rects.
VkClearAttachment lightCol = {
VK_IMAGE_ASPECT_COLOR_BIT, 0, {{{light.x, light.y, light.z, light.w}}}};
VkClearAttachment darkCol = {VK_IMAGE_ASPECT_COLOR_BIT, 0, {{{dark.x, dark.y, dark.z, dark.w}}}};
VkClearRect fullRect = {{
{0, 0},
{outw.width, outw.height},
},
0,
1};
vt->CmdClearAttachments(Unwrap(cmd), 1, &lightCol, 1, &fullRect);
rdcarray<VkClearRect> squares;
for(int32_t y = 0; y < (int32_t)outw.height; y += 128)
{
for(int32_t x = 0; x < (int32_t)outw.width; x += 128)
{
VkClearRect square = {{
{x, y},
{64, 64},
},
0,
1};
squares.push_back(square);
square.rect.offset.x += 64;
square.rect.offset.y += 64;
squares.push_back(square);
}
}
vt->CmdClearAttachments(Unwrap(cmd), 1, &darkCol, (uint32_t)squares.size(), squares.data());
}
vt->CmdEndRenderPass(Unwrap(cmd));
vkr = vt->EndCommandBuffer(Unwrap(cmd));
CheckVkResult(vkr);
if(Vulkan_Debug_SingleSubmitFlushing())
m_pDriver->SubmitCmds();
}
void VulkanReplay::RenderHighlightBox(float w, float h, float scale)
{
auto it = m_OutputWindows.find(m_ActiveWinID);
if(m_ActiveWinID == 0 || it == m_OutputWindows.end())
return;
OutputWindow &outw = it->second;
// if the swapchain failed to create, do nothing. We will try to recreate it
// again in CheckResizeOutputWindow (once per render 'frame')
if(outw.m_WindowSystem != WindowingSystem::Headless && outw.swap == VK_NULL_HANDLE)
return;
VkDevice dev = m_pDriver->GetDev();
VkCommandBuffer cmd = m_pDriver->GetNextCmd();
const VkDevDispatchTable *vt = ObjDisp(dev);
if(cmd == VK_NULL_HANDLE)
return;
VkCommandBufferBeginInfo beginInfo = {VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO, NULL,
VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT};
VkResult vkr = vt->BeginCommandBuffer(Unwrap(cmd), &beginInfo);
CheckVkResult(vkr);
{
VkRenderPassBeginInfo rpbegin = {
VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
NULL,
Unwrap(outw.rp),
Unwrap(outw.fb),
{{
0,
0,
},
{m_DebugWidth, m_DebugHeight}},
0,
NULL,
};
vt->CmdBeginRenderPass(Unwrap(cmd), &rpbegin, VK_SUBPASS_CONTENTS_INLINE);
VkClearAttachment black = {VK_IMAGE_ASPECT_COLOR_BIT, 0, {{{0.0f, 0.0f, 0.0f, 1.0f}}}};
VkClearAttachment white = {VK_IMAGE_ASPECT_COLOR_BIT, 0, {{{1.0f, 1.0f, 1.0f, 1.0f}}}};
uint32_t sz = uint32_t(scale);
VkOffset2D tl = {int32_t(w / 2.0f + 0.5f), int32_t(h / 2.0f + 0.5f)};
VkClearRect rect[4] = {
{{
{tl.x, tl.y},
{1, sz},
},
0,
1},
{{
{tl.x + (int32_t)sz, tl.y},
{1, sz + 1},
},
0,
1},
{{
{tl.x, tl.y},
{sz, 1},
},
0,
1},
{{
{tl.x, tl.y + (int32_t)sz},
{sz, 1},
},
0,
1},
};
// inner
vt->CmdClearAttachments(Unwrap(cmd), 1, &white, 4, rect);
rect[0].rect.offset.x--;
rect[1].rect.offset.x++;
rect[2].rect.offset.x--;
rect[3].rect.offset.x--;
rect[0].rect.offset.y--;
rect[1].rect.offset.y--;
rect[2].rect.offset.y--;
rect[3].rect.offset.y++;
rect[0].rect.extent.height += 2;
rect[1].rect.extent.height += 2;
rect[2].rect.extent.width += 2;
rect[3].rect.extent.width += 2;
// outer
vt->CmdClearAttachments(Unwrap(cmd), 1, &black, 4, rect);
vt->CmdEndRenderPass(Unwrap(cmd));
}
vkr = vt->EndCommandBuffer(Unwrap(cmd));
CheckVkResult(vkr);
if(Vulkan_Debug_SingleSubmitFlushing())
m_pDriver->SubmitCmds();
}
void VulkanReplay::GetBufferData(ResourceId buff, uint64_t offset, uint64_t len, bytebuf &retData)
{
GetDebugManager()->GetBufferData(buff, offset, len, retData);
}
void VulkanReplay::FileChanged()
{
}
void VulkanReplay::GetInitialDriverVersion()
{
RDCEraseEl(m_DriverInfo);
VkInstance inst = m_pDriver->GetInstance();
uint32_t count;
VkResult vkr = ObjDisp(inst)->EnumeratePhysicalDevices(Unwrap(inst), &count, NULL);
if(vkr != VK_SUCCESS)
{
RDCERR("Couldn't enumerate physical devices");
return;
}
if(count == 0)
{
RDCERR("No physical devices available");
}
count = 1;
VkPhysicalDevice firstDevice = VK_NULL_HANDLE;
vkr = ObjDisp(inst)->EnumeratePhysicalDevices(Unwrap(inst), &count, &firstDevice);
// incomplete is expected if multiple GPUs are present, and we're just grabbing the first
if(vkr != VK_SUCCESS && vkr != VK_INCOMPLETE)
{
RDCERR("Couldn't fetch first physical device");
return;
}
VkPhysicalDeviceProperties props = {};
ObjDisp(inst)->GetPhysicalDeviceProperties(firstDevice, &props);
VkPhysicalDeviceDriverProperties driverProps = {};
GetPhysicalDeviceDriverProperties(ObjDisp(inst), firstDevice, driverProps);
SetDriverInformation(props, driverProps);
}
void VulkanReplay::SetDriverInformation(const VkPhysicalDeviceProperties &props,
const VkPhysicalDeviceDriverProperties &driverProps)
{
VkDriverInfo info(props, driverProps);
m_DriverInfo.vendor = info.Vendor();
rdcstr versionString =
StringFormat::Fmt("%s %u.%u.%u", props.deviceName, info.Major(), info.Minor(), info.Patch());
versionString.resize(RDCMIN(versionString.size(), ARRAY_COUNT(m_DriverInfo.version) - 1));
memcpy(m_DriverInfo.version, versionString.c_str(), versionString.size());
}
static TextureSwizzle Convert(VkComponentSwizzle src, int i)
{
switch(src)
{
default: RDCWARN("Unexpected component swizzle value %d", (int)src); DELIBERATE_FALLTHROUGH();
case VK_COMPONENT_SWIZZLE_IDENTITY: break;
case VK_COMPONENT_SWIZZLE_ZERO: return TextureSwizzle::Zero;
case VK_COMPONENT_SWIZZLE_ONE: return TextureSwizzle::One;
case VK_COMPONENT_SWIZZLE_R: return TextureSwizzle::Red;
case VK_COMPONENT_SWIZZLE_G: return TextureSwizzle::Green;
case VK_COMPONENT_SWIZZLE_B: return TextureSwizzle::Blue;
case VK_COMPONENT_SWIZZLE_A: return TextureSwizzle::Alpha;
}
return TextureSwizzle(uint32_t(TextureSwizzle::Red) + i);
}
static void Convert(TextureSwizzle4 &dst, VkComponentMapping src)
{
dst.red = Convert(src.r, 0);
dst.green = Convert(src.g, 1);
dst.blue = Convert(src.b, 2);
dst.alpha = Convert(src.a, 3);
}
void VulkanReplay::SavePipelineState(uint32_t eventId)
{
if(!m_VulkanPipelineState)
return;
const VulkanRenderState &state = m_pDriver->m_RenderState;
VulkanCreationInfo &c = m_pDriver->m_CreationInfo;
VKPipe::State &ret = *m_VulkanPipelineState;
VulkanResourceManager *rm = m_pDriver->GetResourceManager();
VkMarkerRegion::Begin(StringFormat::Fmt("FetchShaderFeedback for %u", eventId));
FetchShaderFeedback(eventId);
VkMarkerRegion::End();
{
// reset the pipeline state, but keep the descriptor set arrays. This prevents needless
// reallocations, we'll ensure that descriptors are fully overwritten below.
rdcarray<VKPipe::DescriptorSet> graphicsDescriptors;
rdcarray<VKPipe::DescriptorSet> computeDescriptors;
ret.graphics.descriptorSets.swap(graphicsDescriptors);
ret.compute.descriptorSets.swap(computeDescriptors);
ret = VKPipe::State();
ret.graphics.descriptorSets.swap(graphicsDescriptors);
ret.compute.descriptorSets.swap(computeDescriptors);
}
ret.pushconsts.resize(state.pushConstSize);
memcpy(ret.pushconsts.data(), state.pushconsts, state.pushConstSize);
// General pipeline properties
ret.compute.pipelineResourceId = rm->GetUnreplacedOriginalID(state.compute.pipeline);
ret.graphics.pipelineResourceId = rm->GetUnreplacedOriginalID(state.graphics.pipeline);
if(state.compute.pipeline != ResourceId())
{
const VulkanCreationInfo::Pipeline &p = c.m_Pipeline[state.compute.pipeline];
ret.compute.pipelineComputeLayoutResourceId = rm->GetOriginalID(p.compLayout);
ret.compute.flags = p.flags;
VKPipe::Shader &stage = ret.computeShader;
int i = 5; // 5 is the CS idx (VS, TCS, TES, GS, FS, CS)
{
stage.resourceId = rm->GetUnreplacedOriginalID(p.shaders[i].module);
stage.entryPoint = p.shaders[i].entryPoint;
stage.stage = ShaderStage::Compute;
if(p.shaders[i].mapping)
stage.bindpointMapping = *p.shaders[i].mapping;
if(p.shaders[i].refl)
stage.reflection = p.shaders[i].refl;
stage.pushConstantRangeByteOffset = stage.pushConstantRangeByteSize = 0;
for(const VkPushConstantRange &pr : c.m_PipelineLayout[p.compLayout].pushRanges)
{
if(pr.stageFlags & VK_SHADER_STAGE_COMPUTE_BIT)
{
stage.pushConstantRangeByteOffset = pr.offset;
stage.pushConstantRangeByteSize = pr.size;
break;
}
}
stage.requiredSubgroupSize = p.shaders[i].requiredSubgroupSize;
stage.specializationData.clear();
// set up the defaults
if(p.shaders[i].mapping && p.shaders[i].refl)
{
for(size_t cb = 0; cb < p.shaders[i].mapping->constantBlocks.size(); cb++)
{
if(p.shaders[i].mapping->constantBlocks[cb].bindset == SpecializationConstantBindSet)
{
for(const ShaderConstant &sc : p.shaders[i].refl->constantBlocks[cb].variables)
{
stage.specializationData.resize_for_index(sc.byteOffset + sizeof(uint64_t));
memcpy(stage.specializationData.data() + sc.byteOffset, &sc.defaultValue,
sizeof(uint64_t));
}
break;
}
}
}
// apply any specializations
for(const SpecConstant &s : p.shaders[i].specialization)
{
int32_t idx = p.shaders[i].patchData->specIDs.indexOf(s.specID);
if(idx == -1)
{
RDCWARN("Couldn't find offset for spec ID %u", s.specID);
continue;
}
size_t offs = idx * sizeof(uint64_t);
stage.specializationData.resize_for_index(offs + sizeof(uint64_t));
memcpy(stage.specializationData.data() + offs, &s.value, s.dataSize);
}
if(p.shaders[i].patchData)
stage.specializationIds = p.shaders[i].patchData->specIDs;
}
}
else
{
ret.compute.pipelineComputeLayoutResourceId = ResourceId();
ret.compute.flags = 0;
ret.computeShader = VKPipe::Shader();
}
if(state.graphics.pipeline != ResourceId())
{
const VulkanCreationInfo::Pipeline &p = c.m_Pipeline[state.graphics.pipeline];
ret.graphics.pipelinePreRastLayoutResourceId = rm->GetOriginalID(p.vertLayout);
ret.graphics.pipelineFragmentLayoutResourceId = rm->GetOriginalID(p.fragLayout);
ret.graphics.flags = p.flags;
// Input Assembly
ret.inputAssembly.indexBuffer.resourceId = rm->GetOriginalID(state.ibuffer.buf);
ret.inputAssembly.indexBuffer.byteOffset = state.ibuffer.offs;
ret.inputAssembly.indexBuffer.byteStride = state.ibuffer.bytewidth;
ret.inputAssembly.primitiveRestartEnable = state.primRestartEnable != VK_FALSE;
ret.inputAssembly.topology =
MakePrimitiveTopology(state.primitiveTopology, state.patchControlPoints);
// Vertex Input
ret.vertexInput.attributes.resize(state.vertexAttributes.size());
for(size_t i = 0; i < state.vertexAttributes.size(); i++)
{
ret.vertexInput.attributes[i].location = state.vertexAttributes[i].location;
ret.vertexInput.attributes[i].binding = state.vertexAttributes[i].binding;
ret.vertexInput.attributes[i].byteOffset = state.vertexAttributes[i].offset;
ret.vertexInput.attributes[i].format = MakeResourceFormat(state.vertexAttributes[i].format);
}
ret.vertexInput.bindings.resize(state.vertexBindings.size());
for(const VkVertexInputBindingDescription2EXT &b : state.vertexBindings)
{
ret.vertexInput.bindings.resize_for_index(b.binding);
ret.vertexInput.bindings[b.binding].vertexBufferBinding = b.binding;
ret.vertexInput.bindings[b.binding].perInstance = b.inputRate == VK_VERTEX_INPUT_RATE_INSTANCE;
ret.vertexInput.bindings[b.binding].instanceDivisor = b.divisor;
}
ret.vertexInput.vertexBuffers.resize(state.vbuffers.size());
for(size_t i = 0; i < state.vbuffers.size(); i++)
{
ret.vertexInput.vertexBuffers[i].resourceId = rm->GetOriginalID(state.vbuffers[i].buf);
ret.vertexInput.vertexBuffers[i].byteOffset = state.vbuffers[i].offs;
ret.vertexInput.vertexBuffers[i].byteStride = (uint32_t)state.vbuffers[i].stride;
ret.vertexInput.vertexBuffers[i].byteSize = (uint32_t)state.vbuffers[i].size;
}
// Shader Stages
VKPipe::Shader *stages[] = {
&ret.vertexShader,
&ret.tessControlShader,
&ret.tessEvalShader,
&ret.geometryShader,
&ret.fragmentShader,
// compute
NULL,
&ret.taskShader,
&ret.meshShader,
};
for(size_t i = 0; i < ARRAY_COUNT(stages); i++)
{
if(stages[i] == NULL)
continue;
stages[i]->resourceId = rm->GetUnreplacedOriginalID(p.shaders[i].module);
stages[i]->entryPoint = p.shaders[i].entryPoint;
stages[i]->stage = StageFromIndex(i);
if(p.shaders[i].mapping)
stages[i]->bindpointMapping = *p.shaders[i].mapping;
if(p.shaders[i].refl)
stages[i]->reflection = p.shaders[i].refl;
stages[i]->pushConstantRangeByteOffset = stages[i]->pushConstantRangeByteSize = 0;
// don't have to handle separate vert/frag layouts as push constant ranges must be identical
for(const VkPushConstantRange &pr : c.m_PipelineLayout[p.vertLayout].pushRanges)
{
if(pr.stageFlags & ShaderMaskFromIndex(i))
{
stages[i]->pushConstantRangeByteOffset = pr.offset;
stages[i]->pushConstantRangeByteSize = pr.size;
break;
}
}
stages[i]->specializationData.clear();
stages[i]->requiredSubgroupSize = p.shaders[i].requiredSubgroupSize;
// set up the defaults
if(p.shaders[i].mapping && p.shaders[i].refl)
{
for(size_t cb = 0; cb < p.shaders[i].mapping->constantBlocks.size(); cb++)
{
if(p.shaders[i].mapping->constantBlocks[cb].bindset == SpecializationConstantBindSet)
{
for(const ShaderConstant &sc : p.shaders[i].refl->constantBlocks[cb].variables)
{
stages[i]->specializationData.resize_for_index(sc.byteOffset + sizeof(uint64_t));
memcpy(stages[i]->specializationData.data() + sc.byteOffset, &sc.defaultValue,
sizeof(uint64_t));
}
break;
}
}
}
// apply any specializations
for(const SpecConstant &s : p.shaders[i].specialization)
{
int32_t idx = p.shaders[i].patchData->specIDs.indexOf(s.specID);
if(idx == -1)
{
RDCWARN("Couldn't find offset for spec ID %u", s.specID);
continue;
}
size_t offs = idx * sizeof(uint64_t);
stages[i]->specializationData.resize_for_index(offs + sizeof(uint64_t));
memcpy(stages[i]->specializationData.data() + offs, &s.value, s.dataSize);
}
if(p.shaders[i].patchData)
stages[i]->specializationIds = p.shaders[i].patchData->specIDs;
}
// Tessellation
ret.tessellation.numControlPoints = p.patchControlPoints;
ret.tessellation.domainOriginUpperLeft =
state.domainOrigin == VK_TESSELLATION_DOMAIN_ORIGIN_UPPER_LEFT;
ret.transformFeedback.rasterizedStream = state.rasterStream;
// Transform feedback
ret.transformFeedback.buffers.resize(state.xfbbuffers.size());
for(size_t i = 0; i < state.xfbbuffers.size(); i++)
{
ret.transformFeedback.buffers[i].bufferResourceId = rm->GetOriginalID(state.xfbbuffers[i].buf);
ret.transformFeedback.buffers[i].byteOffset = state.xfbbuffers[i].offs;
ret.transformFeedback.buffers[i].byteSize = state.xfbbuffers[i].size;
ret.transformFeedback.buffers[i].active = false;
ret.transformFeedback.buffers[i].counterBufferResourceId = ResourceId();
ret.transformFeedback.buffers[i].counterBufferOffset = 0;
if(i >= state.firstxfbcounter)
{
size_t xfb = i - state.firstxfbcounter;
if(xfb < state.xfbcounters.size())
{
ret.transformFeedback.buffers[i].active = true;
ret.transformFeedback.buffers[i].counterBufferResourceId =
rm->GetOriginalID(state.xfbcounters[xfb].buf);
ret.transformFeedback.buffers[i].counterBufferOffset = state.xfbcounters[xfb].offs;
}
}
}
// Viewport/Scissors
size_t numViewScissors = state.views.size();
ret.viewportScissor.viewportScissors.resize(numViewScissors);
for(size_t i = 0; i < numViewScissors; i++)
{
if(i < state.views.size())
{
ret.viewportScissor.viewportScissors[i].vp.x = state.views[i].x;
ret.viewportScissor.viewportScissors[i].vp.y = state.views[i].y;
ret.viewportScissor.viewportScissors[i].vp.width = state.views[i].width;
ret.viewportScissor.viewportScissors[i].vp.height = state.views[i].height;
ret.viewportScissor.viewportScissors[i].vp.minDepth = state.views[i].minDepth;
ret.viewportScissor.viewportScissors[i].vp.maxDepth = state.views[i].maxDepth;
}
else
{
RDCEraseEl(ret.viewportScissor.viewportScissors[i].vp);
}
if(i < state.scissors.size())
{
ret.viewportScissor.viewportScissors[i].scissor.x = state.scissors[i].offset.x;
ret.viewportScissor.viewportScissors[i].scissor.y = state.scissors[i].offset.y;
ret.viewportScissor.viewportScissors[i].scissor.width = state.scissors[i].extent.width;
ret.viewportScissor.viewportScissors[i].scissor.height = state.scissors[i].extent.height;
}
else
{
RDCEraseEl(ret.viewportScissor.viewportScissors[i].scissor);
}
}
{
ret.viewportScissor.discardRectangles.resize(p.discardRectangles.size());
for(size_t i = 0; i < p.discardRectangles.size() && i < state.discardRectangles.size(); i++)
{
ret.viewportScissor.discardRectangles[i].x = state.discardRectangles[i].offset.x;
ret.viewportScissor.discardRectangles[i].y = state.discardRectangles[i].offset.y;
ret.viewportScissor.discardRectangles[i].width = state.discardRectangles[i].extent.width;
ret.viewportScissor.discardRectangles[i].height = state.discardRectangles[i].extent.height;
}
ret.viewportScissor.discardRectanglesExclusive =
(p.discardMode == VK_DISCARD_RECTANGLE_MODE_EXCLUSIVE_EXT);
}
{
ret.viewportScissor.depthNegativeOneToOne = state.negativeOneToOne != VK_FALSE;
}
// Rasterizer
ret.rasterizer.depthClampEnable = state.depthClampEnable != VK_FALSE;
ret.rasterizer.depthClipEnable = state.depthClipEnable != VK_FALSE;
ret.rasterizer.rasterizerDiscardEnable = state.rastDiscardEnable != VK_FALSE;
ret.rasterizer.frontCCW = state.frontFace == VK_FRONT_FACE_COUNTER_CLOCKWISE;
ret.rasterizer.conservativeRasterization = ConservativeRaster::Disabled;
switch(state.conservativeRastMode)
{
case VK_CONSERVATIVE_RASTERIZATION_MODE_UNDERESTIMATE_EXT:
ret.rasterizer.conservativeRasterization = ConservativeRaster::Underestimate;
break;
case VK_CONSERVATIVE_RASTERIZATION_MODE_OVERESTIMATE_EXT:
ret.rasterizer.conservativeRasterization = ConservativeRaster::Overestimate;
break;
default: break;
}
ret.rasterizer.pipelineShadingRate = {state.pipelineShadingRate.width,
state.pipelineShadingRate.height};
ShadingRateCombiner combiners[2] = {};
for(int i = 0; i < 2; i++)
{
switch(state.shadingRateCombiners[i])
{
default:
case VK_FRAGMENT_SHADING_RATE_COMBINER_OP_KEEP_KHR:
combiners[i] = ShadingRateCombiner::Keep;
break;
case VK_FRAGMENT_SHADING_RATE_COMBINER_OP_REPLACE_KHR:
combiners[i] = ShadingRateCombiner::Replace;
break;
case VK_FRAGMENT_SHADING_RATE_COMBINER_OP_MIN_KHR:
combiners[i] = ShadingRateCombiner::Min;
break;
case VK_FRAGMENT_SHADING_RATE_COMBINER_OP_MAX_KHR:
combiners[i] = ShadingRateCombiner::Max;
break;
case VK_FRAGMENT_SHADING_RATE_COMBINER_OP_MUL_KHR:
combiners[i] = ShadingRateCombiner::Multiply;
break;
}
}
ret.rasterizer.shadingRateCombiners = {combiners[0], combiners[1]};
ret.rasterizer.lineRasterMode = LineRaster::Default;
// "VK_LINE_RASTERIZATION_MODE_DEFAULT_HKR is equivalent to
// VK_LINE_RASTERIZATION_MODE_RECTANGULAR_KHR if VkPhysicalDeviceLimits::strictLines is VK_TRUE"
if(m_pDriver->GetDeviceProps().limits.strictLines)
ret.rasterizer.lineRasterMode = LineRaster::Rectangular;
switch(state.lineRasterMode)
{
case VK_LINE_RASTERIZATION_MODE_RECTANGULAR_KHR:
ret.rasterizer.lineRasterMode = LineRaster::Rectangular;
break;
case VK_LINE_RASTERIZATION_MODE_BRESENHAM_KHR:
ret.rasterizer.lineRasterMode = LineRaster::Bresenham;
break;
case VK_LINE_RASTERIZATION_MODE_RECTANGULAR_SMOOTH_KHR:
ret.rasterizer.lineRasterMode = LineRaster::RectangularSmooth;
break;
default: break;
}
ret.rasterizer.lineStippleFactor = 0; // stippled line disable
ret.rasterizer.lineStipplePattern = 0;
if(state.stippledLineEnable)
{
ret.rasterizer.lineStippleFactor = state.stippleFactor;
ret.rasterizer.lineStipplePattern = state.stipplePattern;
}
ret.rasterizer.extraPrimitiveOverestimationSize = state.primOverestimationSize;
switch(state.polygonMode)
{
case VK_POLYGON_MODE_POINT: ret.rasterizer.fillMode = FillMode::Point; break;
case VK_POLYGON_MODE_LINE: ret.rasterizer.fillMode = FillMode::Wireframe; break;
case VK_POLYGON_MODE_FILL: ret.rasterizer.fillMode = FillMode::Solid; break;
default:
ret.rasterizer.fillMode = FillMode::Solid;
RDCERR("Unexpected value for FillMode %x", state.polygonMode);
break;
}
switch(state.cullMode)
{
case VK_CULL_MODE_NONE: ret.rasterizer.cullMode = CullMode::NoCull; break;
case VK_CULL_MODE_FRONT_BIT: ret.rasterizer.cullMode = CullMode::Front; break;
case VK_CULL_MODE_BACK_BIT: ret.rasterizer.cullMode = CullMode::Back; break;
case VK_CULL_MODE_FRONT_AND_BACK: ret.rasterizer.cullMode = CullMode::FrontAndBack; break;
default:
ret.rasterizer.cullMode = CullMode::NoCull;
RDCERR("Unexpected value for CullMode %x", state.cullMode);
break;
}
ret.rasterizer.provokingVertexFirst =
state.provokingVertexMode == VK_PROVOKING_VERTEX_MODE_FIRST_VERTEX_EXT;
ret.rasterizer.depthBiasEnable = state.depthBiasEnable != VK_FALSE;
ret.rasterizer.depthBias = state.bias.depth;
ret.rasterizer.depthBiasClamp = state.bias.biasclamp;
ret.rasterizer.slopeScaledDepthBias = state.bias.slope;
ret.rasterizer.lineWidth = state.lineWidth;
// MSAA
ret.multisample.rasterSamples = state.rastSamples;
ret.multisample.sampleShadingEnable = p.sampleShadingEnable;
ret.multisample.minSampleShading = p.minSampleShading;
ret.multisample.sampleMask = state.sampleMask[0];
ret.multisample.sampleLocations.customLocations.clear();
if(state.sampleLocEnable)
{
ret.multisample.sampleLocations.gridWidth = state.sampleLocations.gridSize.width;
ret.multisample.sampleLocations.gridHeight = state.sampleLocations.gridSize.height;
ret.multisample.sampleLocations.customLocations.reserve(state.sampleLocations.locations.size());
for(const VkSampleLocationEXT &loc : state.sampleLocations.locations)
{
ret.multisample.sampleLocations.customLocations.push_back({loc.x, loc.y, 0.0f, 0.0f});
}
}
// Color Blend
ret.colorBlend.alphaToCoverageEnable = state.alphaToCoverageEnable != VK_FALSE;
ret.colorBlend.alphaToOneEnable = state.alphaToOneEnable != VK_FALSE;
// find size if no static pipeline state
size_t numAttach = RDCMAX(state.colorBlendEnable.size(), state.colorBlendEquation.size());
numAttach = RDCMAX(numAttach, state.colorWriteEnable.size());
numAttach = RDCMAX(numAttach, state.colorWriteMask.size());
ret.colorBlend.blends.resize(numAttach);
for(size_t i = 0; i < numAttach; i++)
{
ret.colorBlend.blends[i].enabled =
(i < state.colorBlendEnable.size()) ? state.colorBlendEnable[i] != VK_FALSE : VK_FALSE;
// due to shared structs, this is slightly duplicated - Vulkan doesn't have separate states
// for logic operations
ret.colorBlend.blends[i].logicOperationEnabled = state.logicOpEnable != VK_FALSE;
ret.colorBlend.blends[i].logicOperation = MakeLogicOp(state.logicOp);
if(ret.colorBlend.blends[i].enabled && i < state.colorBlendEquation.size())
{
ret.colorBlend.blends[i].colorBlend.source =
MakeBlendMultiplier(state.colorBlendEquation[i].srcColorBlendFactor);
ret.colorBlend.blends[i].colorBlend.destination =
MakeBlendMultiplier(state.colorBlendEquation[i].dstColorBlendFactor);
ret.colorBlend.blends[i].colorBlend.operation =
MakeBlendOp(state.colorBlendEquation[i].colorBlendOp);
ret.colorBlend.blends[i].alphaBlend.source =
MakeBlendMultiplier(state.colorBlendEquation[i].srcAlphaBlendFactor);
ret.colorBlend.blends[i].alphaBlend.destination =
MakeBlendMultiplier(state.colorBlendEquation[i].dstAlphaBlendFactor);
ret.colorBlend.blends[i].alphaBlend.operation =
MakeBlendOp(state.colorBlendEquation[i].alphaBlendOp);
}
else
{
ret.colorBlend.blends[i].colorBlend.source = MakeBlendMultiplier(VK_BLEND_FACTOR_ZERO);
ret.colorBlend.blends[i].colorBlend.destination = MakeBlendMultiplier(VK_BLEND_FACTOR_ZERO);
ret.colorBlend.blends[i].colorBlend.operation = MakeBlendOp(VK_BLEND_OP_ADD);
ret.colorBlend.blends[i].alphaBlend.source = MakeBlendMultiplier(VK_BLEND_FACTOR_ZERO);
ret.colorBlend.blends[i].alphaBlend.destination = MakeBlendMultiplier(VK_BLEND_FACTOR_ZERO);
ret.colorBlend.blends[i].alphaBlend.operation = MakeBlendOp(VK_BLEND_OP_ADD);
}
ret.colorBlend.blends[i].writeMask =
(i < state.colorWriteMask.size()) ? (uint8_t)state.colorWriteMask[i] : 0;
if(i < state.colorWriteEnable.size() && !state.colorWriteEnable[i])
ret.colorBlend.blends[i].writeMask = 0;
}
ret.colorBlend.blendFactor = state.blendConst;
// Depth Stencil
ret.depthStencil.depthTestEnable = state.depthTestEnable != VK_FALSE;
ret.depthStencil.depthWriteEnable = state.depthWriteEnable != VK_FALSE;
ret.depthStencil.depthBoundsEnable = state.depthBoundsTestEnable != VK_FALSE;
ret.depthStencil.depthFunction = MakeCompareFunc(state.depthCompareOp);
ret.depthStencil.stencilTestEnable = state.stencilTestEnable != VK_FALSE;
ret.depthStencil.frontFace.passOperation = MakeStencilOp(state.front.passOp);
ret.depthStencil.frontFace.failOperation = MakeStencilOp(state.front.failOp);
ret.depthStencil.frontFace.depthFailOperation = MakeStencilOp(state.front.depthFailOp);
ret.depthStencil.frontFace.function = MakeCompareFunc(state.front.compareOp);
ret.depthStencil.backFace.passOperation = MakeStencilOp(state.back.passOp);
ret.depthStencil.backFace.failOperation = MakeStencilOp(state.back.failOp);
ret.depthStencil.backFace.depthFailOperation = MakeStencilOp(state.back.depthFailOp);
ret.depthStencil.backFace.function = MakeCompareFunc(state.back.compareOp);
ret.depthStencil.minDepthBounds = state.mindepth;
ret.depthStencil.maxDepthBounds = state.maxdepth;
ret.depthStencil.frontFace.reference = state.front.ref;
ret.depthStencil.frontFace.compareMask = state.front.compare;
ret.depthStencil.frontFace.writeMask = state.front.write;
ret.depthStencil.backFace.reference = state.back.ref;
ret.depthStencil.backFace.compareMask = state.back.compare;
ret.depthStencil.backFace.writeMask = state.back.write;
}
else
{
ret.graphics.pipelinePreRastLayoutResourceId = ResourceId();
ret.graphics.pipelineFragmentLayoutResourceId = ResourceId();
ret.graphics.flags = 0;
ret.vertexInput.attributes.clear();
ret.vertexInput.bindings.clear();
ret.vertexInput.vertexBuffers.clear();
VKPipe::Shader *stages[] = {
&ret.vertexShader, &ret.tessControlShader, &ret.tessEvalShader, &ret.geometryShader,
&ret.fragmentShader, &ret.taskShader, &ret.meshShader,
};
for(size_t i = 0; i < ARRAY_COUNT(stages); i++)
*stages[i] = VKPipe::Shader();
ret.viewportScissor.viewportScissors.clear();
ret.viewportScissor.discardRectangles.clear();
ret.viewportScissor.discardRectanglesExclusive = true;
ret.viewportScissor.depthNegativeOneToOne = false;
ret.colorBlend.blends.clear();
}
if(state.dynamicRendering.active)
{
VKPipe::RenderPass &rpState = ret.currentPass.renderpass;
VKPipe::Framebuffer &fbState = ret.currentPass.framebuffer;
const VulkanRenderState::DynamicRendering &dyn = state.dynamicRendering;
rpState.dynamic = true;
rpState.suspended = dyn.suspended;
rpState.feedbackLoop = false;
rpState.resourceId = ResourceId();
rpState.subpass = 0;
rpState.fragmentDensityOffsets.clear();
rpState.tileOnlyMSAASampleCount = 0;
fbState.resourceId = ResourceId();
// dynamic rendering does not provide a framebuffer dimension, it's implicit from the image
// views
fbState.width = 0;
fbState.height = 0;
fbState.layers = dyn.layerCount;
fbState.attachments.clear();
rpState.inputAttachments.clear();
rpState.colorAttachments.clear();
rpState.resolveAttachments.clear();
size_t attIdx = 0;
for(size_t i = 0; i < dyn.color.size(); i++)
{
fbState.attachments.push_back({});
ResourceId viewid = GetResID(dyn.color[i].imageView);
if(viewid != ResourceId())
{
fbState.attachments.back().viewResourceId = rm->GetOriginalID(viewid);
ret.currentPass.framebuffer.attachments[attIdx].imageResourceId =
rm->GetOriginalID(c.m_ImageView[viewid].image);
fbState.attachments.back().viewFormat = MakeResourceFormat(c.m_ImageView[viewid].format);
fbState.attachments.back().firstMip = c.m_ImageView[viewid].range.baseMipLevel;
fbState.attachments.back().firstSlice = c.m_ImageView[viewid].range.baseArrayLayer;
fbState.attachments.back().numMips = c.m_ImageView[viewid].range.levelCount;
fbState.attachments.back().numSlices = c.m_ImageView[viewid].range.layerCount;
Convert(fbState.attachments.back().swizzle, c.m_ImageView[viewid].componentMapping);
}
else
{
fbState.attachments.back().viewResourceId = ResourceId();
fbState.attachments.back().imageResourceId = ResourceId();
fbState.attachments.back().firstMip = 0;
fbState.attachments.back().firstSlice = 0;
fbState.attachments.back().numMips = 1;
fbState.attachments.back().numSlices = 1;
}
rpState.colorAttachments.push_back(uint32_t(attIdx++));
if(dyn.color[i].resolveMode && dyn.color[i].resolveImageView != VK_NULL_HANDLE)
{
fbState.attachments.push_back({});
viewid = GetResID(dyn.color[i].resolveImageView);
fbState.attachments.back().viewResourceId = rm->GetOriginalID(viewid);
ret.currentPass.framebuffer.attachments[attIdx].imageResourceId =
rm->GetOriginalID(c.m_ImageView[viewid].image);
fbState.attachments.back().viewFormat = MakeResourceFormat(c.m_ImageView[viewid].format);
fbState.attachments.back().firstMip = c.m_ImageView[viewid].range.baseMipLevel;
fbState.attachments.back().firstSlice = c.m_ImageView[viewid].range.baseArrayLayer;
fbState.attachments.back().numMips = c.m_ImageView[viewid].range.levelCount;
fbState.attachments.back().numSlices = c.m_ImageView[viewid].range.layerCount;
Convert(fbState.attachments.back().swizzle, c.m_ImageView[viewid].componentMapping);
rpState.resolveAttachments.push_back(uint32_t(attIdx++));
}
}
if(dyn.depth.imageView != VK_NULL_HANDLE || dyn.stencil.imageView != VK_NULL_HANDLE)
{
fbState.attachments.push_back({});
ResourceId viewid = GetResID(dyn.depth.imageView);
if(dyn.depth.imageView == VK_NULL_HANDLE)
viewid = GetResID(dyn.stencil.imageView);
fbState.attachments.back().viewResourceId = rm->GetOriginalID(viewid);
ret.currentPass.framebuffer.attachments[attIdx].imageResourceId =
rm->GetOriginalID(c.m_ImageView[viewid].image);
fbState.attachments.back().viewFormat = MakeResourceFormat(c.m_ImageView[viewid].format);
fbState.attachments.back().firstMip = c.m_ImageView[viewid].range.baseMipLevel;
fbState.attachments.back().firstSlice = c.m_ImageView[viewid].range.baseArrayLayer;
fbState.attachments.back().numMips = c.m_ImageView[viewid].range.levelCount;
fbState.attachments.back().numSlices = c.m_ImageView[viewid].range.layerCount;
Convert(fbState.attachments.back().swizzle, c.m_ImageView[viewid].componentMapping);
rpState.depthstencilAttachment = int32_t(attIdx++);
}
else
{
rpState.depthstencilAttachment = -1;
}
if(dyn.fragmentDensityView != VK_NULL_HANDLE)
{
fbState.attachments.push_back({});
ResourceId viewid = GetResID(dyn.fragmentDensityView);
fbState.attachments.back().viewResourceId = rm->GetOriginalID(viewid);
ret.currentPass.framebuffer.attachments[attIdx].imageResourceId =
rm->GetOriginalID(c.m_ImageView[viewid].image);
fbState.attachments.back().viewFormat = MakeResourceFormat(c.m_ImageView[viewid].format);
fbState.attachments.back().firstMip = c.m_ImageView[viewid].range.baseMipLevel;
fbState.attachments.back().firstSlice = c.m_ImageView[viewid].range.baseArrayLayer;
fbState.attachments.back().numMips = c.m_ImageView[viewid].range.levelCount;
fbState.attachments.back().numSlices = c.m_ImageView[viewid].range.layerCount;
Convert(fbState.attachments.back().swizzle, c.m_ImageView[viewid].componentMapping);
rpState.fragmentDensityAttachment = int32_t(attIdx++);
}
else
{
rpState.fragmentDensityAttachment = -1;
}
if(dyn.shadingRateView != VK_NULL_HANDLE)
{
fbState.attachments.push_back({});
ResourceId viewid = GetResID(dyn.shadingRateView);
fbState.attachments.back().viewResourceId = rm->GetOriginalID(viewid);
ret.currentPass.framebuffer.attachments[attIdx].imageResourceId =
rm->GetOriginalID(c.m_ImageView[viewid].image);
fbState.attachments.back().viewFormat = MakeResourceFormat(c.m_ImageView[viewid].format);
fbState.attachments.back().firstMip = c.m_ImageView[viewid].range.baseMipLevel;
fbState.attachments.back().firstSlice = c.m_ImageView[viewid].range.baseArrayLayer;
fbState.attachments.back().numMips = c.m_ImageView[viewid].range.levelCount;
fbState.attachments.back().numSlices = c.m_ImageView[viewid].range.layerCount;
Convert(fbState.attachments.back().swizzle, c.m_ImageView[viewid].componentMapping);
rpState.shadingRateAttachment = int32_t(attIdx++);
rpState.shadingRateTexelSize = {dyn.shadingRateTexelSize.width,
dyn.shadingRateTexelSize.height};
}
else
{
rpState.shadingRateAttachment = -1;
rpState.shadingRateTexelSize = {1, 1};
}
rpState.multiviews.clear();
for(uint32_t v = 0; v < 32; v++)
{
if(dyn.viewMask & (1 << v))
rpState.multiviews.push_back(v);
}
}
else if(state.GetRenderPass() != ResourceId())
{
// Renderpass
ret.currentPass.renderpass.dynamic = false;
ret.currentPass.renderpass.resourceId = rm->GetOriginalID(state.GetRenderPass());
ret.currentPass.renderpass.subpass = state.subpass;
ret.currentPass.renderpass.inputAttachments =
c.m_RenderPass[state.GetRenderPass()].subpasses[state.subpass].inputAttachments;
ret.currentPass.renderpass.colorAttachments =
c.m_RenderPass[state.GetRenderPass()].subpasses[state.subpass].colorAttachments;
ret.currentPass.renderpass.resolveAttachments =
c.m_RenderPass[state.GetRenderPass()].subpasses[state.subpass].resolveAttachments;
ret.currentPass.renderpass.depthstencilAttachment =
c.m_RenderPass[state.GetRenderPass()].subpasses[state.subpass].depthstencilAttachment;
ret.currentPass.renderpass.depthstencilResolveAttachment =
c.m_RenderPass[state.GetRenderPass()].subpasses[state.subpass].depthstencilResolveAttachment;
ret.currentPass.renderpass.fragmentDensityAttachment =
c.m_RenderPass[state.GetRenderPass()].subpasses[state.subpass].fragmentDensityAttachment;
ret.currentPass.renderpass.shadingRateAttachment =
c.m_RenderPass[state.GetRenderPass()].subpasses[state.subpass].shadingRateAttachment;
VkExtent2D texelSize =
c.m_RenderPass[state.GetRenderPass()].subpasses[state.subpass].shadingRateTexelSize;
ret.currentPass.renderpass.shadingRateTexelSize = {texelSize.width, texelSize.height};
ret.currentPass.renderpass.multiviews =
c.m_RenderPass[state.GetRenderPass()].subpasses[state.subpass].multiviews;
ret.currentPass.renderpass.feedbackLoop =
c.m_RenderPass[state.GetRenderPass()].subpasses[state.subpass].feedbackLoop;
ret.currentPass.renderpass.tileOnlyMSAASampleCount =
c.m_RenderPass[state.GetRenderPass()].subpasses[state.subpass].tileOnlyMSAASampleCount;
ResourceId fb = state.GetFramebuffer();
ret.currentPass.framebuffer.resourceId = rm->GetOriginalID(fb);
if(fb != ResourceId())
{
ret.currentPass.framebuffer.width = c.m_Framebuffer[fb].width;
ret.currentPass.framebuffer.height = c.m_Framebuffer[fb].height;
ret.currentPass.framebuffer.layers = c.m_Framebuffer[fb].layers;
ret.currentPass.framebuffer.attachments.resize(c.m_Framebuffer[fb].attachments.size());
for(size_t i = 0; i < c.m_Framebuffer[fb].attachments.size(); i++)
{
ResourceId viewid = state.GetFramebufferAttachments()[i];
if(viewid != ResourceId())
{
ret.currentPass.framebuffer.attachments[i].viewResourceId = rm->GetOriginalID(viewid);
ret.currentPass.framebuffer.attachments[i].imageResourceId =
rm->GetOriginalID(c.m_ImageView[viewid].image);
ret.currentPass.framebuffer.attachments[i].viewFormat =
MakeResourceFormat(c.m_ImageView[viewid].format);
ret.currentPass.framebuffer.attachments[i].firstMip =
c.m_ImageView[viewid].range.baseMipLevel;
ret.currentPass.framebuffer.attachments[i].firstSlice =
c.m_ImageView[viewid].range.baseArrayLayer;
ret.currentPass.framebuffer.attachments[i].numMips = c.m_ImageView[viewid].range.levelCount;
ret.currentPass.framebuffer.attachments[i].numSlices =
c.m_ImageView[viewid].range.layerCount;
Convert(ret.currentPass.framebuffer.attachments[i].swizzle,
c.m_ImageView[viewid].componentMapping);
}
else
{
ret.currentPass.framebuffer.attachments[i].viewResourceId = ResourceId();
ret.currentPass.framebuffer.attachments[i].imageResourceId = ResourceId();
ret.currentPass.framebuffer.attachments[i].firstMip = 0;
ret.currentPass.framebuffer.attachments[i].firstSlice = 0;
ret.currentPass.framebuffer.attachments[i].numMips = 1;
ret.currentPass.framebuffer.attachments[i].numSlices = 1;
}
}
}
else
{
ret.currentPass.framebuffer.width = 0;
ret.currentPass.framebuffer.height = 0;
ret.currentPass.framebuffer.layers = 0;
}
ret.currentPass.renderpass.fragmentDensityOffsets.resize(state.fragmentDensityMapOffsets.size());
for(size_t i = 0; i < state.fragmentDensityMapOffsets.size(); i++)
{
const VkOffset2D &o = state.fragmentDensityMapOffsets[i];
ret.currentPass.renderpass.fragmentDensityOffsets[i] = Offset(o.x, o.y);
}
}
else
{
ret.currentPass.renderpass.resourceId = ResourceId();
ret.currentPass.renderpass.subpass = 0;
ret.currentPass.renderpass.inputAttachments.clear();
ret.currentPass.renderpass.colorAttachments.clear();
ret.currentPass.renderpass.resolveAttachments.clear();
ret.currentPass.renderpass.fragmentDensityOffsets.clear();
ret.currentPass.renderpass.depthstencilAttachment = -1;
ret.currentPass.renderpass.depthstencilResolveAttachment = -1;
ret.currentPass.renderpass.fragmentDensityAttachment = -1;
ret.currentPass.renderpass.shadingRateAttachment = -1;
ret.currentPass.renderpass.shadingRateTexelSize = {1, 1};
ret.currentPass.renderpass.tileOnlyMSAASampleCount = 0;
ret.currentPass.framebuffer.resourceId = ResourceId();
ret.currentPass.framebuffer.attachments.clear();
}
if(state.GetRenderPass() != ResourceId() || (state.dynamicRendering.active))
{
ret.currentPass.renderArea.x = state.renderArea.offset.x;
ret.currentPass.renderArea.y = state.renderArea.offset.y;
ret.currentPass.renderArea.width = state.renderArea.extent.width;
ret.currentPass.renderArea.height = state.renderArea.extent.height;
}
ret.currentPass.colorFeedbackAllowed = (state.feedbackAspects & VK_IMAGE_ASPECT_COLOR_BIT) != 0;
ret.currentPass.depthFeedbackAllowed = (state.feedbackAspects & VK_IMAGE_ASPECT_DEPTH_BIT) != 0;
ret.currentPass.stencilFeedbackAllowed = (state.feedbackAspects & VK_IMAGE_ASPECT_STENCIL_BIT) != 0;
// Descriptor sets
ret.graphics.descriptorSets.resize(state.graphics.descSets.size());
ret.compute.descriptorSets.resize(state.compute.descSets.size());
{
rdcarray<VKPipe::DescriptorSet> *dsts[] = {
&ret.graphics.descriptorSets,
&ret.compute.descriptorSets,
};
const rdcarray<VulkanStatePipeline::DescriptorAndOffsets> *srcs[] = {
&state.graphics.descSets,
&state.compute.descSets,
};
const VKDynamicShaderFeedback &usage = m_BindlessFeedback.Usage[eventId];
ret.shaderMessages = usage.messages;
for(size_t p = 0; p < ARRAY_COUNT(srcs); p++)
{
bool hasUsedBinds = false;
const BindpointIndex *usedBindsData = NULL;
size_t usedBindsSize = 0;
{
bool curCompute = (p == 1);
if(usage.valid && usage.compute == curCompute)
{
hasUsedBinds = true;
usedBindsData = usage.used.data();
usedBindsSize = usage.used.size();
}
const ActionDescription *action = m_pDriver->GetAction(eventId);
if(action)
{
bool isDispatch = bool(action->flags & ActionFlags::Dispatch);
// ifor compute stage on draws, and non-compute stages on dispatches, pretend all
// resources are dynamically unused, to prevent the lack of data from causing large arrays
// to be force-expanded
if((curCompute && !isDispatch) || (!curCompute && isDispatch))
{
hasUsedBinds = true;
usedBindsData = NULL;
usedBindsSize = 0;
}
}
}
BindpointIndex curBind;
for(size_t i = 0; i < srcs[p]->size(); i++)
{
ResourceId sourceSet = (*srcs[p])[i].descSet;
const uint32_t *srcOffset = (*srcs[p])[i].offsets.begin();
VKPipe::DescriptorSet &destSet = (*dsts[p])[i];
if(sourceSet == ResourceId())
{
destSet.inlineData.clear();
destSet.descriptorSetResourceId = ResourceId();
destSet.pushDescriptor = false;
destSet.layoutResourceId = ResourceId();
destSet.bindings.clear();
continue;
}
destSet.inlineData = m_pDriver->m_DescriptorSetState[sourceSet].data.inlineBytes;
curBind.bindset = (uint32_t)i;
ResourceId layoutId = m_pDriver->m_DescriptorSetState[sourceSet].layout;
// push descriptors don't have a real descriptor set backing them
if(c.m_DescSetLayout[layoutId].flags & VK_DESCRIPTOR_SET_LAYOUT_CREATE_PUSH_DESCRIPTOR_BIT_KHR)
{
destSet.descriptorSetResourceId = ResourceId();
destSet.pushDescriptor = true;
}
else
{
destSet.descriptorSetResourceId = rm->GetOriginalID(sourceSet);
destSet.pushDescriptor = false;
}
destSet.layoutResourceId = rm->GetOriginalID(layoutId);
destSet.bindings.resize(m_pDriver->m_DescriptorSetState[sourceSet].data.binds.size());
for(size_t b = 0; b < m_pDriver->m_DescriptorSetState[sourceSet].data.binds.size(); b++)
{
DescriptorSetSlot *sourceSlots = m_pDriver->m_DescriptorSetState[sourceSet].data.binds[b];
VKPipe::DescriptorBinding &destSlots = destSet.bindings[b];
const DescSetLayout::Binding &layoutBind = c.m_DescSetLayout[layoutId].bindings[b];
curBind.bind = (uint32_t)b;
uint32_t descriptorCount = layoutBind.descriptorCount;
if(layoutBind.variableSize)
descriptorCount = m_pDriver->m_DescriptorSetState[sourceSet].data.variableDescriptorCount;
destSlots.descriptorCount = descriptorCount;
if(layoutBind.layoutDescType == VK_DESCRIPTOR_TYPE_INLINE_UNIFORM_BLOCK)
destSlots.descriptorCount = 1;
destSlots.stageFlags = (ShaderStageMask)layoutBind.stageFlags;
destSlots.firstUsedIndex = -1;
destSlots.lastUsedIndex = -1;
destSlots.dynamicallyUsedCount = 0;
destSlots.binds.resize(destSlots.descriptorCount);
for(uint32_t a = 0; a < destSlots.descriptorCount; a++)
{
const DescriptorSetSlot &srcel = sourceSlots[a];
VKPipe::BindingElement &dstel = destSlots.binds[a];
// clear it so we don't have to manually reset all elements back to normal
memset(&dstel, 0, sizeof(dstel));
curBind.arrayIndex = a;
// if we have a list of used binds, and this is an array descriptor (so would be
// expected to be in the list), check it for dynamic usage.
if(destSlots.descriptorCount > 1 && hasUsedBinds)
{
// if we exhausted the list, all other elements are unused
if(usedBindsSize == 0)
{
dstel.dynamicallyUsed = false;
}
else
{
// we never saw the current value of usedBindsData (which is odd, we should have
// when iterating over all descriptors. This could only happen if there's some
// layout mismatch or a feedback bug that lead to an invalid entry in the list).
// Keep advancing until we get to one that is >= our current bind
while(curBind > *usedBindsData && usedBindsSize)
{
usedBindsData++;
usedBindsSize--;
}
// the next used bind is equal to this one. Mark it as dynamically used, and consume
if(usedBindsSize > 0 && curBind == *usedBindsData)
{
dstel.dynamicallyUsed = true;
usedBindsData++;
usedBindsSize--;
}
// the next used bind is after the current one, this is not used.
else if(usedBindsSize > 0 && curBind < *usedBindsData)
{
dstel.dynamicallyUsed = false;
}
}
}
else
{
dstel.dynamicallyUsed = true;
}
if(dstel.dynamicallyUsed)
{
destSlots.dynamicallyUsedCount++;
// we iterate in forward order, so we can unconditinoally set the last bind to the
// current one, and only set the first bind if we haven't encountered one before
destSlots.lastUsedIndex = a;
if(destSlots.firstUsedIndex < 0)
destSlots.firstUsedIndex = a;
}
DescriptorSlotType descriptorType = srcel.type;
// immutable samplers cannot be used with mutable descriptors, so if we have immutable
// samplers set the type from the layout. That way even if the descriptor is never
// written we still process immutable samplers properly.
if(layoutBind.immutableSampler)
descriptorType = convert(layoutBind.layoutDescType);
switch(descriptorType)
{
case DescriptorSlotType::Sampler: dstel.type = BindType::Sampler; break;
case DescriptorSlotType::CombinedImageSampler:
dstel.type = BindType::ImageSampler;
break;
case DescriptorSlotType::SampledImage: dstel.type = BindType::ReadOnlyImage; break;
case DescriptorSlotType::StorageImage: dstel.type = BindType::ReadWriteImage; break;
case DescriptorSlotType::UniformTexelBuffer:
dstel.type = BindType::ReadOnlyTBuffer;
break;
case DescriptorSlotType::StorageTexelBuffer:
dstel.type = BindType::ReadWriteTBuffer;
break;
case DescriptorSlotType::UniformBuffer: dstel.type = BindType::ConstantBuffer; break;
case DescriptorSlotType::StorageBuffer: dstel.type = BindType::ReadWriteBuffer; break;
case DescriptorSlotType::UniformBufferDynamic:
dstel.type = BindType::ConstantBuffer;
break;
case DescriptorSlotType::StorageBufferDynamic:
dstel.type = BindType::ReadWriteBuffer;
break;
case DescriptorSlotType::InputAttachment:
dstel.type = BindType::InputAttachment;
break;
case DescriptorSlotType::InlineBlock: dstel.type = BindType::ConstantBuffer; break;
case DescriptorSlotType::Unwritten:
case DescriptorSlotType::Count: dstel.type = BindType::Unknown; break;
}
// first handle the sampler separately because it might be in a combined descriptor
if(descriptorType == DescriptorSlotType::Sampler ||
descriptorType == DescriptorSlotType::CombinedImageSampler)
{
if(layoutBind.immutableSampler)
{
destSlots.binds[a].samplerResourceId = layoutBind.immutableSampler[a];
destSlots.binds[a].immutableSampler = true;
}
else if(srcel.sampler != ResourceId())
{
destSlots.binds[a].samplerResourceId = srcel.sampler;
}
if(destSlots.binds[a].samplerResourceId != ResourceId())
{
VKPipe::BindingElement &el = destSlots.binds[a];
const VulkanCreationInfo::Sampler &sampl = c.m_Sampler[el.samplerResourceId];
ResourceId liveId = el.samplerResourceId;
el.samplerResourceId = rm->GetOriginalID(el.samplerResourceId);
// sampler info
el.filter = MakeFilter(sampl.minFilter, sampl.magFilter, sampl.mipmapMode,
sampl.maxAnisotropy >= 1.0f, sampl.compareEnable,
sampl.reductionMode);
el.addressU = MakeAddressMode(sampl.address[0]);
el.addressV = MakeAddressMode(sampl.address[1]);
el.addressW = MakeAddressMode(sampl.address[2]);
el.mipBias = sampl.mipLodBias;
el.maxAnisotropy = sampl.maxAnisotropy;
el.compareFunction = MakeCompareFunc(sampl.compareOp);
el.minLOD = sampl.minLod;
el.maxLOD = sampl.maxLod;
MakeBorderColor(sampl.borderColor, el.borderColorValue.floatValue);
el.borderColorType = CompType::Float;
el.unnormalized = sampl.unnormalizedCoordinates;
el.seamless = sampl.seamless;
if(sampl.ycbcr != ResourceId())
{
const VulkanCreationInfo::YCbCrSampler &ycbcr = c.m_YCbCrSampler[sampl.ycbcr];
el.ycbcrSampler = rm->GetOriginalID(sampl.ycbcr);
el.ycbcrModel = ycbcr.ycbcrModel;
el.ycbcrRange = ycbcr.ycbcrRange;
Convert(el.samplerSwizzle, ycbcr.componentMapping);
el.xChromaOffset = ycbcr.xChromaOffset;
el.yChromaOffset = ycbcr.yChromaOffset;
el.chromaFilter = ycbcr.chromaFilter;
el.forceExplicitReconstruction = ycbcr.forceExplicitReconstruction;
}
else
{
Convert(el.samplerSwizzle, sampl.componentMapping);
el.srgbBorder = sampl.srgbBorder;
}
if(sampl.customBorder)
{
if(sampl.borderColor == VK_BORDER_COLOR_INT_CUSTOM_EXT)
{
el.borderColorValue.uintValue = sampl.customBorderColor.uint32;
el.borderColorType = CompType::UInt;
}
else
{
el.borderColorValue.floatValue = sampl.customBorderColor.float32;
}
}
}
}
// now look at the 'base' type. Sampler is excluded from these ifs
if(descriptorType == DescriptorSlotType::SampledImage ||
descriptorType == DescriptorSlotType::CombinedImageSampler ||
descriptorType == DescriptorSlotType::InputAttachment ||
descriptorType == DescriptorSlotType::StorageImage)
{
ResourceId viewid = srcel.resource;
if(viewid != ResourceId())
{
destSlots.binds[a].viewResourceId = rm->GetOriginalID(viewid);
destSlots.binds[a].resourceResourceId =
rm->GetOriginalID(c.m_ImageView[viewid].image);
destSlots.binds[a].viewFormat = MakeResourceFormat(c.m_ImageView[viewid].format);
Convert(destSlots.binds[a].swizzle, c.m_ImageView[viewid].componentMapping);
destSlots.binds[a].firstMip = c.m_ImageView[viewid].range.baseMipLevel;
destSlots.binds[a].firstSlice = c.m_ImageView[viewid].range.baseArrayLayer;
destSlots.binds[a].numMips = c.m_ImageView[viewid].range.levelCount;
destSlots.binds[a].numSlices = c.m_ImageView[viewid].range.layerCount;
if(c.m_ImageView[viewid].viewType == VK_IMAGE_VIEW_TYPE_3D)
destSlots.binds[a].firstSlice = destSlots.binds[a].numSlices = 0;
// temporary hack, store image layout enum in byteOffset as it's not used for images
destSlots.binds[a].byteOffset = convert(srcel.imageLayout);
destSlots.binds[a].minLOD = c.m_ImageView[viewid].minLOD;
}
else
{
destSlots.binds[a].viewResourceId = ResourceId();
destSlots.binds[a].resourceResourceId = ResourceId();
destSlots.binds[a].firstMip = 0;
destSlots.binds[a].firstSlice = 0;
destSlots.binds[a].numMips = 1;
destSlots.binds[a].numSlices = 1;
destSlots.binds[a].minLOD = 0.0f;
}
}
else if(descriptorType == DescriptorSlotType::UniformTexelBuffer ||
descriptorType == DescriptorSlotType::StorageTexelBuffer)
{
ResourceId viewid = srcel.resource;
if(viewid != ResourceId())
{
destSlots.binds[a].viewResourceId = rm->GetOriginalID(viewid);
destSlots.binds[a].resourceResourceId =
rm->GetOriginalID(c.m_BufferView[viewid].buffer);
destSlots.binds[a].byteOffset = c.m_BufferView[viewid].offset;
destSlots.binds[a].viewFormat = MakeResourceFormat(c.m_BufferView[viewid].format);
destSlots.binds[a].byteSize = c.m_BufferView[viewid].size;
}
else
{
destSlots.binds[a].viewResourceId = ResourceId();
destSlots.binds[a].resourceResourceId = ResourceId();
destSlots.binds[a].byteOffset = 0;
destSlots.binds[a].byteSize = 0;
}
}
else if(descriptorType == DescriptorSlotType::InlineBlock)
{
destSlots.binds[a].viewResourceId = ResourceId();
destSlots.binds[a].resourceResourceId = ResourceId();
destSlots.binds[a].inlineBlock = true;
destSlots.binds[a].byteOffset = srcel.offset;
destSlots.binds[a].byteSize = descriptorCount;
}
else if(descriptorType == DescriptorSlotType::StorageBuffer ||
descriptorType == DescriptorSlotType::StorageBufferDynamic ||
descriptorType == DescriptorSlotType::UniformBuffer ||
descriptorType == DescriptorSlotType::UniformBufferDynamic)
{
destSlots.binds[a].viewResourceId = ResourceId();
if(srcel.resource != ResourceId())
destSlots.binds[a].resourceResourceId = rm->GetOriginalID(srcel.resource);
destSlots.binds[a].byteOffset = srcel.offset;
if(descriptorType == DescriptorSlotType::StorageBufferDynamic ||
descriptorType == DescriptorSlotType::UniformBufferDynamic)
{
destSlots.binds[a].byteOffset += *srcOffset;
srcOffset++;
}
destSlots.binds[a].byteSize = srcel.GetRange();
}
}
// if no bindings were set these will still be negative. Set them to something sensible.
if(destSlots.firstUsedIndex < 0)
{
destSlots.firstUsedIndex = 0;
destSlots.lastUsedIndex = 0x7fffffff;
}
}
}
}
}
// image layouts
{
size_t i = 0;
ret.images.resize(m_pDriver->m_ImageStates.size());
for(auto it = m_pDriver->m_ImageStates.begin(); it != m_pDriver->m_ImageStates.end(); ++it)
{
VKPipe::ImageData &img = ret.images[i];
if(rm->GetOriginalID(it->first) == it->first)
continue;
img.resourceId = rm->GetOriginalID(it->first);
LockedConstImageStateRef imState = it->second.LockRead();
img.layouts.resize(imState->subresourceStates.size());
auto subIt = imState->subresourceStates.begin();
for(size_t l = 0; l < img.layouts.size(); ++l, ++subIt)
{
img.layouts[l].name = ToStr(subIt->state().newLayout);
img.layouts[l].baseMip = subIt->range().baseMipLevel;
img.layouts[l].numMip = subIt->range().levelCount;
img.layouts[l].baseLayer = subIt->range().baseArrayLayer;
img.layouts[l].numLayer = subIt->range().layerCount;
}
if(img.layouts.empty())
{
img.layouts.push_back(VKPipe::ImageLayout());
img.layouts[0].name = "Unknown";
}
i++;
}
ret.images.resize(i);
}
if(state.conditionalRendering.buffer != ResourceId())
{
ret.conditionalRendering.bufferId = rm->GetOriginalID(state.conditionalRendering.buffer);
ret.conditionalRendering.byteOffset = state.conditionalRendering.offset;
ret.conditionalRendering.isInverted =
state.conditionalRendering.flags == VK_CONDITIONAL_RENDERING_INVERTED_BIT_EXT;
bytebuf data;
GetBufferData(state.conditionalRendering.buffer, state.conditionalRendering.offset,
sizeof(uint32_t), data);
uint32_t value;
memcpy(&value, data.data(), sizeof(uint32_t));
ret.conditionalRendering.isPassing = value != 0;
if(ret.conditionalRendering.isInverted)
ret.conditionalRendering.isPassing = !ret.conditionalRendering.isPassing;
}
}
void VulkanReplay::FillCBufferVariables(ResourceId pipeline, ResourceId shader, ShaderStage stage,
rdcstr entryPoint, uint32_t cbufSlot,
rdcarray<ShaderVariable> &outvars, const bytebuf &data)
{
auto it = m_pDriver->m_CreationInfo.m_ShaderModule.find(shader);
if(it == m_pDriver->m_CreationInfo.m_ShaderModule.end())
{
RDCERR("Can't get shader details");
return;
}
ShaderReflection &refl = *it->second.GetReflection(stage, entryPoint, pipeline).refl;
ShaderBindpointMapping &mapping = it->second.GetReflection(stage, entryPoint, pipeline).mapping;
if(cbufSlot >= (uint32_t)refl.constantBlocks.count())
{
RDCERR("Invalid cbuffer slot");
return;
}
ConstantBlock &c = refl.constantBlocks[cbufSlot];
if(c.bufferBacked)
{
const rdcarray<VulkanStatePipeline::DescriptorAndOffsets> &descSets =
(refl.stage == ShaderStage::Compute) ? m_pDriver->m_RenderState.compute.descSets
: m_pDriver->m_RenderState.graphics.descSets;
Bindpoint bind = mapping.constantBlocks[c.bindPoint];
if(bind.bindset < descSets.count())
{
ResourceId set = descSets[bind.bindset].descSet;
const WrappedVulkan::DescriptorSetInfo &setData = m_pDriver->m_DescriptorSetState[set];
ResourceId layoutId = setData.layout;
if(bind.bind < m_pDriver->m_CreationInfo.m_DescSetLayout[layoutId].bindings.count())
{
const DescSetLayout::Binding &layoutBind =
m_pDriver->m_CreationInfo.m_DescSetLayout[layoutId].bindings[bind.bind];
if(layoutBind.layoutDescType == VK_DESCRIPTOR_TYPE_INLINE_UNIFORM_BLOCK)
{
bytebuf inlineData;
inlineData.assign(setData.data.inlineBytes.data() + setData.data.binds[bind.bind]->offset,
layoutBind.variableSize ? setData.data.variableDescriptorCount
: layoutBind.descriptorCount);
StandardFillCBufferVariables(refl.resourceId, c.variables, outvars, inlineData);
return;
}
}
}
StandardFillCBufferVariables(refl.resourceId, c.variables, outvars, data);
}
else
{
// specialised path to display specialization constants
if(mapping.constantBlocks[c.bindPoint].bindset == SpecializationConstantBindSet)
{
auto pipeIt = m_pDriver->m_CreationInfo.m_Pipeline.find(pipeline);
if(pipeIt != m_pDriver->m_CreationInfo.m_Pipeline.end())
{
const VulkanCreationInfo::ShaderModuleReflection &reflection =
it->second.GetReflection(stage, entryPoint, pipeline);
const rdcarray<SpecConstant> &specInfo =
pipeIt->second.shaders[reflection.stageIndex].specialization;
FillSpecConstantVariables(refl.resourceId, reflection.patchData, c.variables, outvars,
specInfo);
}
}
else
{
bytebuf pushdata;
pushdata.resize(sizeof(m_pDriver->m_RenderState.pushconsts));
memcpy(&pushdata[0], m_pDriver->m_RenderState.pushconsts, pushdata.size());
StandardFillCBufferVariables(refl.resourceId, c.variables, outvars, pushdata);
}
}
}
void VulkanReplay::PickPixel(ResourceId texture, uint32_t x, uint32_t y, const Subresource &sub,
CompType typeCast, float pixel[4])
{
int oldW = m_DebugWidth, oldH = m_DebugHeight;
m_DebugWidth = m_DebugHeight = 1;
VulkanCreationInfo::Image &iminfo = m_pDriver->m_CreationInfo.m_Image[texture];
LockedConstImageStateRef imageState = m_pDriver->FindConstImageState(texture);
if(!imageState)
{
RDCWARN("Could not find image info for image %s", ToStr(texture).c_str());
return;
}
if(!imageState->isMemoryBound)
return;
bool isStencil = IsStencilFormat(iminfo.format);
// do a second pass to render the stencil, if needed
for(int pass = 0; pass < (isStencil ? 2 : 1); pass++)
{
// render picked pixel to readback F32 RGBA texture
{
TextureDisplay texDisplay;
texDisplay.red = texDisplay.green = texDisplay.blue = texDisplay.alpha = true;
texDisplay.hdrMultiplier = -1.0f;
texDisplay.linearDisplayAsGamma = true;
texDisplay.flipY = false;
texDisplay.subresource = sub;
texDisplay.customShaderId = ResourceId();
texDisplay.overlay = DebugOverlay::NoOverlay;
texDisplay.rangeMin = 0.0f;
texDisplay.rangeMax = 1.0f;
texDisplay.scale = 1.0f;
texDisplay.resourceId = texture;
texDisplay.typeCast = typeCast;
texDisplay.rawOutput = true;
uint32_t mipWidth = RDCMAX(1U, iminfo.extent.width >> sub.mip);
uint32_t mipHeight = RDCMAX(1U, iminfo.extent.height >> sub.mip);
texDisplay.xOffset = -(float(x) / float(mipWidth)) * iminfo.extent.width;
texDisplay.yOffset = -(float(y) / float(mipHeight)) * iminfo.extent.height;
// only render green (stencil) in second pass
if(pass == 1)
{
texDisplay.green = true;
texDisplay.red = texDisplay.blue = texDisplay.alpha = false;
}
VkClearValue clearval = {};
VkRenderPassBeginInfo rpbegin = {
VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
NULL,
Unwrap(m_PixelPick.RP),
Unwrap(m_PixelPick.FB),
{{
0,
0,
},
{1, 1}},
1,
&clearval,
};
RenderTextureInternal(texDisplay, *imageState, rpbegin,
eTexDisplay_32Render | eTexDisplay_MipShift);
}
VkDevice dev = m_pDriver->GetDev();
VkCommandBuffer cmd = m_pDriver->GetNextCmd();
const VkDevDispatchTable *vt = ObjDisp(dev);
if(cmd == VK_NULL_HANDLE)
return;
VkResult vkr = VK_SUCCESS;
{
VkImageMemoryBarrier pickimBarrier = {VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
NULL,
0,
0,
VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL,
VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,
VK_QUEUE_FAMILY_IGNORED,
VK_QUEUE_FAMILY_IGNORED,
Unwrap(m_PixelPick.Image),
{VK_IMAGE_ASPECT_COLOR_BIT, 0, 1, 0, 1}};
// update image layout from color attachment to transfer source, with proper memory barriers
pickimBarrier.srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
pickimBarrier.dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT;
VkCommandBufferBeginInfo beginInfo = {VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO, NULL,
VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT};
vkr = vt->BeginCommandBuffer(Unwrap(cmd), &beginInfo);
CheckVkResult(vkr);
DoPipelineBarrier(cmd, 1, &pickimBarrier);
pickimBarrier.oldLayout = pickimBarrier.newLayout;
pickimBarrier.srcAccessMask = pickimBarrier.dstAccessMask;
// do copy
VkBufferImageCopy region = {
0, 128, 1, {VK_IMAGE_ASPECT_COLOR_BIT, 0, 0, 1}, {0, 0, 0}, {1, 1, 1},
};
vt->CmdCopyImageToBuffer(Unwrap(cmd), Unwrap(m_PixelPick.Image),
VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,
Unwrap(m_PixelPick.ReadbackBuffer.buf), 1, &region);
// update image layout back to color attachment
pickimBarrier.newLayout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL;
pickimBarrier.dstAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
DoPipelineBarrier(cmd, 1, &pickimBarrier);
vt->EndCommandBuffer(Unwrap(cmd));
}
// submit cmds and wait for idle so we can readback
m_pDriver->SubmitCmds();
m_pDriver->FlushQ();
float *pData = NULL;
vkr = vt->MapMemory(Unwrap(dev), Unwrap(m_PixelPick.ReadbackBuffer.mem), 0, VK_WHOLE_SIZE, 0,
(void **)&pData);
CheckVkResult(vkr);
if(vkr != VK_SUCCESS)
return;
if(!pData)
{
RDCERR("Manually reporting failed memory map");
CheckVkResult(VK_ERROR_MEMORY_MAP_FAILED);
return;
}
VkMappedMemoryRange range = {
VK_STRUCTURE_TYPE_MAPPED_MEMORY_RANGE,
NULL,
Unwrap(m_PixelPick.ReadbackBuffer.mem),
0,
VK_WHOLE_SIZE,
};
vkr = vt->InvalidateMappedMemoryRanges(Unwrap(dev), 1, &range);
CheckVkResult(vkr);
RDCASSERT(pData != NULL);
if(pData == NULL)
{
RDCERR("Failed ot map readback buffer memory");
}
else
{
// only write stencil to .y
if(pass == 1)
{
pixel[1] = ((uint32_t *)pData)[0] / 255.0f;
}
else
{
pixel[0] = pData[0];
pixel[1] = pData[1];
pixel[2] = pData[2];
pixel[3] = pData[3];
}
}
vt->UnmapMemory(Unwrap(dev), Unwrap(m_PixelPick.ReadbackBuffer.mem));
}
m_DebugWidth = oldW;
m_DebugHeight = oldH;
}
bool VulkanReplay::GetMinMax(ResourceId texid, const Subresource &sub, CompType typeCast,
float *minval, float *maxval)
{
const ImageInfo *imageInfo = NULL;
{
LockedConstImageStateRef state = m_pDriver->FindConstImageState(texid);
if(!state)
return false;
imageInfo = &state->GetImageInfo();
}
if(IsDepthAndStencilFormat(imageInfo->format))
{
// for depth/stencil we need to run the code twice - once to fetch depth and once to fetch
// stencil - since we can't process float depth and int stencil at the same time
Vec4f depth[2] = {
{0.0f, 0.0f, 0.0f, 0.0f},
{1.0f, 1.0f, 1.0f, 1.0f},
};
Vec4u stencil[2] = {{0, 0, 0, 0}, {1, 1, 1, 1}};
bool success = GetMinMax(texid, sub, typeCast, false, &depth[0].x, &depth[1].x);
if(!success)
return false;
success = GetMinMax(texid, sub, typeCast, true, (float *)&stencil[0].x, (float *)&stencil[1].x);
if(!success)
return false;
// copy across into green channel, casting up to float, dividing by the range for this texture
depth[0].y = float(stencil[0].x) / 255.0f;
depth[1].y = float(stencil[1].x) / 255.0f;
memcpy(minval, &depth[0], sizeof(depth[0]));
memcpy(maxval, &depth[1], sizeof(depth[1]));
return true;
}
return GetMinMax(texid, sub, typeCast, false, minval, maxval);
}
bool VulkanReplay::GetMinMax(ResourceId texid, const Subresource &sub, CompType typeCast,
bool stencil, float *minval, float *maxval)
{
VkDevice dev = m_pDriver->GetDev();
const VkDevDispatchTable *vt = ObjDisp(dev);
LockedConstImageStateRef state = m_pDriver->FindConstImageState(texid);
if(!state)
return false;
bool isMemoryBound = state->isMemoryBound;
VulkanCreationInfo::Image &iminfo = m_pDriver->m_CreationInfo.m_Image[texid];
TextureDisplayViews &texviews = m_TexRender.TextureViews[texid];
VkImage liveIm = m_pDriver->GetResourceManager()->GetCurrentHandle<VkImage>(texid);
if(!isMemoryBound)
return false;
if(!IsStencilFormat(iminfo.format))
stencil = false;
CreateTexImageView(liveIm, iminfo, typeCast, texviews);
VkImageView liveImView = texviews.views[0];
// if it's not stencil-only and we're displaying stencil, use view 1
if(texviews.castedFormat != VK_FORMAT_S8_UINT && stencil)
liveImView = texviews.views[1];
RDCASSERT(liveImView != VK_NULL_HANDLE);
VkDescriptorImageInfo imdesc = {0};
imdesc.imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
imdesc.imageView = Unwrap(liveImView);
imdesc.sampler = Unwrap(m_General.PointSampler);
uint32_t descSetBinding = 0;
uint32_t intTypeIndex = 0;
if(IsUIntFormat(texviews.castedFormat))
{
descSetBinding = 10;
intTypeIndex = 1;
}
else if(IsSIntFormat(texviews.castedFormat))
{
descSetBinding = 15;
intTypeIndex = 2;
}
else
{
descSetBinding = 5;
}
int textype = 0;
if(iminfo.type == VK_IMAGE_TYPE_1D)
{
textype = RESTYPE_TEX1D;
}
else if(iminfo.type == VK_IMAGE_TYPE_3D)
{
textype = RESTYPE_TEX3D;
}
else if(iminfo.type == VK_IMAGE_TYPE_2D)
{
textype = RESTYPE_TEX2D;
if(iminfo.samples != VK_SAMPLE_COUNT_1_BIT)
textype = RESTYPE_TEX2DMS;
}
if(stencil)
{
descSetBinding = 10;
intTypeIndex = 1;
}
descSetBinding += textype;
if(m_Histogram.m_MinMaxTilePipe[textype][intTypeIndex] == VK_NULL_HANDLE)
return false;
VkDescriptorBufferInfo bufdescs[3];
RDCEraseEl(bufdescs);
m_Histogram.m_MinMaxTileResult.FillDescriptor(bufdescs[0]);
m_Histogram.m_MinMaxResult.FillDescriptor(bufdescs[1]);
m_Histogram.m_HistogramUBO.FillDescriptor(bufdescs[2]);
VkDescriptorImageInfo altimdesc[2] = {};
for(uint32_t i = 1; i < GetYUVPlaneCount(texviews.castedFormat); i++)
{
RDCASSERT(texviews.views[i] != VK_NULL_HANDLE);
altimdesc[i - 1].imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
altimdesc[i - 1].imageView = Unwrap(texviews.views[i]);
altimdesc[i - 1].sampler = Unwrap(m_General.PointSampler);
}
VkWriteDescriptorSet writeSet[] = {
// first pass on tiles
{
VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET, NULL, Unwrap(m_Histogram.m_HistogramDescSet[0]),
0, 0, 1, VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, NULL, &bufdescs[0],
NULL // destination = tile result
},
{
VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET, NULL, Unwrap(m_Histogram.m_HistogramDescSet[0]),
1, 0, 1, VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, NULL, &bufdescs[0],
NULL // source = unused, bind tile result
},
{VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET, NULL, Unwrap(m_Histogram.m_HistogramDescSet[0]), 2,
0, 1, VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, NULL, &bufdescs[2], NULL},
// sampled view
{VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET, NULL, Unwrap(m_Histogram.m_HistogramDescSet[0]),
descSetBinding, 0, 1, VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, &imdesc, NULL, NULL},
// YUV secondary planes (if needed)
{VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET, NULL, Unwrap(m_Histogram.m_HistogramDescSet[0]), 10,
0, GetYUVPlaneCount(texviews.castedFormat) - 1, VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER,
altimdesc, NULL, NULL},
// second pass from tiles to result
{
VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET, NULL, Unwrap(m_Histogram.m_HistogramDescSet[1]),
0, 0, 1, VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, NULL, &bufdescs[1],
NULL // destination = result
},
{
VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET, NULL, Unwrap(m_Histogram.m_HistogramDescSet[1]),
1, 0, 1, VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, NULL, &bufdescs[0],
NULL // source = tile result
},
{VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET, NULL, Unwrap(m_Histogram.m_HistogramDescSet[1]), 2,
0, 1, VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, NULL, &bufdescs[2], NULL},
};
rdcarray<VkWriteDescriptorSet> writeSets;
for(size_t i = 0; i < ARRAY_COUNT(writeSet); i++)
{
if(writeSet[i].descriptorCount > 0)
writeSets.push_back(writeSet[i]);
}
for(size_t i = 0; i < ARRAY_COUNT(m_TexRender.DummyWrites); i++)
{
VkWriteDescriptorSet &write = m_TexRender.DummyWrites[i];
// don't write dummy data in the actual slot
if(write.dstBinding == descSetBinding)
continue;
// don't overwrite YUV texture slots if it's a YUV planar format
if(write.dstBinding == 10)
{
if(write.dstArrayElement == 0 && GetYUVPlaneCount(texviews.castedFormat) >= 2)
continue;
if(write.dstArrayElement == 1 && GetYUVPlaneCount(texviews.castedFormat) >= 3)
continue;
}
write.dstSet = Unwrap(m_Histogram.m_HistogramDescSet[0]);
writeSets.push_back(write);
}
vt->UpdateDescriptorSets(Unwrap(dev), (uint32_t)writeSets.size(), &writeSets[0], 0, NULL);
HistogramUBOData *data = (HistogramUBOData *)m_Histogram.m_HistogramUBO.Map(NULL);
if(!data)
return false;
data->HistogramTextureResolution.x = (float)RDCMAX(uint32_t(iminfo.extent.width) >> sub.mip, 1U);
data->HistogramTextureResolution.y = (float)RDCMAX(uint32_t(iminfo.extent.height) >> sub.mip, 1U);
data->HistogramTextureResolution.z = (float)RDCMAX(uint32_t(iminfo.extent.depth) >> sub.mip, 1U);
if(iminfo.type == VK_IMAGE_TYPE_3D)
data->HistogramSlice =
(float)RDCCLAMP(sub.slice, 0U, uint32_t(iminfo.extent.depth >> sub.mip) - 1) + 0.001f;
else
data->HistogramSlice = (float)RDCCLAMP(sub.slice, 0U, iminfo.arrayLayers - 1) + 0.001f;
data->HistogramMip = (int)sub.mip;
data->HistogramNumSamples = iminfo.samples;
data->HistogramSample = (int)RDCCLAMP(sub.sample, 0U, uint32_t(iminfo.samples) - 1);
if(sub.sample == ~0U)
data->HistogramSample = -iminfo.samples;
data->HistogramMin = 0.0f;
data->HistogramMax = 1.0f;
data->HistogramChannels = 0xf;
Vec4u YUVDownsampleRate = {};
Vec4u YUVAChannels = {};
GetYUVShaderParameters(texviews.castedFormat, YUVDownsampleRate, YUVAChannels);
data->HistogramYUVDownsampleRate = YUVDownsampleRate;
data->HistogramYUVAChannels = YUVAChannels;
m_Histogram.m_HistogramUBO.Unmap();
VkCommandBufferBeginInfo beginInfo = {VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO, NULL,
VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT};
VkCommandBuffer cmd = m_pDriver->GetNextCmd();
if(cmd == VK_NULL_HANDLE)
return false;
vt->BeginCommandBuffer(Unwrap(cmd), &beginInfo);
ImageBarrierSequence setupBarriers, cleanupBarriers;
state->TempTransition(m_pDriver->m_QueueFamilyIdx, VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL,
VK_ACCESS_SHADER_READ_BIT, setupBarriers, cleanupBarriers,
m_pDriver->GetImageTransitionInfo());
m_pDriver->InlineSetupImageBarriers(cmd, setupBarriers);
m_pDriver->SubmitAndFlushImageStateBarriers(setupBarriers);
int blocksX = (int)ceil(iminfo.extent.width / float(HGRAM_PIXELS_PER_TILE * HGRAM_TILES_PER_BLOCK));
int blocksY =
(int)ceil(iminfo.extent.height / float(HGRAM_PIXELS_PER_TILE * HGRAM_TILES_PER_BLOCK));
vt->CmdBindPipeline(Unwrap(cmd), VK_PIPELINE_BIND_POINT_COMPUTE,
Unwrap(m_Histogram.m_MinMaxTilePipe[textype][intTypeIndex]));
vt->CmdBindDescriptorSets(Unwrap(cmd), VK_PIPELINE_BIND_POINT_COMPUTE,
Unwrap(m_Histogram.m_HistogramPipeLayout), 0, 1,
UnwrapPtr(m_Histogram.m_HistogramDescSet[0]), 0, NULL);
vt->CmdDispatch(Unwrap(cmd), blocksX, blocksY, 1);
m_pDriver->InlineCleanupImageBarriers(cmd, cleanupBarriers);
if(!cleanupBarriers.empty())
{
vt->EndCommandBuffer(Unwrap(cmd));
m_pDriver->SubmitCmds();
m_pDriver->FlushQ();
m_pDriver->SubmitAndFlushImageStateBarriers(cleanupBarriers);
cmd = m_pDriver->GetNextCmd();
if(cmd == VK_NULL_HANDLE)
return false;
vt->BeginCommandBuffer(Unwrap(cmd), &beginInfo);
}
VkBufferMemoryBarrier tilebarrier = {
VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,
NULL,
VK_ACCESS_SHADER_WRITE_BIT,
VK_ACCESS_SHADER_READ_BIT,
VK_QUEUE_FAMILY_IGNORED,
VK_QUEUE_FAMILY_IGNORED,
Unwrap(m_Histogram.m_MinMaxTileResult.buf),
0,
m_Histogram.m_MinMaxTileResult.totalsize,
};
// ensure shader writes complete before coalescing the tiles
DoPipelineBarrier(cmd, 1, &tilebarrier);
vt->CmdBindPipeline(Unwrap(cmd), VK_PIPELINE_BIND_POINT_COMPUTE,
Unwrap(m_Histogram.m_MinMaxResultPipe[intTypeIndex]));
vt->CmdBindDescriptorSets(Unwrap(cmd), VK_PIPELINE_BIND_POINT_COMPUTE,
Unwrap(m_Histogram.m_HistogramPipeLayout), 0, 1,
UnwrapPtr(m_Histogram.m_HistogramDescSet[1]), 0, NULL);
vt->CmdDispatch(Unwrap(cmd), 1, 1, 1);
// ensure shader writes complete before copying back to readback buffer
tilebarrier.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT;
tilebarrier.dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT;
tilebarrier.buffer = Unwrap(m_Histogram.m_MinMaxResult.buf);
tilebarrier.size = m_Histogram.m_MinMaxResult.totalsize;
DoPipelineBarrier(cmd, 1, &tilebarrier);
VkBufferCopy bufcopy = {
0,
0,
m_Histogram.m_MinMaxResult.totalsize,
};
vt->CmdCopyBuffer(Unwrap(cmd), Unwrap(m_Histogram.m_MinMaxResult.buf),
Unwrap(m_Histogram.m_MinMaxReadback.buf), 1, &bufcopy);
// wait for copy to complete before mapping
tilebarrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
tilebarrier.dstAccessMask = VK_ACCESS_HOST_READ_BIT;
tilebarrier.buffer = Unwrap(m_Histogram.m_MinMaxReadback.buf);
tilebarrier.size = m_Histogram.m_MinMaxResult.totalsize;
DoPipelineBarrier(cmd, 1, &tilebarrier);
vt->EndCommandBuffer(Unwrap(cmd));
// submit cmds and wait for idle so we can readback
m_pDriver->SubmitCmds();
m_pDriver->FlushQ();
Vec4f *minmax = (Vec4f *)m_Histogram.m_MinMaxReadback.Map(NULL);
if(!minmax)
return false;
minval[0] = minmax[0].x;
minval[1] = minmax[0].y;
minval[2] = minmax[0].z;
minval[3] = minmax[0].w;
maxval[0] = minmax[1].x;
maxval[1] = minmax[1].y;
maxval[2] = minmax[1].z;
maxval[3] = minmax[1].w;
m_Histogram.m_MinMaxReadback.Unmap();
return true;
}
void VulkanReplay::CheckVkResult(VkResult vkr)
{
return m_pDriver->CheckVkResult(vkr);
}
bool VulkanReplay::GetHistogram(ResourceId texid, const Subresource &sub, CompType typeCast,
float minval, float maxval, const rdcfixedarray<bool, 4> &channels,
rdcarray<uint32_t> &histogram)
{
if(minval >= maxval)
return false;
VkDevice dev = m_pDriver->GetDev();
const VkDevDispatchTable *vt = ObjDisp(dev);
LockedConstImageStateRef state = m_pDriver->FindConstImageState(texid);
if(!state->isMemoryBound)
return false;
VulkanCreationInfo::Image &iminfo = m_pDriver->m_CreationInfo.m_Image[texid];
TextureDisplayViews &texviews = m_TexRender.TextureViews[texid];
VkImage liveIm = m_pDriver->GetResourceManager()->GetCurrentHandle<VkImage>(texid);
bool stencil = false;
// detect if stencil is selected
if(IsStencilFormat(iminfo.format) && !channels[0] && channels[1] && !channels[2] && !channels[3])
stencil = true;
CreateTexImageView(liveIm, iminfo, typeCast, texviews);
uint32_t descSetBinding = 0;
uint32_t intTypeIndex = 0;
if(IsUIntFormat(texviews.castedFormat))
{
descSetBinding = 10;
intTypeIndex = 1;
}
else if(IsSIntFormat(texviews.castedFormat))
{
descSetBinding = 15;
intTypeIndex = 2;
}
else
{
descSetBinding = 5;
}
int textype = 0;
if(iminfo.type == VK_IMAGE_TYPE_1D)
{
textype = RESTYPE_TEX1D;
}
else if(iminfo.type == VK_IMAGE_TYPE_3D)
{
textype = RESTYPE_TEX3D;
}
else if(iminfo.type == VK_IMAGE_TYPE_2D)
{
textype = RESTYPE_TEX2D;
if(iminfo.samples != VK_SAMPLE_COUNT_1_BIT)
textype = RESTYPE_TEX2DMS;
}
if(stencil)
{
descSetBinding = 10;
intTypeIndex = 1;
// rescale the range so that stencil seems to fit to 0-1
minval *= 255.0f;
maxval *= 255.0f;
}
descSetBinding += textype;
if(m_Histogram.m_HistogramPipe[textype][intTypeIndex] == VK_NULL_HANDLE)
{
histogram.resize(HGRAM_NUM_BUCKETS);
for(size_t i = 0; i < HGRAM_NUM_BUCKETS; i++)
histogram[i] = 1;
return false;
}
VkImageView liveImView = texviews.views[0];
// if it's not stencil-only and we're displaying stencil, use view 1
if(stencil && texviews.castedFormat != VK_FORMAT_S8_UINT)
liveImView = texviews.views[1];
RDCASSERT(liveImView != VK_NULL_HANDLE);
VkDescriptorImageInfo imdesc = {0};
imdesc.imageLayout = VK_IMAGE_LAYOUT_GENERAL;
imdesc.imageView = Unwrap(liveImView);
imdesc.sampler = Unwrap(m_General.PointSampler);
VkDescriptorBufferInfo bufdescs[2];
RDCEraseEl(bufdescs);
m_Histogram.m_HistogramBuf.FillDescriptor(bufdescs[0]);
m_Histogram.m_HistogramUBO.FillDescriptor(bufdescs[1]);
VkDescriptorImageInfo altimdesc[2] = {};
for(uint32_t i = 1; i < GetYUVPlaneCount(texviews.castedFormat); i++)
{
RDCASSERT(texviews.views[i] != VK_NULL_HANDLE);
altimdesc[i - 1].imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
altimdesc[i - 1].imageView = Unwrap(texviews.views[i]);
altimdesc[i - 1].sampler = Unwrap(m_General.PointSampler);
}
VkWriteDescriptorSet writeSet[] = {
// histogram pass
{
VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET, NULL, Unwrap(m_Histogram.m_HistogramDescSet[0]),
0, 0, 1, VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, NULL, &bufdescs[0],
NULL // destination = histogram result
},
{
VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET, NULL, Unwrap(m_Histogram.m_HistogramDescSet[0]),
1, 0, 1, VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, NULL, &bufdescs[0],
NULL // source = unused, bind histogram result
},
{VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET, NULL, Unwrap(m_Histogram.m_HistogramDescSet[0]), 2,
0, 1, VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, NULL, &bufdescs[1], NULL},
// sampled view
{VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET, NULL, Unwrap(m_Histogram.m_HistogramDescSet[0]),
descSetBinding, 0, 1, VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, &imdesc, NULL, NULL},
// YUV secondary planes (if needed)
{VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET, NULL, Unwrap(m_Histogram.m_HistogramDescSet[0]), 10,
0, GetYUVPlaneCount(texviews.castedFormat) - 1, VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER,
altimdesc, NULL, NULL},
};
rdcarray<VkWriteDescriptorSet> writeSets;
for(size_t i = 0; i < ARRAY_COUNT(writeSet); i++)
{
if(writeSet[i].descriptorCount > 0)
writeSets.push_back(writeSet[i]);
}
for(size_t i = 0; i < ARRAY_COUNT(m_TexRender.DummyWrites); i++)
{
VkWriteDescriptorSet &write = m_TexRender.DummyWrites[i];
// don't write dummy data in the actual slot
if(write.dstBinding == descSetBinding)
continue;
// don't overwrite YUV texture slots if it's a YUV planar format
if(write.dstBinding == 10)
{
if(write.dstArrayElement == 0 && GetYUVPlaneCount(texviews.castedFormat) >= 2)
continue;
if(write.dstArrayElement == 1 && GetYUVPlaneCount(texviews.castedFormat) >= 3)
continue;
}
write.dstSet = Unwrap(m_Histogram.m_HistogramDescSet[0]);
writeSets.push_back(write);
}
vt->UpdateDescriptorSets(Unwrap(dev), (uint32_t)writeSets.size(), &writeSets[0], 0, NULL);
HistogramUBOData *data = (HistogramUBOData *)m_Histogram.m_HistogramUBO.Map(NULL);
if(!data)
return false;
data->HistogramTextureResolution.x = (float)RDCMAX(uint32_t(iminfo.extent.width) >> sub.mip, 1U);
data->HistogramTextureResolution.y = (float)RDCMAX(uint32_t(iminfo.extent.height) >> sub.mip, 1U);
data->HistogramTextureResolution.z = (float)RDCMAX(uint32_t(iminfo.extent.depth) >> sub.mip, 1U);
if(iminfo.type == VK_IMAGE_TYPE_3D)
data->HistogramSlice =
(float)RDCCLAMP(sub.slice, 0U, uint32_t(iminfo.extent.depth >> sub.mip) - 1) + 0.001f;
else
data->HistogramSlice = (float)RDCCLAMP(sub.slice, 0U, iminfo.arrayLayers - 1) + 0.001f;
data->HistogramMip = (int)sub.mip;
data->HistogramNumSamples = iminfo.samples;
data->HistogramSample = (int)RDCCLAMP(sub.sample, 0U, uint32_t(iminfo.samples) - 1);
if(sub.sample == ~0U)
data->HistogramSample = -iminfo.samples;
data->HistogramMin = minval;
// The calculation in the shader normalises each value between min and max, then multiplies by the
// number of buckets.
// But any value equal to HistogramMax must go into NUM_BUCKETS-1, so add a small delta.
data->HistogramMax = maxval + maxval * 1e-6f;
uint32_t chans = 0;
if(channels[0])
chans |= 0x1;
if(channels[1])
chans |= 0x2;
if(channels[2])
chans |= 0x4;
if(channels[3])
chans |= 0x8;
// shuffle the channel selection, since stencil comes back in red
if(stencil)
chans = 0x1;
data->HistogramChannels = chans;
data->HistogramFlags = 0;
Vec4u YUVDownsampleRate = {};
Vec4u YUVAChannels = {};
GetYUVShaderParameters(texviews.castedFormat, YUVDownsampleRate, YUVAChannels);
data->HistogramYUVDownsampleRate = YUVDownsampleRate;
data->HistogramYUVAChannels = YUVAChannels;
m_Histogram.m_HistogramUBO.Unmap();
VkCommandBufferBeginInfo beginInfo = {VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO, NULL,
VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT};
VkCommandBuffer cmd = m_pDriver->GetNextCmd();
if(cmd == VK_NULL_HANDLE)
return false;
vt->BeginCommandBuffer(Unwrap(cmd), &beginInfo);
ImageBarrierSequence setupBarriers, cleanupBarriers;
state->TempTransition(m_pDriver->m_QueueFamilyIdx, VK_IMAGE_LAYOUT_GENERAL,
VK_ACCESS_SHADER_READ_BIT, setupBarriers, cleanupBarriers,
m_pDriver->GetImageTransitionInfo());
m_pDriver->InlineSetupImageBarriers(cmd, setupBarriers);
m_pDriver->SubmitAndFlushImageStateBarriers(setupBarriers);
int blocksX = (int)ceil(iminfo.extent.width / float(HGRAM_PIXELS_PER_TILE * HGRAM_TILES_PER_BLOCK));
int blocksY =
(int)ceil(iminfo.extent.height / float(HGRAM_PIXELS_PER_TILE * HGRAM_TILES_PER_BLOCK));
vt->CmdFillBuffer(Unwrap(cmd), Unwrap(m_Histogram.m_HistogramBuf.buf), 0,
m_Histogram.m_HistogramBuf.totalsize, 0);
vt->CmdBindPipeline(Unwrap(cmd), VK_PIPELINE_BIND_POINT_COMPUTE,
Unwrap(m_Histogram.m_HistogramPipe[textype][intTypeIndex]));
vt->CmdBindDescriptorSets(Unwrap(cmd), VK_PIPELINE_BIND_POINT_COMPUTE,
Unwrap(m_Histogram.m_HistogramPipeLayout), 0, 1,
UnwrapPtr(m_Histogram.m_HistogramDescSet[0]), 0, NULL);
vt->CmdDispatch(Unwrap(cmd), blocksX, blocksY, 1);
m_pDriver->InlineCleanupImageBarriers(cmd, cleanupBarriers);
if(!cleanupBarriers.empty())
{
vt->EndCommandBuffer(Unwrap(cmd));
m_pDriver->SubmitCmds();
m_pDriver->FlushQ();
m_pDriver->SubmitAndFlushImageStateBarriers(cleanupBarriers);
cmd = m_pDriver->GetNextCmd();
if(cmd == VK_NULL_HANDLE)
return false;
vt->BeginCommandBuffer(Unwrap(cmd), &beginInfo);
}
VkBufferMemoryBarrier tilebarrier = {
VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,
NULL,
VK_ACCESS_SHADER_WRITE_BIT,
VK_ACCESS_TRANSFER_READ_BIT,
VK_QUEUE_FAMILY_IGNORED,
VK_QUEUE_FAMILY_IGNORED,
Unwrap(m_Histogram.m_HistogramBuf.buf),
0,
m_Histogram.m_HistogramBuf.totalsize,
};
// ensure shader writes complete before copying to readback buf
DoPipelineBarrier(cmd, 1, &tilebarrier);
VkBufferCopy bufcopy = {
0,
0,
m_Histogram.m_HistogramBuf.totalsize,
};
vt->CmdCopyBuffer(Unwrap(cmd), Unwrap(m_Histogram.m_HistogramBuf.buf),
Unwrap(m_Histogram.m_HistogramReadback.buf), 1, &bufcopy);
// wait for copy to complete before mapping
tilebarrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
tilebarrier.dstAccessMask = VK_ACCESS_HOST_READ_BIT;
tilebarrier.buffer = Unwrap(m_Histogram.m_HistogramReadback.buf);
tilebarrier.size = m_Histogram.m_HistogramReadback.totalsize;
DoPipelineBarrier(cmd, 1, &tilebarrier);
vt->EndCommandBuffer(Unwrap(cmd));
// submit cmds and wait for idle so we can readback
m_pDriver->SubmitCmds();
m_pDriver->FlushQ();
uint32_t *buckets = (uint32_t *)m_Histogram.m_HistogramReadback.Map(NULL);
if(!buckets)
return false;
histogram.assign(buckets, HGRAM_NUM_BUCKETS);
m_Histogram.m_HistogramReadback.Unmap();
return true;
}
rdcarray<EventUsage> VulkanReplay::GetUsage(ResourceId id)
{
return m_pDriver->GetUsage(id);
}
void VulkanReplay::CopyPixelForPixelHistory(VkCommandBuffer cmd, VkOffset2D offset, uint32_t sample,
uint32_t bufferOffset, VkFormat format,
VkDescriptorSet descSet)
{
VkPipeline pipe;
if(IsDepthOrStencilFormat(format))
pipe = m_PixelHistory.MSCopyDepthPipe;
else
pipe = m_PixelHistory.MSCopyPipe;
if(pipe == VK_NULL_HANDLE)
return;
if(!m_pDriver->GetDeviceEnabledFeatures().shaderStorageImageWriteWithoutFormat)
return;
ObjDisp(cmd)->CmdBindPipeline(Unwrap(cmd), VK_PIPELINE_BIND_POINT_COMPUTE, Unwrap(pipe));
int32_t params[8] = {(int32_t)sample,
offset.x,
offset.y,
(int32_t)bufferOffset,
!IsStencilOnlyFormat(format),
IsStencilFormat(format),
0,
0};
ObjDisp(cmd)->CmdBindDescriptorSets(Unwrap(cmd), VK_PIPELINE_BIND_POINT_COMPUTE,
Unwrap(m_PixelHistory.MSCopyPipeLayout), 0, 1,
UnwrapPtr(descSet), 0, NULL);
ObjDisp(cmd)->CmdPushConstants(Unwrap(cmd), Unwrap(m_PixelHistory.MSCopyPipeLayout),
VK_SHADER_STAGE_ALL, 0, 8 * 4, params);
ObjDisp(cmd)->CmdDispatch(Unwrap(cmd), 1, 1, 1);
}
void VulkanReplay::GetTextureData(ResourceId tex, const Subresource &sub,
const GetTextureDataParams &params, bytebuf &data)
{
bool wasms = false;
bool resolve = params.resolve;
bool copyToBuffer = true;
if(m_pDriver->m_CreationInfo.m_Image.find(tex) == m_pDriver->m_CreationInfo.m_Image.end())
{
RDCERR("Trying to get texture data for unknown ID %s!", ToStr(tex).c_str());
return;
}
const VulkanCreationInfo::Image &imInfo = m_pDriver->m_CreationInfo.m_Image[tex];
LockedConstImageStateRef lockedImage = m_pDriver->FindConstImageState(tex);
if(!lockedImage || !lockedImage->isMemoryBound)
return;
const ImageState *srcImageState = &*lockedImage;
ImageState tmpImageState;
VkMarkerRegion region(StringFormat::Fmt("GetTextureData(%u, %u, %u, remap=%d)", sub.mip,
sub.slice, sub.sample, params.remap));
Subresource s = sub;
s.slice = RDCMIN(uint32_t(imInfo.arrayLayers - 1), s.slice);
s.sample = RDCMIN(uint32_t(imInfo.samples - 1), s.sample);
s.mip = RDCMIN(uint32_t(imInfo.mipLevels - 1), s.mip);
VkImageCreateInfo imCreateInfo = {
VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,
NULL,
0,
imInfo.type,
imInfo.format,
imInfo.extent,
imInfo.mipLevels,
imInfo.arrayLayers,
imInfo.samples,
VK_IMAGE_TILING_OPTIMAL,
VK_IMAGE_USAGE_TRANSFER_SRC_BIT | VK_IMAGE_USAGE_TRANSFER_DST_BIT,
VK_SHARING_MODE_EXCLUSIVE,
0,
NULL,
VK_IMAGE_LAYOUT_UNDEFINED,
};
VkImageAspectFlags imageAspects = FormatImageAspects(imInfo.format);
bool isDepth = (imageAspects & VK_IMAGE_ASPECT_DEPTH_BIT) != 0;
bool isStencil = (imageAspects & VK_IMAGE_ASPECT_STENCIL_BIT) != 0;
bool isPlanar = (imageAspects & VK_IMAGE_ASPECT_PLANE_0_BIT) != 0;
uint32_t planeCount = GetYUVPlaneCount(imInfo.format);
VkImage liveWrappedImage = GetResourceManager()->GetCurrentHandle<VkImage>(tex);
VkImage srcImage = Unwrap(liveWrappedImage);
VkImage tmpImage = VK_NULL_HANDLE;
VkImage wrappedTmpImage = VK_NULL_HANDLE;
VkDeviceMemory tmpMemory = VK_NULL_HANDLE;
VkFramebuffer *tmpFB = NULL;
VkImageView *tmpView = NULL;
uint32_t numFBs = 0;
VkRenderPass tmpRP = VK_NULL_HANDLE;
VkRenderPass tmpRPStencil = VK_NULL_HANDLE;
VkDevice dev = m_pDriver->GetDev();
VkCommandBuffer cmd = m_pDriver->GetNextCmd();
const VkDevDispatchTable *vt = ObjDisp(dev);
if(cmd == VK_NULL_HANDLE)
return;
VkCommandBufferBeginInfo beginInfo = {VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO, NULL,
VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT};
VkResult vkr = vt->BeginCommandBuffer(Unwrap(cmd), &beginInfo);
CheckVkResult(vkr);
uint32_t dataSize = 0;
VkBuffer readbackBuf = VK_NULL_HANDLE;
VkDeviceMemory readbackMem = VK_NULL_HANDLE;
if(imInfo.samples > 1)
{
// make image n-array instead of n-samples
imCreateInfo.arrayLayers *= imCreateInfo.samples;
imCreateInfo.samples = VK_SAMPLE_COUNT_1_BIT;
wasms = true;
}
if(wasms && (isDepth || isStencil))
resolve = false;
if(params.remap != RemapTexture::NoRemap)
{
int renderFlags = 0;
// force readback texture to RGBA8 unorm
if(params.remap == RemapTexture::RGBA8)
{
if(IsSRGBFormat(imCreateInfo.format))
{
imCreateInfo.format = VK_FORMAT_R8G8B8A8_SRGB;
renderFlags |= eTexDisplay_RemapSRGB;
}
else
{
imCreateInfo.format = VK_FORMAT_R8G8B8A8_UNORM;
}
}
else if(params.remap == RemapTexture::RGBA16)
{
imCreateInfo.format = VK_FORMAT_R16G16B16A16_SFLOAT;
renderFlags = eTexDisplay_16Render;
}
else if(params.remap == RemapTexture::RGBA32)
{
imCreateInfo.format = VK_FORMAT_R32G32B32A32_SFLOAT;
renderFlags = eTexDisplay_32Render;
}
else
{
RDCERR("Unsupported remap format: %u", params.remap);
}
imCreateInfo.format = GetViewCastedFormat(imCreateInfo.format, BaseRemapType(params));
if(IsUIntFormat(imCreateInfo.format))
renderFlags |= eTexDisplay_RemapUInt;
else if(IsSIntFormat(imCreateInfo.format))
renderFlags |= eTexDisplay_RemapSInt;
else
renderFlags |= eTexDisplay_RemapFloat;
// force to 1 array slice, 1 mip
imCreateInfo.arrayLayers = 1;
imCreateInfo.mipLevels = 1;
// force to 2D
imCreateInfo.imageType = VK_IMAGE_TYPE_2D;
imCreateInfo.usage |= VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT;
// we'll need to cast to remap the stencil part
if(IsStencilFormat(imInfo.format))
imCreateInfo.flags |= VK_IMAGE_CREATE_MUTABLE_FORMAT_BIT;
imCreateInfo.extent.width = RDCMAX(1U, imCreateInfo.extent.width >> s.mip);
imCreateInfo.extent.height = RDCMAX(1U, imCreateInfo.extent.height >> s.mip);
imCreateInfo.extent.depth = RDCMAX(1U, imCreateInfo.extent.depth >> s.mip);
// convert a 3D texture into a 2D array, so we can render to the slices without needing
// KHR_maintenance1
if(imCreateInfo.extent.depth > 1)
{
imCreateInfo.arrayLayers = imCreateInfo.extent.depth;
imCreateInfo.extent.depth = 1;
}
// create render texture similar to readback texture
vt->CreateImage(Unwrap(dev), &imCreateInfo, NULL, &tmpImage);
wrappedTmpImage = tmpImage;
GetResourceManager()->WrapResource(Unwrap(dev), wrappedTmpImage);
tmpImageState = ImageState(wrappedTmpImage, ImageInfo(imCreateInfo), eFrameRef_None);
NameVulkanObject(wrappedTmpImage, "GetTextureData tmpImage");
VkMemoryRequirements mrq = {0};
vt->GetImageMemoryRequirements(Unwrap(dev), tmpImage, &mrq);
VkMemoryAllocateInfo allocInfo = {
VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO,
NULL,
mrq.size,
m_pDriver->GetGPULocalMemoryIndex(mrq.memoryTypeBits),
};
vkr = vt->AllocateMemory(Unwrap(dev), &allocInfo, NULL, &tmpMemory);
CheckVkResult(vkr);
if(vkr != VK_SUCCESS)
return;
vkr = vt->BindImageMemory(Unwrap(dev), tmpImage, tmpMemory, 0);
CheckVkResult(vkr);
tmpImageState.InlineTransition(
cmd, m_pDriver->m_QueueFamilyIdx, VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL, 0,
VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT, m_pDriver->GetImageTransitionInfo());
// end this command buffer, the rendertexture below will use its own and we want to ensure
// ordering
vt->EndCommandBuffer(Unwrap(cmd));
if(Vulkan_Debug_SingleSubmitFlushing())
m_pDriver->SubmitCmds();
// create framebuffer/render pass to render to
VkAttachmentDescription attDesc = {0,
imCreateInfo.format,
VK_SAMPLE_COUNT_1_BIT,
VK_ATTACHMENT_LOAD_OP_LOAD,
VK_ATTACHMENT_STORE_OP_STORE,
VK_ATTACHMENT_LOAD_OP_DONT_CARE,
VK_ATTACHMENT_STORE_OP_DONT_CARE,
VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL,
VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL};
VkAttachmentReference attRef = {0, VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL};
VkSubpassDescription subpass = {
0, VK_PIPELINE_BIND_POINT_GRAPHICS,
0, NULL, // inputs
1, &attRef, // color
NULL, // resolve
NULL, // depth-stencil
0, NULL, // preserve
};
VkRenderPassCreateInfo rpinfo = {
VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO,
NULL,
0,
1,
&attDesc,
1,
&subpass,
0,
NULL, // dependencies
};
vt->CreateRenderPass(Unwrap(dev), &rpinfo, NULL, &tmpRP);
numFBs = imCreateInfo.arrayLayers;
// we'll need twice as many temp views/FBs for stencil views
if(IsStencilFormat(imInfo.format))
{
tmpFB = new VkFramebuffer[numFBs * 2];
tmpView = new VkImageView[numFBs * 2];
}
else
{
tmpFB = new VkFramebuffer[numFBs];
tmpView = new VkImageView[numFBs];
}
int oldW = m_DebugWidth, oldH = m_DebugHeight;
m_DebugWidth = imCreateInfo.extent.width;
m_DebugHeight = imCreateInfo.extent.height;
int renderCount = 0;
// if 3d texture, render each slice separately, otherwise render once
for(uint32_t i = 0; i < numFBs; i++)
{
if(numFBs > 1 && (renderCount % m_TexRender.UBO.GetRingCount()) == 0)
{
m_pDriver->SubmitCmds();
m_pDriver->FlushQ();
}
TextureDisplay texDisplay;
texDisplay.red = texDisplay.green = texDisplay.blue = texDisplay.alpha = true;
texDisplay.hdrMultiplier = -1.0f;
texDisplay.linearDisplayAsGamma = false;
texDisplay.overlay = DebugOverlay::NoOverlay;
texDisplay.flipY = false;
texDisplay.subresource.mip = s.mip;
texDisplay.subresource.slice = imInfo.type == VK_IMAGE_TYPE_3D ? i : s.slice;
texDisplay.subresource.sample =
imInfo.type == VK_IMAGE_TYPE_3D ? 0 : (resolve ? ~0U : s.sample);
texDisplay.customShaderId = ResourceId();
texDisplay.rangeMin = params.blackPoint;
texDisplay.rangeMax = params.whitePoint;
texDisplay.scale = 1.0f;
texDisplay.resourceId = tex;
texDisplay.typeCast = params.typeCast;
texDisplay.rawOutput = false;
texDisplay.xOffset = 0;
texDisplay.yOffset = 0;
VkImageViewCreateInfo viewInfo = {
VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
NULL,
0,
tmpImage,
VK_IMAGE_VIEW_TYPE_2D,
imCreateInfo.format,
{VK_COMPONENT_SWIZZLE_IDENTITY, VK_COMPONENT_SWIZZLE_IDENTITY,
VK_COMPONENT_SWIZZLE_IDENTITY, VK_COMPONENT_SWIZZLE_IDENTITY},
{
VK_IMAGE_ASPECT_COLOR_BIT,
0,
VK_REMAINING_MIP_LEVELS,
i,
1,
},
};
vkr = vt->CreateImageView(Unwrap(dev), &viewInfo, NULL, &tmpView[i]);
CheckVkResult(vkr);
NameUnwrappedVulkanObject(tmpView[i], "GetTextureData tmpView[i]");
VkFramebufferCreateInfo fbinfo = {
VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO,
NULL,
0,
tmpRP,
1,
&tmpView[i],
(uint32_t)imCreateInfo.extent.width,
(uint32_t)imCreateInfo.extent.height,
1,
};
vkr = vt->CreateFramebuffer(Unwrap(dev), &fbinfo, NULL, &tmpFB[i]);
CheckVkResult(vkr);
VkClearValue clearval = {};
VkRenderPassBeginInfo rpbegin = {
VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
NULL,
tmpRP,
tmpFB[i],
{{
0,
0,
},
{imCreateInfo.extent.width, imCreateInfo.extent.height}},
1,
&clearval,
};
RenderTextureInternal(texDisplay, *srcImageState, rpbegin, renderFlags);
renderCount++;
// for textures with stencil, do another draw to copy the stencil
if(isStencil)
{
viewInfo.format = GetViewCastedFormat(viewInfo.format, CompType::UInt);
attDesc.format = viewInfo.format;
vkr = vt->CreateRenderPass(Unwrap(dev), &rpinfo, NULL, &tmpRPStencil);
CheckVkResult(vkr);
fbinfo.renderPass = tmpRPStencil;
rpbegin.renderPass = tmpRPStencil;
vkr = vt->CreateImageView(Unwrap(dev), &viewInfo, NULL, &tmpView[i + numFBs]);
CheckVkResult(vkr);
NameUnwrappedVulkanObject(tmpView[i + numFBs], "GetTextureData tmpView[i]");
fbinfo.pAttachments = &tmpView[i + numFBs];
vkr = vt->CreateFramebuffer(Unwrap(dev), &fbinfo, NULL, &tmpFB[i + numFBs]);
CheckVkResult(vkr);
rpbegin.framebuffer = tmpFB[i + numFBs];
int stencilFlags = renderFlags;
stencilFlags &= ~eTexDisplay_RemapFloat;
stencilFlags &= ~eTexDisplay_RemapSRGB;
stencilFlags |= eTexDisplay_RemapUInt | eTexDisplay_GreenOnly;
texDisplay.red = texDisplay.blue = texDisplay.alpha = false;
// S8 renders into red
if(IsStencilOnlyFormat(imInfo.format))
{
texDisplay.red = true;
texDisplay.green = false;
stencilFlags &= ~eTexDisplay_GreenOnly;
}
RenderTextureInternal(texDisplay, *srcImageState, rpbegin, stencilFlags);
renderCount++;
}
}
m_DebugWidth = oldW;
m_DebugHeight = oldH;
srcImage = tmpImage;
srcImageState = &tmpImageState;
// fetch a new command buffer for copy & readback
cmd = m_pDriver->GetNextCmd();
if(cmd == VK_NULL_HANDLE)
return;
vkr = vt->BeginCommandBuffer(Unwrap(cmd), &beginInfo);
CheckVkResult(vkr);
tmpImageState.InlineTransition(cmd, m_pDriver->m_QueueFamilyIdx,
VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,
VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT,
VK_ACCESS_TRANSFER_READ_BIT, m_pDriver->GetImageTransitionInfo());
// these have already been selected, don't need to fetch that subresource
// when copying back to readback buffer
s.slice = 0;
s.mip = 0;
// no longer depth, if it was
isDepth = false;
isStencil = false;
}
else if(wasms && resolve)
{
// force to 1 array slice, 1 mip
imCreateInfo.arrayLayers = 1;
imCreateInfo.mipLevels = 1;
imCreateInfo.extent.width = RDCMAX(1U, imCreateInfo.extent.width >> s.mip);
imCreateInfo.extent.height = RDCMAX(1U, imCreateInfo.extent.height >> s.mip);
// create resolve texture
vt->CreateImage(Unwrap(dev), &imCreateInfo, NULL, &tmpImage);
wrappedTmpImage = tmpImage;
GetResourceManager()->WrapResource(Unwrap(dev), wrappedTmpImage);
tmpImageState = ImageState(wrappedTmpImage, ImageInfo(imCreateInfo), eFrameRef_None);
NameVulkanObject(wrappedTmpImage, "GetTextureData tmpImage");
VkMemoryRequirements mrq = {0};
vt->GetImageMemoryRequirements(Unwrap(dev), tmpImage, &mrq);
VkMemoryAllocateInfo allocInfo = {
VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO,
NULL,
mrq.size,
m_pDriver->GetGPULocalMemoryIndex(mrq.memoryTypeBits),
};
vkr = vt->AllocateMemory(Unwrap(dev), &allocInfo, NULL, &tmpMemory);
CheckVkResult(vkr);
if(vkr != VK_SUCCESS)
return;
vkr = vt->BindImageMemory(Unwrap(dev), tmpImage, tmpMemory, 0);
CheckVkResult(vkr);
RDCASSERT(!isDepth && !isStencil);
VkImageResolve resolveRegion = {
{VK_IMAGE_ASPECT_COLOR_BIT, s.mip, s.slice, 1},
{0, 0, 0},
{VK_IMAGE_ASPECT_COLOR_BIT, 0, 0, 1},
{0, 0, 0},
imCreateInfo.extent,
};
tmpImageState.InlineTransition(
cmd, m_pDriver->m_QueueFamilyIdx, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 0,
VK_ACCESS_TRANSFER_WRITE_BIT, m_pDriver->GetImageTransitionInfo());
ImageBarrierSequence setupBarriers, cleanupBarriers;
srcImageState->TempTransition(m_pDriver->m_QueueFamilyIdx, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,
VK_ACCESS_TRANSFER_READ_BIT, setupBarriers, cleanupBarriers,
m_pDriver->GetImageTransitionInfo());
m_pDriver->InlineSetupImageBarriers(cmd, setupBarriers);
m_pDriver->SubmitAndFlushImageStateBarriers(setupBarriers);
// resolve from live texture to resolve texture
vt->CmdResolveImage(Unwrap(cmd), srcImage, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, tmpImage,
VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 1, &resolveRegion);
tmpImageState.InlineTransition(cmd, m_pDriver->m_QueueFamilyIdx,
VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, VK_ACCESS_TRANSFER_WRITE_BIT,
VK_ACCESS_TRANSFER_READ_BIT, m_pDriver->GetImageTransitionInfo());
m_pDriver->InlineCleanupImageBarriers(cmd, cleanupBarriers);
if(!cleanupBarriers.empty())
{
// ensure this resolve happens before handing back the source image to the original queue
vkr = vt->EndCommandBuffer(Unwrap(cmd));
CheckVkResult(vkr);
m_pDriver->SubmitCmds();
m_pDriver->FlushQ();
m_pDriver->SubmitAndFlushImageStateBarriers(cleanupBarriers);
// fetch a new command buffer for remaining work
cmd = m_pDriver->GetNextCmd();
if(cmd == VK_NULL_HANDLE)
return;
vkr = vt->BeginCommandBuffer(Unwrap(cmd), &beginInfo);
CheckVkResult(vkr);
}
srcImageState = &tmpImageState;
srcImage = tmpImage;
// these have already been selected, don't need to fetch that subresource
// when copying back to readback buffer
s.slice = 0;
s.mip = 0;
}
else if(wasms)
{
dataSize = GetByteSize(imInfo.extent.width, imInfo.extent.height, imInfo.extent.depth,
imCreateInfo.format, s.mip);
// buffer size needs to be align to the int for shader writing
VkBufferCreateInfo bufInfo = {
VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
NULL,
0,
AlignUp(dataSize, 4U),
VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT |
VK_BUFFER_USAGE_TRANSFER_DST_BIT,
};
vkr = vt->CreateBuffer(Unwrap(dev), &bufInfo, NULL, &readbackBuf);
CheckVkResult(vkr);
VkMemoryRequirements mrq = {0};
vt->GetBufferMemoryRequirements(Unwrap(dev), readbackBuf, &mrq);
VkMemoryAllocateInfo allocInfo = {
VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO,
NULL,
mrq.size,
m_pDriver->GetReadbackMemoryIndex(mrq.memoryTypeBits),
};
vkr = vt->AllocateMemory(Unwrap(dev), &allocInfo, NULL, &readbackMem);
CheckVkResult(vkr);
if(vkr != VK_SUCCESS)
return;
vkr = vt->BindBufferMemory(Unwrap(dev), readbackBuf, readbackMem, 0);
CheckVkResult(vkr);
// copy/expand multisampled live texture to readback buffer
ImageBarrierSequence setupBarriers, cleanupBarriers;
srcImageState->TempTransition(m_pDriver->m_QueueFamilyIdx,
VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL,
VK_ACCESS_SHADER_READ_BIT, setupBarriers, cleanupBarriers,
m_pDriver->GetImageTransitionInfo());
m_pDriver->InlineSetupImageBarriers(cmd, setupBarriers);
m_pDriver->SubmitAndFlushImageStateBarriers(setupBarriers);
GetDebugManager()->CopyTex2DMSToBuffer(cmd, readbackBuf, srcImage, imCreateInfo.extent, s.slice,
1, s.sample, 1, imCreateInfo.format);
m_pDriver->InlineCleanupImageBarriers(cmd, cleanupBarriers);
if(!cleanupBarriers.empty())
{
// ensure this resolve happens before handing back the source image to the original queue
vkr = vt->EndCommandBuffer(Unwrap(cmd));
CheckVkResult(vkr);
m_pDriver->SubmitCmds();
m_pDriver->FlushQ();
m_pDriver->SubmitAndFlushImageStateBarriers(cleanupBarriers);
// fetch a new command buffer for remaining work
cmd = m_pDriver->GetNextCmd();
if(cmd == VK_NULL_HANDLE)
return;
vkr = vt->BeginCommandBuffer(Unwrap(cmd), &beginInfo);
CheckVkResult(vkr);
}
// readback buffer has already been populated, no need to call CmdCopyImageToBuffer
copyToBuffer = false;
}
VkDeviceSize stencilOffset = 0;
// if we have no tmpImage, we're copying directly from the real image
if(copyToBuffer)
{
ImageBarrierSequence cleanupBarriers;
if(tmpImage == VK_NULL_HANDLE)
{
ImageBarrierSequence setupBarriers;
srcImageState->TempTransition(m_pDriver->m_QueueFamilyIdx, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,
VK_ACCESS_TRANSFER_READ_BIT, setupBarriers, cleanupBarriers,
m_pDriver->GetImageTransitionInfo());
m_pDriver->InlineSetupImageBarriers(cmd, setupBarriers);
m_pDriver->SubmitAndFlushImageStateBarriers(setupBarriers);
}
rdcarray<VkBufferImageCopy> copyregions;
VkBufferImageCopy copyRegionTemplate = {
0,
0,
0,
{VK_IMAGE_ASPECT_NONE, s.mip, s.slice, 1},
{
0,
0,
0,
},
{RDCMAX(1U, imCreateInfo.extent.width >> s.mip),
RDCMAX(1U, imCreateInfo.extent.height >> s.mip),
RDCMAX(1U, imCreateInfo.extent.depth >> s.mip)},
};
if(isDepth || isStencil)
{
if(isDepth)
{
copyRegionTemplate.imageSubresource.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT;
copyregions.push_back(copyRegionTemplate);
// Stencil offset (if present)
copyRegionTemplate.bufferOffset = stencilOffset =
GetByteSize(imInfo.extent.width, imInfo.extent.height, imInfo.extent.depth,
GetDepthOnlyFormat(imCreateInfo.format), s.mip);
copyRegionTemplate.bufferOffset = AlignUp(copyRegionTemplate.bufferOffset, (VkDeviceSize)4);
}
if(isStencil)
{
copyRegionTemplate.imageSubresource.aspectMask = VK_IMAGE_ASPECT_STENCIL_BIT;
copyregions.push_back(copyRegionTemplate);
}
}
else if(isPlanar)
{
for(uint32_t i = 0; i < planeCount; i++)
{
copyRegionTemplate.imageSubresource.aspectMask = VK_IMAGE_ASPECT_PLANE_0_BIT << i;
VkExtent2D planeExtent =
GetPlaneShape(RDCMAX(1U, imCreateInfo.extent.width >> s.mip),
RDCMAX(1U, imCreateInfo.extent.height >> s.mip), imCreateInfo.format, i);
copyRegionTemplate.imageExtent.width = planeExtent.width;
copyRegionTemplate.imageExtent.height = planeExtent.height;
copyregions.push_back(copyRegionTemplate);
copyRegionTemplate.bufferOffset +=
GetPlaneByteSize(imCreateInfo.extent.width, imCreateInfo.extent.height,
imCreateInfo.extent.depth, imCreateInfo.format, s.mip, i);
}
}
else
{
copyRegionTemplate.imageSubresource.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
copyregions.push_back(copyRegionTemplate);
}
dataSize = GetByteSize(imInfo.extent.width, imInfo.extent.height, imInfo.extent.depth,
imCreateInfo.format, s.mip);
if(imCreateInfo.format == VK_FORMAT_D24_UNORM_S8_UINT)
{
// for most combined depth-stencil images this will be large enough for both to be copied
// separately, but for D24S8 we need to add extra space since they won't be copied packed
dataSize = AlignUp(dataSize, 4U);
dataSize += GetByteSize(imInfo.extent.width, imInfo.extent.height, imInfo.extent.depth,
VK_FORMAT_S8_UINT, s.mip);
}
VkBufferCreateInfo bufInfo = {
VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
NULL,
0,
dataSize,
VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT,
};
vkr = vt->CreateBuffer(Unwrap(dev), &bufInfo, NULL, &readbackBuf);
CheckVkResult(vkr);
VkMemoryRequirements mrq = {0};
vt->GetBufferMemoryRequirements(Unwrap(dev), readbackBuf, &mrq);
VkMemoryAllocateInfo allocInfo = {
VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO,
NULL,
mrq.size,
m_pDriver->GetReadbackMemoryIndex(mrq.memoryTypeBits),
};
vkr = vt->AllocateMemory(Unwrap(dev), &allocInfo, NULL, &readbackMem);
CheckVkResult(vkr);
if(vkr != VK_SUCCESS)
return;
vkr = vt->BindBufferMemory(Unwrap(dev), readbackBuf, readbackMem, 0);
CheckVkResult(vkr);
if(imInfo.type == VK_IMAGE_TYPE_3D && params.remap != RemapTexture::NoRemap)
{
// copy in each slice from the 2D array we created to render out the 3D texture
for(uint32_t i = 0; i < imCreateInfo.arrayLayers; i++)
{
copyregions[0].imageSubresource.baseArrayLayer = i;
copyregions[0].bufferOffset =
i * GetByteSize(imCreateInfo.extent.width, imCreateInfo.extent.height, 1,
imCreateInfo.format, s.mip);
vt->CmdCopyImageToBuffer(Unwrap(cmd), srcImage, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,
readbackBuf, (uint32_t)copyregions.size(), copyregions.data());
}
}
else
{
if(imInfo.type == VK_IMAGE_TYPE_3D)
copyregions[0].imageSubresource.baseArrayLayer = 0;
// copy from desired subresource in srcImage to buffer
vt->CmdCopyImageToBuffer(Unwrap(cmd), srcImage, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,
readbackBuf, (uint32_t)copyregions.size(), copyregions.data());
}
// if we have no tmpImage, we're copying directly from the real image
if(tmpImage == VK_NULL_HANDLE)
{
m_pDriver->InlineCleanupImageBarriers(cmd, cleanupBarriers);
if(!cleanupBarriers.empty())
{
// ensure this resolve happens before handing back the source image to the original queue
vkr = vt->EndCommandBuffer(Unwrap(cmd));
CheckVkResult(vkr);
m_pDriver->SubmitCmds();
m_pDriver->FlushQ();
m_pDriver->SubmitAndFlushImageStateBarriers(cleanupBarriers);
// fetch a new command buffer for remaining work
cmd = m_pDriver->GetNextCmd();
if(cmd == VK_NULL_HANDLE)
return;
vkr = vt->BeginCommandBuffer(Unwrap(cmd), &beginInfo);
CheckVkResult(vkr);
}
}
}
VkBufferMemoryBarrier bufBarrier = {
VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,
NULL,
VK_ACCESS_TRANSFER_WRITE_BIT,
VK_ACCESS_HOST_READ_BIT,
VK_QUEUE_FAMILY_IGNORED,
VK_QUEUE_FAMILY_IGNORED,
readbackBuf,
0,
dataSize,
};
// wait for copy to finish before reading back to host
DoPipelineBarrier(cmd, 1, &bufBarrier);
vt->EndCommandBuffer(Unwrap(cmd));
m_pDriver->SubmitCmds();
m_pDriver->FlushQ();
// map the buffer and copy to return buffer
byte *pData = NULL;
vkr = vt->MapMemory(Unwrap(dev), readbackMem, 0, VK_WHOLE_SIZE, 0, (void **)&pData);
CheckVkResult(vkr);
if(vkr != VK_SUCCESS)
return;
if(!pData)
{
RDCERR("Manually reporting failed memory map");
CheckVkResult(VK_ERROR_MEMORY_MAP_FAILED);
return;
}
VkMappedMemoryRange range = {
VK_STRUCTURE_TYPE_MAPPED_MEMORY_RANGE, NULL, readbackMem, 0, VK_WHOLE_SIZE,
};
vkr = vt->InvalidateMappedMemoryRanges(Unwrap(dev), 1, &range);
CheckVkResult(vkr);
RDCASSERT(pData != NULL);
data.resize(dataSize);
if(params.remap == RemapTexture::RGBA32 && IsDepthAndStencilFormat(imInfo.format))
{
memcpy(data.data(), pData, dataSize);
Vec4f *output = (Vec4f *)data.data();
Vec4u *input = (Vec4u *)pData;
for(size_t i = 0; i < dataSize / sizeof(Vec4u); i++)
output[i].y = float(input[i].y) / 255.0f;
}
else if(isDepth && isStencil && copyToBuffer)
{
// We only need to manually interleave if we use CmdCopyImageToBuffer.
// CopyDepthTex2DMS2Buffer will produce interleaved results.
size_t pixelCount = std::max(1U, imCreateInfo.extent.width >> s.mip) *
std::max(1U, imCreateInfo.extent.height >> s.mip) *
std::max(1U, imCreateInfo.extent.depth >> s.mip);
// for some reason reading direct from mapped memory here is *super* slow on android (1.5s to
// iterate over the image), so we memcpy to a temporary buffer.
rdcarray<byte> tmp;
tmp.resize((size_t)stencilOffset + pixelCount * sizeof(uint8_t));
memcpy(tmp.data(), pData, tmp.size());
if(imCreateInfo.format == VK_FORMAT_D16_UNORM_S8_UINT)
{
uint16_t *dSrc = (uint16_t *)tmp.data();
uint8_t *sSrc = (uint8_t *)(tmp.data() + stencilOffset);
uint16_t *dDst = (uint16_t *)data.data();
uint16_t *sDst = dDst + 1; // interleaved, next pixel
for(size_t i = 0; i < pixelCount; i++)
{
*dDst = *dSrc;
*sDst = *sSrc;
// increment source pointers by 1 since they're separate, and dest pointers by 2 since
// they're interleaved
dDst += 2;
sDst += 2;
sSrc++;
dSrc++;
}
}
else if(imCreateInfo.format == VK_FORMAT_D24_UNORM_S8_UINT)
{
// we can copy the depth from D24 as a 32-bit integer, since the remaining bits are garbage
// and we overwrite them with stencil
uint32_t *dSrc = (uint32_t *)tmp.data();
uint8_t *sSrc = (uint8_t *)(tmp.data() + stencilOffset);
uint32_t *dst = (uint32_t *)data.data();
for(size_t i = 0; i < pixelCount; i++)
{
// pack the data together again, stencil in top bits
*dst = (*dSrc & 0x00ffffff) | (uint32_t(*sSrc) << 24);
dst++;
sSrc++;
dSrc++;
}
}
else
{
uint32_t *dSrc = (uint32_t *)tmp.data();
uint8_t *sSrc = (uint8_t *)(tmp.data() + stencilOffset);
uint32_t *dDst = (uint32_t *)data.data();
uint32_t *sDst = dDst + 1; // interleaved, next pixel
for(size_t i = 0; i < pixelCount; i++)
{
*dDst = *dSrc;
*sDst = *sSrc;
// increment source pointers by 1 since they're separate, and dest pointers by 2 since
// they're interleaved
dDst += 2;
sDst += 2;
sSrc++;
dSrc++;
}
}
// need to manually copy to interleave pixels
}
else
{
memcpy(data.data(), pData, dataSize);
// vulkan's bitpacking of some layouts puts alpha in the low bits, which is not our 'standard'
// layout and is not representable in our resource formats
if(params.standardLayout)
{
if(imCreateInfo.format == VK_FORMAT_R4G4B4A4_UNORM_PACK16 ||
imCreateInfo.format == VK_FORMAT_B4G4R4A4_UNORM_PACK16)
{
uint16_t *ptr = (uint16_t *)data.data();
for(uint32_t i = 0; i < dataSize; i += sizeof(uint16_t))
{
const uint16_t val = *ptr;
*ptr = (val >> 4) | ((val & 0xf) << 12);
ptr++;
}
}
else if(imCreateInfo.format == VK_FORMAT_R5G5B5A1_UNORM_PACK16 ||
imCreateInfo.format == VK_FORMAT_B5G5R5A1_UNORM_PACK16)
{
uint16_t *ptr = (uint16_t *)data.data();
for(uint32_t i = 0; i < dataSize; i += sizeof(uint16_t))
{
const uint16_t val = *ptr;
*ptr = (val >> 1) | ((val & 0x1) << 15);
ptr++;
}
}
}
}
vt->UnmapMemory(Unwrap(dev), readbackMem);
// clean up temporary objects
vt->DestroyBuffer(Unwrap(dev), readbackBuf, NULL);
vt->FreeMemory(Unwrap(dev), readbackMem, NULL);
if(tmpImage != VK_NULL_HANDLE)
{
GetResourceManager()->ReleaseWrappedResource(wrappedTmpImage, true);
vt->DestroyImage(Unwrap(dev), tmpImage, NULL);
vt->FreeMemory(Unwrap(dev), tmpMemory, NULL);
}
if(tmpFB != NULL)
{
if(IsStencilFormat(imInfo.format))
numFBs *= 2;
for(uint32_t i = 0; i < numFBs; i++)
{
vt->DestroyFramebuffer(Unwrap(dev), tmpFB[i], NULL);
vt->DestroyImageView(Unwrap(dev), tmpView[i], NULL);
}
delete[] tmpFB;
delete[] tmpView;
vt->DestroyRenderPass(Unwrap(dev), tmpRP, NULL);
vt->DestroyRenderPass(Unwrap(dev), tmpRPStencil, NULL);
}
}
void VulkanReplay::SetCustomShaderIncludes(const rdcarray<rdcstr> &directories)
{
}
void VulkanReplay::BuildCustomShader(ShaderEncoding sourceEncoding, const bytebuf &source,
const rdcstr &entry, const ShaderCompileFlags &compileFlags,
ShaderStage type, ResourceId &id, rdcstr &errors)
{
if(sourceEncoding == ShaderEncoding::GLSL)
{
rdcstr sourceText = InsertSnippetAfterVersion(ShaderType::Vulkan, (const char *)source.data(),
source.count(), GLSL_CUSTOM_PREFIX);
bytebuf patchedSource;
patchedSource.assign((byte *)sourceText.begin(), sourceText.size());
return BuildTargetShader(sourceEncoding, patchedSource, entry, compileFlags, type, id, errors);
}
BuildTargetShader(sourceEncoding, source, entry, compileFlags, type, id, errors);
}
void VulkanReplay::FreeCustomShader(ResourceId id)
{
if(id == ResourceId())
return;
m_pDriver->ReleaseResource(GetResourceManager()->GetCurrentResource(id));
}
ResourceId VulkanReplay::ApplyCustomShader(TextureDisplay &display)
{
if(display.customShaderId == ResourceId() || display.resourceId == ResourceId())
return ResourceId();
VulkanCreationInfo::Image &iminfo = m_pDriver->m_CreationInfo.m_Image[display.resourceId];
GetDebugManager()->CreateCustomShaderTex(iminfo.extent.width, iminfo.extent.height,
display.subresource.mip);
int oldW = m_DebugWidth, oldH = m_DebugHeight;
m_DebugWidth = RDCMAX(1U, iminfo.extent.width);
m_DebugHeight = RDCMAX(1U, iminfo.extent.height);
TextureDisplay disp;
disp.red = disp.green = disp.blue = disp.alpha = true;
disp.flipY = false;
disp.xOffset = 0.0f;
disp.yOffset = 0.0f;
disp.customShaderId = display.customShaderId;
disp.resourceId = display.resourceId;
disp.typeCast = display.typeCast;
disp.hdrMultiplier = -1.0f;
disp.linearDisplayAsGamma = false;
disp.subresource = display.subresource;
disp.overlay = DebugOverlay::NoOverlay;
disp.rangeMin = 0.0f;
disp.rangeMax = 1.0f;
disp.rawOutput = false;
disp.scale = 1.0f;
VkClearValue clearval = {{{0.0f, 0.0f, 0.0f, 1.0f}}};
VkRenderPassBeginInfo rpbegin = {
VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
NULL,
Unwrap(GetDebugManager()->GetCustomRenderpass()),
Unwrap(GetDebugManager()->GetCustomFramebuffer()),
{{
0,
0,
},
{RDCMAX(1U, iminfo.extent.width >> display.subresource.mip),
RDCMAX(1U, iminfo.extent.height >> display.subresource.mip)}},
1,
&clearval,
};
LockedConstImageStateRef imageState = m_pDriver->FindConstImageState(display.resourceId);
if(!imageState)
{
RDCWARN("Could not find image info for image %s", ToStr(display.resourceId).c_str());
return ResourceId();
}
if(!imageState->isMemoryBound)
return ResourceId();
RenderTextureInternal(disp, *imageState, rpbegin, eTexDisplay_MipShift);
m_DebugWidth = oldW;
m_DebugHeight = oldH;
return GetResID(GetDebugManager()->GetCustomTexture());
}
rdcarray<ShaderSourcePrefix> VulkanReplay::GetCustomShaderSourcePrefixes()
{
// this is a complete hack, since we *do* want to define a prefix for GLSL. However GLSL sucks
// and has the #version as the first thing, so we can't do a simple prepend of some defines.
// Instead we will return no prefix and insert our own in BuildCustomShader if we see GLSL
// coming in.
// For SPIR-V no prefix is needed (or possible)
// For HLSL however we define our HLSL prefix so that custom-compiled HLSL to SPIR-V gets the
// right binding and helper definitions
return {
{ShaderEncoding::HLSL, HLSL_CUSTOM_PREFIX},
{ShaderEncoding::Slang, HLSL_CUSTOM_PREFIX},
};
}
void VulkanReplay::BuildTargetShader(ShaderEncoding sourceEncoding, const bytebuf &source,
const rdcstr &entry, const ShaderCompileFlags &compileFlags,
ShaderStage type, ResourceId &id, rdcstr &errors)
{
rdcarray<uint32_t> spirv;
if(sourceEncoding == ShaderEncoding::GLSL)
{
rdcspv::ShaderStage stage = rdcspv::ShaderStage::Invalid;
switch(type)
{
case ShaderStage::Vertex: stage = rdcspv::ShaderStage::Vertex; break;
case ShaderStage::Hull: stage = rdcspv::ShaderStage::TessControl; break;
case ShaderStage::Domain: stage = rdcspv::ShaderStage::TessEvaluation; break;
case ShaderStage::Geometry: stage = rdcspv::ShaderStage::Geometry; break;
case ShaderStage::Pixel: stage = rdcspv::ShaderStage::Fragment; break;
case ShaderStage::Compute: stage = rdcspv::ShaderStage::Compute; break;
case ShaderStage::Task: stage = rdcspv::ShaderStage::Task; break;
case ShaderStage::Mesh: stage = rdcspv::ShaderStage::Mesh; break;
default:
RDCERR("Unexpected type in BuildShader!");
id = ResourceId();
return;
}
rdcarray<rdcstr> sources;
sources.push_back(rdcstr((char *)source.begin(), source.size()));
rdcspv::CompilationSettings settings(rdcspv::InputLanguage::VulkanGLSL, stage);
rdcstr output = rdcspv::Compile(settings, sources, spirv);
if(spirv.empty())
{
id = ResourceId();
errors = output;
return;
}
}
else
{
spirv.resize(source.size() / 4);
memcpy(&spirv[0], source.data(), source.size());
}
VkShaderModuleCreateInfo modinfo = {
VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO,
NULL,
0,
spirv.size() * sizeof(uint32_t),
&spirv[0],
};
VkShaderModule module;
VkResult vkr = m_pDriver->vkCreateShaderModule(m_pDriver->GetDev(), &modinfo, NULL, &module);
CheckVkResult(vkr);
id = GetResID(module);
}
void VulkanReplay::FreeTargetResource(ResourceId id)
{
if(id == ResourceId())
return;
m_pDriver->ReleaseResource(GetResourceManager()->GetCurrentResource(id));
}
void VulkanReplay::ReplaceResource(ResourceId from, ResourceId to)
{
// replace the shader module
m_pDriver->GetResourceManager()->ReplaceResource(from, to);
// now update any derived resources
RefreshDerivedReplacements();
ClearPostVSCache();
ClearFeedbackCache();
}
void VulkanReplay::RemoveReplacement(ResourceId id)
{
if(m_pDriver->GetResourceManager()->HasReplacement(id))
{
m_pDriver->GetResourceManager()->RemoveReplacement(id);
RefreshDerivedReplacements();
ClearPostVSCache();
ClearFeedbackCache();
}
}
void VulkanReplay::RefreshDerivedReplacements()
{
VkDevice dev = m_pDriver->GetDev();
VulkanResourceManager *rm = m_pDriver->GetResourceManager();
// we defer deletes of old replaced resources since it will invalidate elements in the vector
// we're iterating
rdcarray<VkPipeline> deletequeue;
// remake and replace any pipelines that reference a replaced shader
for(auto it = m_pDriver->m_CreationInfo.m_Pipeline.begin();
it != m_pDriver->m_CreationInfo.m_Pipeline.end(); ++it)
{
ResourceId pipesrcid = it->first;
const VulkanCreationInfo::Pipeline &pipeInfo = it->second;
ResourceId origsrcid = rm->GetOriginalID(pipesrcid);
// only look at pipelines from the capture, no replay-time programs.
if(origsrcid == pipesrcid)
continue;
// if this pipeline has a replacement, remove it and delete the program generated for it
if(rm->HasReplacement(origsrcid))
{
deletequeue.push_back(rm->GetLiveHandle<VkPipeline>(origsrcid));
rm->RemoveReplacement(origsrcid);
}
bool usesReplacedShader = false;
for(size_t i = 0; i < ARRAY_COUNT(it->second.shaders); i++)
{
if(rm->HasReplacement(rm->GetOriginalID(it->second.shaders[i].module)))
{
usesReplacedShader = true;
break;
}
}
// if there are replaced shaders in use, create a new pipeline with any/all replaced shaders.
if(usesReplacedShader)
{
VkPipeline pipe = VK_NULL_HANDLE;
// check if this is a graphics or compute pipeline
if(pipeInfo.graphicsPipe)
{
VkGraphicsPipelineCreateInfo pipeCreateInfo;
m_pDriver->GetShaderCache()->MakeGraphicsPipelineInfo(pipeCreateInfo, it->first);
rdcarray<rdcstr> entrynames;
entrynames.reserve(pipeCreateInfo.stageCount);
// replace the modules by going via the live ID to pick up any replacements
for(uint32_t i = 0; i < pipeCreateInfo.stageCount; i++)
{
VkPipelineShaderStageCreateInfo &sh =
(VkPipelineShaderStageCreateInfo &)pipeCreateInfo.pStages[i];
ResourceId shadOrigId = rm->GetOriginalID(GetResID(sh.module));
sh.module = rm->GetLiveHandle<VkShaderModule>(shadOrigId);
if(rm->HasReplacement(shadOrigId))
{
rdcarray<ShaderEntryPoint> entries =
m_pDriver->m_CreationInfo.m_ShaderModule[GetResID(sh.module)].spirv.EntryPoints();
if(entries.size() > 1)
{
if(entries.contains({sh.pName, ShaderStage(StageIndex(sh.stage))}))
{
// nothing to do!
}
else
{
RDCWARN(
"Multiple entry points in edited shader, none matching original, using first "
"one '%s'",
entries[0].name.c_str());
entrynames.push_back(entries[0].name);
sh.pName = entrynames.back().c_str();
}
}
else
{
entrynames.push_back(entries[0].name);
sh.pName = entrynames.back().c_str();
}
}
}
// if we have pipeline executable properties, capture the data
if(m_pDriver->GetExtensions(NULL).ext_KHR_pipeline_executable_properties)
{
pipeCreateInfo.flags |= (VK_PIPELINE_CREATE_CAPTURE_STATISTICS_BIT_KHR |
VK_PIPELINE_CREATE_CAPTURE_INTERNAL_REPRESENTATIONS_BIT_KHR);
}
// create the new graphics pipeline
VkResult vkr = m_pDriver->vkCreateGraphicsPipelines(dev, VK_NULL_HANDLE, 1, &pipeCreateInfo,
NULL, &pipe);
CheckVkResult(vkr);
}
else
{
VkComputePipelineCreateInfo pipeCreateInfo;
m_pDriver->GetShaderCache()->MakeComputePipelineInfo(pipeCreateInfo, it->first);
// replace the module by going via the live ID to pick up any replacements
VkPipelineShaderStageCreateInfo &sh = pipeCreateInfo.stage;
ResourceId shadOrigId = rm->GetOriginalID(pipeInfo.shaders[5].module);
sh.module = rm->GetLiveHandle<VkShaderModule>(shadOrigId);
rdcarray<ShaderEntryPoint> entries;
if(rm->HasReplacement(shadOrigId))
{
entries = m_pDriver->m_CreationInfo.m_ShaderModule[GetResID(sh.module)].spirv.EntryPoints();
if(entries.size() > 1)
{
if(entries.contains({sh.pName, ShaderStage(StageIndex(sh.stage))}))
{
// nothing to do!
}
else
{
RDCWARN(
"Multiple entry points in edited shader, none matching original, using first "
"one '%s'",
entries[0].name.c_str());
sh.pName = entries[0].name.c_str();
}
}
else
{
sh.pName = entries[0].name.c_str();
}
}
// if we have pipeline executable properties, capture the data
if(m_pDriver->GetExtensions(NULL).ext_KHR_pipeline_executable_properties)
{
pipeCreateInfo.flags |= (VK_PIPELINE_CREATE_CAPTURE_STATISTICS_BIT_KHR |
VK_PIPELINE_CREATE_CAPTURE_INTERNAL_REPRESENTATIONS_BIT_KHR);
}
// create the new compute pipeline
VkResult vkr = m_pDriver->vkCreateComputePipelines(dev, VK_NULL_HANDLE, 1, &pipeCreateInfo,
NULL, &pipe);
CheckVkResult(vkr);
}
// remove the replacements
rm->ReplaceResource(origsrcid, GetResID(pipe));
}
}
for(VkPipeline pipe : deletequeue)
m_pDriver->vkDestroyPipeline(dev, pipe, NULL);
}
ResourceId VulkanReplay::CreateProxyTexture(const TextureDescription &templateTex)
{
VULKANNOTIMP("CreateProxyTexture");
return ResourceId();
}
void VulkanReplay::SetProxyTextureData(ResourceId texid, const Subresource &sub, byte *data,
size_t dataSize)
{
VULKANNOTIMP("SetProxyTextureData");
}
bool VulkanReplay::IsTextureSupported(const TextureDescription &tex)
{
return true;
}
bool VulkanReplay::NeedRemapForFetch(const ResourceFormat &format)
{
return false;
}
ResourceId VulkanReplay::CreateProxyBuffer(const BufferDescription &templateBuf)
{
VULKANNOTIMP("CreateProxyBuffer");
return ResourceId();
}
void VulkanReplay::SetProxyBufferData(ResourceId bufid, byte *data, size_t dataSize)
{
VULKANNOTIMP("SetProxyTextureData");
}
RDResult Vulkan_CreateReplayDevice(RDCFile *rdc, const ReplayOptions &opts, IReplayDriver **driver)
{
RDCDEBUG("Creating a VulkanReplay replay device");
// disable the layer env var, just in case the user left it set from a previous capture run
Process::RegisterEnvironmentModification(
EnvironmentModification(EnvMod::Set, EnvSep::NoSep, RENDERDOC_VULKAN_LAYER_VAR, "0"));
// disable buggy and user-hostile NV optimus layer, which can completely delete physical devices
// (not just rearrange them) and cause problems between capture and replay.
Process::RegisterEnvironmentModification(
EnvironmentModification(EnvMod::Set, EnvSep::NoSep, "DISABLE_LAYER_NV_OPTIMUS_1", ""));
// RTSS layer is buggy, disable it to avoid bug reports that are caused by it
Process::RegisterEnvironmentModification(
EnvironmentModification(EnvMod::Set, EnvSep::NoSep, "DISABLE_RTSS_LAYER", "1"));
// OBS's layer causes crashes, disable it too.
Process::RegisterEnvironmentModification(
EnvironmentModification(EnvMod::Set, EnvSep::NoSep, "DISABLE_VULKAN_OBS_CAPTURE", "1"));
// OverWolf is some shitty software that forked OBS and changed the layer value
Process::RegisterEnvironmentModification(
EnvironmentModification(EnvMod::Set, EnvSep::NoSep, "DISABLE_VULKAN_OW_OBS_CAPTURE", "1"));
// buggy program AgaueEye which also doesn't have a proper layer configuration. As a result
// this is likely to have side-effects but probably also on other buggy layers that duplicate
// sample code without even changing the layer json
Process::RegisterEnvironmentModification(
EnvironmentModification(EnvMod::Set, EnvSep::NoSep, "DISABLE_SAMPLE_LAYER", "1"));
// buggy overlay gamepp
Process::RegisterEnvironmentModification(
EnvironmentModification(EnvMod::Set, EnvSep::NoSep, "DISABLE_GAMEPP_LAYER", "1"));
// mesa device select layer crashes when it calls GPDP2 inside vkCreateInstance, which fails on
// the current loader.
Process::RegisterEnvironmentModification(
EnvironmentModification(EnvMod::Set, EnvSep::NoSep, "NODEVICE_SELECT", "1"));
Process::RegisterEnvironmentModification(EnvironmentModification(
EnvMod::Set, EnvSep::NoSep, "DISABLE_LAYER_AMD_SWITCHABLE_GRAPHICS_1", "1"));
Process::RegisterEnvironmentModification(
EnvironmentModification(EnvMod::Set, EnvSep::NoSep, "VK_LAYER_bandicam_helper_DEBUG_1", "1"));
// fpsmon not only has a buggy layer but it also picks an absurdly generic disable environment
// variable :(. Hopefully no other program picks this, or if it does then it's probably not a
// bad thing to disable too
Process::RegisterEnvironmentModification(
EnvironmentModification(EnvMod::Set, EnvSep::NoSep, "DISABLE_LAYER", "1"));
Process::ApplyEnvironmentModification();
void *module = LoadVulkanLibrary();
if(module == NULL)
{
RETURN_ERROR_RESULT(ResultCode::APIInitFailed, "Failed to load vulkan library");
}
VkInitParams initParams;
uint64_t ver = VkInitParams::CurrentVersion;
// if we have an RDCFile, open the frame capture section and serialise the init params.
// if not, we're creating a proxy-capable device so use default-initialised init params.
if(rdc)
{
int sectionIdx = rdc->SectionIndex(SectionType::FrameCapture);
if(sectionIdx < 0)
RETURN_ERROR_RESULT(ResultCode::FileCorrupted, "File does not contain captured API data");
ver = rdc->GetSectionProperties(sectionIdx).version;
if(!VkInitParams::IsSupportedVersion(ver))
{
RETURN_ERROR_RESULT(ResultCode::APIIncompatibleVersion,
"Vulkan capture is incompatible version %llu, newest supported by this "
"build of RenderDoc is %llu",
ver, VkInitParams::CurrentVersion);
}
StreamReader *reader = rdc->ReadSection(sectionIdx);
ReadSerialiser ser(reader, Ownership::Stream);
ser.SetVersion(ver);
SystemChunk chunk = ser.ReadChunk<SystemChunk>();
if(chunk != SystemChunk::DriverInit)
{
RETURN_ERROR_RESULT(ResultCode::FileCorrupted,
"Expected to get a DriverInit chunk, instead got %u", chunk);
}
SERIALISE_ELEMENT(initParams);
if(ser.IsErrored())
{
return ser.GetError();
}
}
InitReplayTables(module);
const bool isProxy = (rdc == NULL);
AMDRGPControl *rgp = NULL;
if(!isProxy)
{
rgp = new AMDRGPControl();
if(!rgp->Initialised())
SAFE_DELETE(rgp);
}
WrappedVulkan *vk = new WrappedVulkan();
VulkanReplay *replay = vk->GetReplay();
replay->SetProxy(isProxy);
RDResult status = vk->Initialise(initParams, ver, opts);
if(status != ResultCode::Succeeded)
{
SAFE_DELETE(rgp);
delete vk;
return status;
}
RDCLOG("Created device.");
replay->SetRGP(rgp);
*driver = (IReplayDriver *)replay;
replay->GetInitialDriverVersion();
return ResultCode::Succeeded;
}
struct VulkanDriverRegistration
{
VulkanDriverRegistration()
{
RenderDoc::Inst().RegisterReplayProvider(RDCDriver::Vulkan, &Vulkan_CreateReplayDevice);
RenderDoc::Inst().SetVulkanLayerCheck(&VulkanReplay::CheckVulkanLayer);
RenderDoc::Inst().SetVulkanLayerInstall(&VulkanReplay::InstallVulkanLayer);
}
};
static VulkanDriverRegistration VkDriverRegistration;
RDResult Vulkan_ProcessStructured(RDCFile *rdc, SDFile &output)
{
WrappedVulkan vulkan;
int sectionIdx = rdc->SectionIndex(SectionType::FrameCapture);
if(sectionIdx < 0)
RETURN_ERROR_RESULT(ResultCode::FileCorrupted, "File does not contain captured API data");
vulkan.SetStructuredExport(rdc->GetSectionProperties(sectionIdx).version);
RDResult status = vulkan.ReadLogInitialisation(rdc, true);
if(status == ResultCode::Succeeded)
vulkan.GetStructuredFile()->Swap(output);
return status;
}
static StructuredProcessRegistration VulkanProcessRegistration(RDCDriver::Vulkan,
&Vulkan_ProcessStructured);
Reference in New Issue View Git Blame Copy Permalink