blob: 940f49e2d5129616a30644c1d349ddad7cd7bd53 [file] [log] [blame]
/* Copyright (c) 2015-2016 The Khronos Group Inc.
* Copyright (c) 2015-2016 Valve Corporation
* Copyright (c) 2015-2016 LunarG, Inc.
* Copyright (C) 2015-2016 Google Inc.
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and/or associated documentation files (the "Materials"), to
* deal in the Materials without restriction, including without limitation the
* rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
* sell copies of the Materials, and to permit persons to whom the Materials
* are furnished to do so, subject to the following conditions:
*
* The above copyright notice(s) and this permission notice shall be included
* in all copies or substantial portions of the Materials.
*
* THE MATERIALS ARE 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 MATERIALS OR THE
* USE OR OTHER DEALINGS IN THE MATERIALS
*
* Author: Cody Northrop <cnorthrop@google.com>
* Author: Michael Lentine <mlentine@google.com>
* Author: Tobin Ehlis <tobine@google.com>
* Author: Chia-I Wu <olv@google.com>
* Author: Chris Forbes <chrisf@ijw.co.nz>
* Author: Mark Lobodzinski <mark@lunarg.com>
* Author: Ian Elliott <ianelliott@google.com>
*/
// Allow use of STL min and max functions in Windows
#define NOMINMAX
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <assert.h>
#include <unordered_map>
#include <unordered_set>
#include <map>
#include <string>
#include <iostream>
#include <algorithm>
#include <list>
#include <spirv.hpp>
#include <set>
#include "vk_loader_platform.h"
#include "vk_dispatch_table_helper.h"
#include "vk_struct_string_helper_cpp.h"
#if defined(__GNUC__)
#pragma GCC diagnostic ignored "-Wwrite-strings"
#endif
#if defined(__GNUC__)
#pragma GCC diagnostic warning "-Wwrite-strings"
#endif
#include "vk_struct_size_helper.h"
#include "draw_state.h"
#include "vk_layer_config.h"
#include "vulkan/vk_debug_marker_layer.h"
#include "vk_layer_table.h"
#include "vk_layer_debug_marker_table.h"
#include "vk_layer_data.h"
#include "vk_layer_logging.h"
#include "vk_layer_extension_utils.h"
#include "vk_layer_utils.h"
using std::unordered_map;
using std::unordered_set;
// Track command pools and their command buffers
struct CMD_POOL_INFO {
VkCommandPoolCreateFlags createFlags;
uint32_t queueFamilyIndex;
list<VkCommandBuffer> commandBuffers; // list container of cmd buffers allocated from this pool
};
struct devExts {
VkBool32 debug_marker_enabled;
VkBool32 wsi_enabled;
unordered_map<VkSwapchainKHR, SWAPCHAIN_NODE*> swapchainMap;
};
// fwd decls
struct shader_module;
struct render_pass;
struct layer_data {
debug_report_data* report_data;
std::vector<VkDebugReportCallbackEXT> logging_callback;
VkLayerDispatchTable* device_dispatch_table;
VkLayerInstanceDispatchTable* instance_dispatch_table;
devExts device_extensions;
vector<VkQueue> queues; // all queues under given device
// Global set of all cmdBuffers that are inFlight on this device
unordered_set<VkCommandBuffer> globalInFlightCmdBuffers;
// Layer specific data
unordered_map<VkSampler, unique_ptr<SAMPLER_NODE>> sampleMap;
unordered_map<VkImageView, unique_ptr<VkImageViewCreateInfo>> imageViewMap;
unordered_map<VkImage, unique_ptr<VkImageCreateInfo>> imageMap;
unordered_map<VkBufferView, unique_ptr<VkBufferViewCreateInfo>> bufferViewMap;
unordered_map<VkBuffer, BUFFER_NODE> bufferMap;
unordered_map<VkPipeline, PIPELINE_NODE*> pipelineMap;
unordered_map<VkCommandPool, CMD_POOL_INFO> commandPoolMap;
unordered_map<VkDescriptorPool, DESCRIPTOR_POOL_NODE*> descriptorPoolMap;
unordered_map<VkDescriptorSet, SET_NODE*> setMap;
unordered_map<VkDescriptorSetLayout, LAYOUT_NODE*> descriptorSetLayoutMap;
unordered_map<VkPipelineLayout, PIPELINE_LAYOUT_NODE> pipelineLayoutMap;
unordered_map<VkDeviceMemory, VkImage> memImageMap;
unordered_map<VkFence, FENCE_NODE> fenceMap;
unordered_map<VkQueue, QUEUE_NODE> queueMap;
unordered_map<VkEvent, EVENT_NODE> eventMap;
unordered_map<QueryObject, bool> queryToStateMap;
unordered_map<VkQueryPool, QUERY_POOL_NODE> queryPoolMap;
unordered_map<VkSemaphore, SEMAPHORE_NODE> semaphoreMap;
unordered_map<void*, GLOBAL_CB_NODE*> commandBufferMap;
unordered_map<VkFramebuffer, VkFramebufferCreateInfo*> frameBufferMap;
unordered_map<VkImage, vector<ImageSubresourcePair>> imageSubresourceMap;
unordered_map<ImageSubresourcePair, IMAGE_NODE> imageLayoutMap;
unordered_map<VkRenderPass, RENDER_PASS_NODE*> renderPassMap;
unordered_map<VkShaderModule, shader_module*> shaderModuleMap;
// Current render pass
VkRenderPassBeginInfo renderPassBeginInfo;
uint32_t currentSubpass;
// Device specific data
PHYS_DEV_PROPERTIES_NODE physDevProperties;
layer_data() :
report_data(nullptr),
device_dispatch_table(nullptr),
instance_dispatch_table(nullptr),
device_extensions()
{};
};
// Code imported from ShaderChecker
static void
build_def_index(shader_module *);
// A forward iterator over spirv instructions. Provides easy access to len, opcode, and content words
// without the caller needing to care too much about the physical SPIRV module layout.
struct spirv_inst_iter {
std::vector<uint32_t>::const_iterator zero;
std::vector<uint32_t>::const_iterator it;
uint32_t len() { return *it >> 16; }
uint32_t opcode() { return *it & 0x0ffffu; }
uint32_t const & word(unsigned n) { return it[n]; }
uint32_t offset() { return (uint32_t)(it - zero); }
spirv_inst_iter() {}
spirv_inst_iter(std::vector<uint32_t>::const_iterator zero,
std::vector<uint32_t>::const_iterator it) : zero(zero), it(it) {}
bool operator== (spirv_inst_iter const & other) {
return it == other.it;
}
bool operator!= (spirv_inst_iter const & other) {
return it != other.it;
}
spirv_inst_iter operator++ (int) { /* x++ */
spirv_inst_iter ii = *this;
it += len();
return ii;
}
spirv_inst_iter operator++ () { /* ++x; */
it += len();
return *this;
}
/* The iterator and the value are the same thing. */
spirv_inst_iter & operator* () { return *this; }
spirv_inst_iter const & operator* () const { return *this; }
};
struct shader_module {
/* the spirv image itself */
vector<uint32_t> words;
/* a mapping of <id> to the first word of its def. this is useful because walking type
* trees, constant expressions, etc requires jumping all over the instruction stream.
*/
unordered_map<unsigned, unsigned> def_index;
shader_module(VkShaderModuleCreateInfo const *pCreateInfo) :
words((uint32_t *)pCreateInfo->pCode, (uint32_t *)pCreateInfo->pCode + pCreateInfo->codeSize / sizeof(uint32_t)),
def_index() {
build_def_index(this);
}
/* expose begin() / end() to enable range-based for */
spirv_inst_iter begin() const { return spirv_inst_iter(words.begin(), words.begin() + 5); } /* first insn */
spirv_inst_iter end() const { return spirv_inst_iter(words.begin(), words.end()); } /* just past last insn */
/* given an offset into the module, produce an iterator there. */
spirv_inst_iter at(unsigned offset) const { return spirv_inst_iter(words.begin(), words.begin() + offset); }
/* gets an iterator to the definition of an id */
spirv_inst_iter get_def(unsigned id) const {
auto it = def_index.find(id);
if (it == def_index.end()) {
return end();
}
return at(it->second);
}
};
// TODO : Do we need to guard access to layer_data_map w/ lock?
static unordered_map<void*, layer_data*> layer_data_map;
// TODO : This can be much smarter, using separate locks for separate global data
static int globalLockInitialized = 0;
static loader_platform_thread_mutex globalLock;
#define MAX_TID 513
static loader_platform_thread_id g_tidMapping[MAX_TID] = {0};
static uint32_t g_maxTID = 0;
template layer_data *get_my_data_ptr<layer_data>(
void *data_key,
std::unordered_map<void *, layer_data *> &data_map);
// Map actual TID to an index value and return that index
// This keeps TIDs in range from 0-MAX_TID and simplifies compares between runs
static uint32_t getTIDIndex() {
loader_platform_thread_id tid = loader_platform_get_thread_id();
for (uint32_t i = 0; i < g_maxTID; i++) {
if (tid == g_tidMapping[i])
return i;
}
// Don't yet have mapping, set it and return newly set index
uint32_t retVal = (uint32_t) g_maxTID;
g_tidMapping[g_maxTID++] = tid;
assert(g_maxTID < MAX_TID);
return retVal;
}
// Return a string representation of CMD_TYPE enum
static string cmdTypeToString(CMD_TYPE cmd)
{
switch (cmd)
{
case CMD_BINDPIPELINE:
return "CMD_BINDPIPELINE";
case CMD_BINDPIPELINEDELTA:
return "CMD_BINDPIPELINEDELTA";
case CMD_SETVIEWPORTSTATE:
return "CMD_SETVIEWPORTSTATE";
case CMD_SETLINEWIDTHSTATE:
return "CMD_SETLINEWIDTHSTATE";
case CMD_SETDEPTHBIASSTATE:
return "CMD_SETDEPTHBIASSTATE";
case CMD_SETBLENDSTATE:
return "CMD_SETBLENDSTATE";
case CMD_SETDEPTHBOUNDSSTATE:
return "CMD_SETDEPTHBOUNDSSTATE";
case CMD_SETSTENCILREADMASKSTATE:
return "CMD_SETSTENCILREADMASKSTATE";
case CMD_SETSTENCILWRITEMASKSTATE:
return "CMD_SETSTENCILWRITEMASKSTATE";
case CMD_SETSTENCILREFERENCESTATE:
return "CMD_SETSTENCILREFERENCESTATE";
case CMD_BINDDESCRIPTORSETS:
return "CMD_BINDDESCRIPTORSETS";
case CMD_BINDINDEXBUFFER:
return "CMD_BINDINDEXBUFFER";
case CMD_BINDVERTEXBUFFER:
return "CMD_BINDVERTEXBUFFER";
case CMD_DRAW:
return "CMD_DRAW";
case CMD_DRAWINDEXED:
return "CMD_DRAWINDEXED";
case CMD_DRAWINDIRECT:
return "CMD_DRAWINDIRECT";
case CMD_DRAWINDEXEDINDIRECT:
return "CMD_DRAWINDEXEDINDIRECT";
case CMD_DISPATCH:
return "CMD_DISPATCH";
case CMD_DISPATCHINDIRECT:
return "CMD_DISPATCHINDIRECT";
case CMD_COPYBUFFER:
return "CMD_COPYBUFFER";
case CMD_COPYIMAGE:
return "CMD_COPYIMAGE";
case CMD_BLITIMAGE:
return "CMD_BLITIMAGE";
case CMD_COPYBUFFERTOIMAGE:
return "CMD_COPYBUFFERTOIMAGE";
case CMD_COPYIMAGETOBUFFER:
return "CMD_COPYIMAGETOBUFFER";
case CMD_CLONEIMAGEDATA:
return "CMD_CLONEIMAGEDATA";
case CMD_UPDATEBUFFER:
return "CMD_UPDATEBUFFER";
case CMD_FILLBUFFER:
return "CMD_FILLBUFFER";
case CMD_CLEARCOLORIMAGE:
return "CMD_CLEARCOLORIMAGE";
case CMD_CLEARATTACHMENTS:
return "CMD_CLEARCOLORATTACHMENT";
case CMD_CLEARDEPTHSTENCILIMAGE:
return "CMD_CLEARDEPTHSTENCILIMAGE";
case CMD_RESOLVEIMAGE:
return "CMD_RESOLVEIMAGE";
case CMD_SETEVENT:
return "CMD_SETEVENT";
case CMD_RESETEVENT:
return "CMD_RESETEVENT";
case CMD_WAITEVENTS:
return "CMD_WAITEVENTS";
case CMD_PIPELINEBARRIER:
return "CMD_PIPELINEBARRIER";
case CMD_BEGINQUERY:
return "CMD_BEGINQUERY";
case CMD_ENDQUERY:
return "CMD_ENDQUERY";
case CMD_RESETQUERYPOOL:
return "CMD_RESETQUERYPOOL";
case CMD_COPYQUERYPOOLRESULTS:
return "CMD_COPYQUERYPOOLRESULTS";
case CMD_WRITETIMESTAMP:
return "CMD_WRITETIMESTAMP";
case CMD_INITATOMICCOUNTERS:
return "CMD_INITATOMICCOUNTERS";
case CMD_LOADATOMICCOUNTERS:
return "CMD_LOADATOMICCOUNTERS";
case CMD_SAVEATOMICCOUNTERS:
return "CMD_SAVEATOMICCOUNTERS";
case CMD_BEGINRENDERPASS:
return "CMD_BEGINRENDERPASS";
case CMD_ENDRENDERPASS:
return "CMD_ENDRENDERPASS";
case CMD_DBGMARKERBEGIN:
return "CMD_DBGMARKERBEGIN";
case CMD_DBGMARKEREND:
return "CMD_DBGMARKEREND";
default:
return "UNKNOWN";
}
}
// SPIRV utility functions
static void
build_def_index(shader_module *module)
{
for (auto insn : *module) {
switch (insn.opcode()) {
/* Types */
case spv::OpTypeVoid:
case spv::OpTypeBool:
case spv::OpTypeInt:
case spv::OpTypeFloat:
case spv::OpTypeVector:
case spv::OpTypeMatrix:
case spv::OpTypeImage:
case spv::OpTypeSampler:
case spv::OpTypeSampledImage:
case spv::OpTypeArray:
case spv::OpTypeRuntimeArray:
case spv::OpTypeStruct:
case spv::OpTypeOpaque:
case spv::OpTypePointer:
case spv::OpTypeFunction:
case spv::OpTypeEvent:
case spv::OpTypeDeviceEvent:
case spv::OpTypeReserveId:
case spv::OpTypeQueue:
case spv::OpTypePipe:
module->def_index[insn.word(1)] = insn.offset();
break;
/* Fixed constants */
case spv::OpConstantTrue:
case spv::OpConstantFalse:
case spv::OpConstant:
case spv::OpConstantComposite:
case spv::OpConstantSampler:
case spv::OpConstantNull:
module->def_index[insn.word(2)] = insn.offset();
break;
/* Specialization constants */
case spv::OpSpecConstantTrue:
case spv::OpSpecConstantFalse:
case spv::OpSpecConstant:
case spv::OpSpecConstantComposite:
case spv::OpSpecConstantOp:
module->def_index[insn.word(2)] = insn.offset();
break;
/* Variables */
case spv::OpVariable:
module->def_index[insn.word(2)] = insn.offset();
break;
/* Functions */
case spv::OpFunction:
module->def_index[insn.word(2)] = insn.offset();
break;
default:
/* We don't care about any other defs for now. */
break;
}
}
}
static spirv_inst_iter
find_entrypoint(shader_module *src, char const *name, VkShaderStageFlagBits stageBits)
{
for (auto insn : *src) {
if (insn.opcode() == spv::OpEntryPoint) {
auto entrypointName = (char const *) &insn.word(3);
auto entrypointStageBits = 1u << insn.word(1);
if (!strcmp(entrypointName, name) && (entrypointStageBits & stageBits)) {
return insn;
}
}
}
return src->end();
}
bool
shader_is_spirv(VkShaderModuleCreateInfo const *pCreateInfo)
{
uint32_t *words = (uint32_t *)pCreateInfo->pCode;
size_t sizeInWords = pCreateInfo->codeSize / sizeof(uint32_t);
/* Just validate that the header makes sense. */
return sizeInWords >= 5 && words[0] == spv::MagicNumber && words[1] == spv::Version;
}
static char const *
storage_class_name(unsigned sc)
{
switch (sc) {
case spv::StorageClassInput: return "input";
case spv::StorageClassOutput: return "output";
case spv::StorageClassUniformConstant: return "const uniform";
case spv::StorageClassUniform: return "uniform";
case spv::StorageClassWorkgroup: return "workgroup local";
case spv::StorageClassCrossWorkgroup: return "workgroup global";
case spv::StorageClassPrivate: return "private global";
case spv::StorageClassFunction: return "function";
case spv::StorageClassGeneric: return "generic";
case spv::StorageClassAtomicCounter: return "atomic counter";
case spv::StorageClassImage: return "image";
default: return "unknown";
}
}
/* get the value of an integral constant */
unsigned
get_constant_value(shader_module const *src, unsigned id)
{
auto value = src->get_def(id);
assert(value != src->end());
if (value.opcode() != spv::OpConstant) {
/* TODO: Either ensure that the specialization transform is already performed on a module we're
considering here, OR -- specialize on the fly now.
*/
return 1;
}
return value.word(3);
}
/* returns ptr to null terminator */
static char *
describe_type(char *dst, shader_module const *src, unsigned type)
{
auto insn = src->get_def(type);
assert(insn != src->end());
switch (insn.opcode()) {
case spv::OpTypeBool:
return dst + sprintf(dst, "bool");
case spv::OpTypeInt:
return dst + sprintf(dst, "%cint%d", insn.word(3) ? 's' : 'u', insn.word(2));
case spv::OpTypeFloat:
return dst + sprintf(dst, "float%d", insn.word(2));
case spv::OpTypeVector:
dst += sprintf(dst, "vec%d of ", insn.word(3));
return describe_type(dst, src, insn.word(2));
case spv::OpTypeMatrix:
dst += sprintf(dst, "mat%d of ", insn.word(3));
return describe_type(dst, src, insn.word(2));
case spv::OpTypeArray:
dst += sprintf(dst, "arr[%d] of ", get_constant_value(src, insn.word(3)));
return describe_type(dst, src, insn.word(2));
case spv::OpTypePointer:
dst += sprintf(dst, "ptr to %s ", storage_class_name(insn.word(2)));
return describe_type(dst, src, insn.word(3));
case spv::OpTypeStruct:
{
dst += sprintf(dst, "struct of (");
for (unsigned i = 2; i < insn.len(); i++) {
dst = describe_type(dst, src, insn.word(i));
dst += sprintf(dst, i == insn.len()-1 ? ")" : ", ");
}
return dst;
}
case spv::OpTypeSampler:
return dst + sprintf(dst, "sampler");
default:
return dst + sprintf(dst, "oddtype");
}
}
static bool
types_match(shader_module const *a, shader_module const *b, unsigned a_type, unsigned b_type, bool b_arrayed)
{
/* walk two type trees together, and complain about differences */
auto a_insn = a->get_def(a_type);
auto b_insn = b->get_def(b_type);
assert(a_insn != a->end());
assert(b_insn != b->end());
if (b_arrayed && b_insn.opcode() == spv::OpTypeArray) {
/* we probably just found the extra level of arrayness in b_type: compare the type inside it to a_type */
return types_match(a, b, a_type, b_insn.word(2), false);
}
if (a_insn.opcode() != b_insn.opcode()) {
return false;
}
switch (a_insn.opcode()) {
/* if b_arrayed and we hit a leaf type, then we can't match -- there's nowhere for the extra OpTypeArray to be! */
case spv::OpTypeBool:
return true && !b_arrayed;
case spv::OpTypeInt:
/* match on width, signedness */
return a_insn.word(2) == b_insn.word(2) && a_insn.word(3) == b_insn.word(3) && !b_arrayed;
case spv::OpTypeFloat:
/* match on width */
return a_insn.word(2) == b_insn.word(2) && !b_arrayed;
case spv::OpTypeVector:
case spv::OpTypeMatrix:
/* match on element type, count. these all have the same layout. we don't get here if
* b_arrayed -- that is handled above. */
return !b_arrayed &&
types_match(a, b, a_insn.word(2), b_insn.word(2), b_arrayed) &&
a_insn.word(3) == b_insn.word(3);
case spv::OpTypeArray:
/* match on element type, count. these all have the same layout. we don't get here if
* b_arrayed. This differs from vector & matrix types in that the array size is the id of a constant instruction,
* not a literal within OpTypeArray */
return !b_arrayed &&
types_match(a, b, a_insn.word(2), b_insn.word(2), b_arrayed) &&
get_constant_value(a, a_insn.word(3)) == get_constant_value(b, b_insn.word(3));
case spv::OpTypeStruct:
/* match on all element types */
{
if (b_arrayed) {
/* for the purposes of matching different levels of arrayness, structs are leaves. */
return false;
}
if (a_insn.len() != b_insn.len()) {
return false; /* structs cannot match if member counts differ */
}
for (unsigned i = 2; i < a_insn.len(); i++) {
if (!types_match(a, b, a_insn.word(i), b_insn.word(i), b_arrayed)) {
return false;
}
}
return true;
}
case spv::OpTypePointer:
/* match on pointee type. storage class is expected to differ */
return types_match(a, b, a_insn.word(3), b_insn.word(3), b_arrayed);
default:
/* remaining types are CLisms, or may not appear in the interfaces we
* are interested in. Just claim no match.
*/
return false;
}
}
static int
value_or_default(std::unordered_map<unsigned, unsigned> const &map, unsigned id, int def)
{
auto it = map.find(id);
if (it == map.end())
return def;
else
return it->second;
}
static unsigned
get_locations_consumed_by_type(shader_module const *src, unsigned type, bool strip_array_level)
{
auto insn = src->get_def(type);
assert(insn != src->end());
switch (insn.opcode()) {
case spv::OpTypePointer:
/* see through the ptr -- this is only ever at the toplevel for graphics shaders;
* we're never actually passing pointers around. */
return get_locations_consumed_by_type(src, insn.word(3), strip_array_level);
case spv::OpTypeArray:
if (strip_array_level) {
return get_locations_consumed_by_type(src, insn.word(2), false);
}
else {
return get_constant_value(src, insn.word(3)) * get_locations_consumed_by_type(src, insn.word(2), false);
}
case spv::OpTypeMatrix:
/* num locations is the dimension * element size */
return insn.word(3) * get_locations_consumed_by_type(src, insn.word(2), false);
default:
/* everything else is just 1. */
return 1;
/* TODO: extend to handle 64bit scalar types, whose vectors may need
* multiple locations. */
}
}
struct interface_var {
uint32_t id;
uint32_t type_id;
uint32_t offset;
/* TODO: collect the name, too? Isn't required to be present. */
};
static void
collect_interface_block_members(layer_data *my_data, VkDevice dev,
shader_module const *src,
std::map<uint32_t, interface_var> &out,
std::map<uint32_t, interface_var> &builtins_out,
std::unordered_map<unsigned, unsigned> const &blocks,
bool is_array_of_verts,
uint32_t id,
uint32_t type_id)
{
/* Walk down the type_id presented, trying to determine whether it's actually an interface block. */
auto type = src->get_def(type_id);
while (true) {
if (type.opcode() == spv::OpTypePointer) {
type = src->get_def(type.word(3));
}
else if (type.opcode() == spv::OpTypeArray && is_array_of_verts) {
type = src->get_def(type.word(2));
is_array_of_verts = false;
}
else if (type.opcode() == spv::OpTypeStruct) {
if (blocks.find(type.word(1)) == blocks.end()) {
/* This isn't an interface block. */
return;
}
else {
/* We have found the correct type. Walk its members. */
break;
}
}
else {
/* not an interface block */
return;
}
}
/* Walk all the OpMemberDecorate for type's result id. */
for (auto insn : *src) {
if (insn.opcode() == spv::OpMemberDecorate && insn.word(1) == type.word(1)) {
unsigned member_index = insn.word(2);
unsigned member_type_id = type.word(2 + member_index);
if (insn.word(3) == spv::DecorationLocation) {
unsigned location = insn.word(4);
unsigned num_locations = get_locations_consumed_by_type(src, member_type_id, false);
for (unsigned int offset = 0; offset < num_locations; offset++) {
interface_var v;
v.id = id;
/* TODO: member index in interface_var too? */
v.type_id = member_type_id;
v.offset = offset;
out[location + offset] = v;
}
}
else if (insn.word(3) == spv::DecorationBuiltIn) {
unsigned builtin = insn.word(4);
interface_var v;
v.id = id;
v.type_id = member_type_id;
v.offset = 0;
builtins_out[builtin] = v;
}
}
}
}
static void
collect_interface_by_location(layer_data *my_data, VkDevice dev,
shader_module const *src,
spirv_inst_iter entrypoint,
spv::StorageClass sinterface,
std::map<uint32_t, interface_var> &out,
std::map<uint32_t, interface_var> &builtins_out,
bool is_array_of_verts)
{
std::unordered_map<unsigned, unsigned> var_locations;
std::unordered_map<unsigned, unsigned> var_builtins;
std::unordered_map<unsigned, unsigned> blocks;
for (auto insn : *src) {
/* We consider two interface models: SSO rendezvous-by-location, and
* builtins. Complain about anything that fits neither model.
*/
if (insn.opcode() == spv::OpDecorate) {
if (insn.word(2) == spv::DecorationLocation) {
var_locations[insn.word(1)] = insn.word(3);
}
if (insn.word(2) == spv::DecorationBuiltIn) {
var_builtins[insn.word(1)] = insn.word(3);
}
if (insn.word(2) == spv::DecorationBlock) {
blocks[insn.word(1)] = 1;
}
}
}
/* TODO: handle grouped decorations */
/* TODO: handle index=1 dual source outputs from FS -- two vars will
* have the same location, and we DONT want to clobber. */
/* find the end of the entrypoint's name string. additional zero bytes follow the actual null
terminator, to fill out the rest of the word - so we only need to look at the last byte in
the word to determine which word contains the terminator. */
auto word = 3;
while (entrypoint.word(word) & 0xff000000u) {
++word;
}
++word;
for (; word < entrypoint.len(); word++) {
auto insn = src->get_def(entrypoint.word(word));
assert(insn != src->end());
assert(insn.opcode() == spv::OpVariable);
if (insn.word(3) == sinterface) {
unsigned id = insn.word(2);
unsigned type = insn.word(1);
int location = value_or_default(var_locations, id, -1);
int builtin = value_or_default(var_builtins, id, -1);
/* All variables and interface block members in the Input or Output storage classes
* must be decorated with either a builtin or an explicit location.
*
* TODO: integrate the interface block support here. For now, don't complain --
* a valid SPIRV module will only hit this path for the interface block case, as the
* individual members of the type are decorated, rather than variable declarations.
*/
if (location != -1) {
/* A user-defined interface variable, with a location. Where a variable
* occupied multiple locations, emit one result for each. */
unsigned num_locations = get_locations_consumed_by_type(src, type,
is_array_of_verts);
for (unsigned int offset = 0; offset < num_locations; offset++) {
interface_var v;
v.id = id;
v.type_id = type;
v.offset = offset;
out[location + offset] = v;
}
}
else if (builtin != -1) {
/* A builtin interface variable */
/* Note that since builtin interface variables do not consume numbered
* locations, there is no larger-than-vec4 consideration as above
*/
interface_var v;
v.id = id;
v.type_id = type;
v.offset = 0;
builtins_out[builtin] = v;
}
else {
/* An interface block instance */
collect_interface_block_members(my_data, dev, src, out, builtins_out,
blocks, is_array_of_verts, id, type);
}
}
}
}
static void
collect_interface_by_descriptor_slot(layer_data *my_data, VkDevice dev,
shader_module const *src, spv::StorageClass sinterface,
std::unordered_set<uint32_t> const &accessible_ids,
std::map<std::pair<unsigned, unsigned>, interface_var> &out)
{
std::unordered_map<unsigned, unsigned> var_sets;
std::unordered_map<unsigned, unsigned> var_bindings;
for (auto insn : *src) {
/* All variables in the Uniform or UniformConstant storage classes are required to be decorated with both
* DecorationDescriptorSet and DecorationBinding.
*/
if (insn.opcode() == spv::OpDecorate) {
if (insn.word(2) == spv::DecorationDescriptorSet) {
var_sets[insn.word(1)] = insn.word(3);
}
if (insn.word(2) == spv::DecorationBinding) {
var_bindings[insn.word(1)] = insn.word(3);
}
}
}
for (auto id : accessible_ids) {
auto insn = src->get_def(id);
assert(insn != src->end());
if (insn.opcode() == spv::OpVariable &&
(insn.word(3) == spv::StorageClassUniform ||
insn.word(3) == spv::StorageClassUniformConstant)) {
unsigned set = value_or_default(var_sets, insn.word(2), 0);
unsigned binding = value_or_default(var_bindings, insn.word(2), 0);
auto existing_it = out.find(std::make_pair(set, binding));
if (existing_it != out.end()) {
/* conflict within spv image */
log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DEVICE_EXT, /*dev*/0, __LINE__,
SHADER_CHECKER_INCONSISTENT_SPIRV, "SC",
"var %d (type %d) in %s interface in descriptor slot (%u,%u) conflicts with existing definition",
insn.word(2), insn.word(1), storage_class_name(sinterface),
existing_it->first.first, existing_it->first.second);
}
interface_var v;
v.id = insn.word(2);
v.type_id = insn.word(1);
out[std::make_pair(set, binding)] = v;
}
}
}
static bool
validate_interface_between_stages(layer_data *my_data, VkDevice dev,
shader_module const *producer, spirv_inst_iter producer_entrypoint, char const *producer_name,
shader_module const *consumer, spirv_inst_iter consumer_entrypoint, char const *consumer_name,
bool consumer_arrayed_input)
{
std::map<uint32_t, interface_var> outputs;
std::map<uint32_t, interface_var> inputs;
std::map<uint32_t, interface_var> builtin_outputs;
std::map<uint32_t, interface_var> builtin_inputs;
bool pass = true;
collect_interface_by_location(my_data, dev, producer, producer_entrypoint, spv::StorageClassOutput, outputs, builtin_outputs, false);
collect_interface_by_location(my_data, dev, consumer, consumer_entrypoint, spv::StorageClassInput, inputs, builtin_inputs,
consumer_arrayed_input);
auto a_it = outputs.begin();
auto b_it = inputs.begin();
/* maps sorted by key (location); walk them together to find mismatches */
while ((outputs.size() > 0 && a_it != outputs.end()) || ( inputs.size() && b_it != inputs.end())) {
bool a_at_end = outputs.size() == 0 || a_it == outputs.end();
bool b_at_end = inputs.size() == 0 || b_it == inputs.end();
auto a_first = a_at_end ? 0 : a_it->first;
auto b_first = b_at_end ? 0 : b_it->first;
if (b_at_end || ((!a_at_end) && (a_first < b_first))) {
if (log_msg(my_data->report_data, VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DEVICE_EXT, /*dev*/0, __LINE__, SHADER_CHECKER_OUTPUT_NOT_CONSUMED, "SC",
"%s writes to output location %d which is not consumed by %s", producer_name, a_first, consumer_name)) {
pass = false;
}
a_it++;
}
else if (a_at_end || a_first > b_first) {
if (log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DEVICE_EXT, /*dev*/0, __LINE__, SHADER_CHECKER_INPUT_NOT_PRODUCED, "SC",
"%s consumes input location %d which is not written by %s", consumer_name, b_first, producer_name)) {
pass = false;
}
b_it++;
}
else {
if (types_match(producer, consumer, a_it->second.type_id, b_it->second.type_id, consumer_arrayed_input)) {
/* OK! */
}
else {
char producer_type[1024];
char consumer_type[1024];
describe_type(producer_type, producer, a_it->second.type_id);
describe_type(consumer_type, consumer, b_it->second.type_id);
if (log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DEVICE_EXT, /*dev*/0, __LINE__, SHADER_CHECKER_INTERFACE_TYPE_MISMATCH, "SC",
"Type mismatch on location %d: '%s' vs '%s'", a_it->first, producer_type, consumer_type)) {
pass = false;
}
}
a_it++;
b_it++;
}
}
return pass;
}
enum FORMAT_TYPE {
FORMAT_TYPE_UNDEFINED,
FORMAT_TYPE_FLOAT, /* UNORM, SNORM, FLOAT, USCALED, SSCALED, SRGB -- anything we consider float in the shader */
FORMAT_TYPE_SINT,
FORMAT_TYPE_UINT,
};
static unsigned
get_format_type(VkFormat fmt) {
switch (fmt) {
case VK_FORMAT_UNDEFINED:
return FORMAT_TYPE_UNDEFINED;
case VK_FORMAT_R8_SINT:
case VK_FORMAT_R8G8_SINT:
case VK_FORMAT_R8G8B8_SINT:
case VK_FORMAT_R8G8B8A8_SINT:
case VK_FORMAT_R16_SINT:
case VK_FORMAT_R16G16_SINT:
case VK_FORMAT_R16G16B16_SINT:
case VK_FORMAT_R16G16B16A16_SINT:
case VK_FORMAT_R32_SINT:
case VK_FORMAT_R32G32_SINT:
case VK_FORMAT_R32G32B32_SINT:
case VK_FORMAT_R32G32B32A32_SINT:
case VK_FORMAT_B8G8R8_SINT:
case VK_FORMAT_B8G8R8A8_SINT:
case VK_FORMAT_A2B10G10R10_SINT_PACK32:
case VK_FORMAT_A2R10G10B10_SINT_PACK32:
return FORMAT_TYPE_SINT;
case VK_FORMAT_R8_UINT:
case VK_FORMAT_R8G8_UINT:
case VK_FORMAT_R8G8B8_UINT:
case VK_FORMAT_R8G8B8A8_UINT:
case VK_FORMAT_R16_UINT:
case VK_FORMAT_R16G16_UINT:
case VK_FORMAT_R16G16B16_UINT:
case VK_FORMAT_R16G16B16A16_UINT:
case VK_FORMAT_R32_UINT:
case VK_FORMAT_R32G32_UINT:
case VK_FORMAT_R32G32B32_UINT:
case VK_FORMAT_R32G32B32A32_UINT:
case VK_FORMAT_B8G8R8_UINT:
case VK_FORMAT_B8G8R8A8_UINT:
case VK_FORMAT_A2B10G10R10_UINT_PACK32:
case VK_FORMAT_A2R10G10B10_UINT_PACK32:
return FORMAT_TYPE_UINT;
default:
return FORMAT_TYPE_FLOAT;
}
}
/* characterizes a SPIR-V type appearing in an interface to a FF stage,
* for comparison to a VkFormat's characterization above. */
static unsigned
get_fundamental_type(shader_module const *src, unsigned type)
{
auto insn = src->get_def(type);
assert(insn != src->end());
switch (insn.opcode()) {
case spv::OpTypeInt:
return insn.word(3) ? FORMAT_TYPE_SINT : FORMAT_TYPE_UINT;
case spv::OpTypeFloat:
return FORMAT_TYPE_FLOAT;
case spv::OpTypeVector:
return get_fundamental_type(src, insn.word(2));
case spv::OpTypeMatrix:
return get_fundamental_type(src, insn.word(2));
case spv::OpTypeArray:
return get_fundamental_type(src, insn.word(2));
case spv::OpTypePointer:
return get_fundamental_type(src, insn.word(3));
default:
return FORMAT_TYPE_UNDEFINED;
}
}
static bool
validate_vi_consistency(layer_data *my_data, VkDevice dev, VkPipelineVertexInputStateCreateInfo const *vi)
{
/* walk the binding descriptions, which describe the step rate and stride of each vertex buffer.
* each binding should be specified only once.
*/
std::unordered_map<uint32_t, VkVertexInputBindingDescription const *> bindings;
bool pass = true;
for (unsigned i = 0; i < vi->vertexBindingDescriptionCount; i++) {
auto desc = &vi->pVertexBindingDescriptions[i];
auto & binding = bindings[desc->binding];
if (binding) {
if (log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DEVICE_EXT, /*dev*/0, __LINE__, SHADER_CHECKER_INCONSISTENT_VI, "SC",
"Duplicate vertex input binding descriptions for binding %d", desc->binding)) {
pass = false;
}
}
else {
binding = desc;
}
}
return pass;
}
static bool
validate_vi_against_vs_inputs(layer_data *my_data, VkDevice dev, VkPipelineVertexInputStateCreateInfo const *vi, shader_module const *vs, spirv_inst_iter entrypoint)
{
std::map<uint32_t, interface_var> inputs;
/* we collect builtin inputs, but they will never appear in the VI state --
* the vs builtin inputs are generated in the pipeline, not sourced from buffers (VertexID, etc)
*/
std::map<uint32_t, interface_var> builtin_inputs;
bool pass = true;
collect_interface_by_location(my_data, dev, vs, entrypoint, spv::StorageClassInput, inputs, builtin_inputs, false);
/* Build index by location */
std::map<uint32_t, VkVertexInputAttributeDescription const *> attribs;
if (vi) {
for (unsigned i = 0; i < vi->vertexAttributeDescriptionCount; i++)
attribs[vi->pVertexAttributeDescriptions[i].location] = &vi->pVertexAttributeDescriptions[i];
}
auto it_a = attribs.begin();
auto it_b = inputs.begin();
while ((attribs.size() > 0 && it_a != attribs.end()) || (inputs.size() > 0 && it_b != inputs.end())) {
bool a_at_end = attribs.size() == 0 || it_a == attribs.end();
bool b_at_end = inputs.size() == 0 || it_b == inputs.end();
auto a_first = a_at_end ? 0 : it_a->first;
auto b_first = b_at_end ? 0 : it_b->first;
if (!a_at_end && (b_at_end || a_first < b_first)) {
if (log_msg(my_data->report_data, VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DEVICE_EXT, /*dev*/0, __LINE__, SHADER_CHECKER_OUTPUT_NOT_CONSUMED, "SC",
"Vertex attribute at location %d not consumed by VS", a_first)) {
pass = false;
}
it_a++;
}
else if (!b_at_end && (a_at_end || b_first < a_first)) {
if (log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DEVICE_EXT, /*dev*/0, __LINE__, SHADER_CHECKER_INPUT_NOT_PRODUCED, "SC",
"VS consumes input at location %d but not provided", b_first)) {
pass = false;
}
it_b++;
}
else {
unsigned attrib_type = get_format_type(it_a->second->format);
unsigned input_type = get_fundamental_type(vs, it_b->second.type_id);
/* type checking */
if (attrib_type != FORMAT_TYPE_UNDEFINED && input_type != FORMAT_TYPE_UNDEFINED && attrib_type != input_type) {
char vs_type[1024];
describe_type(vs_type, vs, it_b->second.type_id);
if (log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DEVICE_EXT, /*dev*/0, __LINE__, SHADER_CHECKER_INTERFACE_TYPE_MISMATCH, "SC",
"Attribute type of `%s` at location %d does not match VS input type of `%s`",
string_VkFormat(it_a->second->format), a_first, vs_type)) {
pass = false;
}
}
/* OK! */
it_a++;
it_b++;
}
}
return pass;
}
static bool
validate_fs_outputs_against_render_pass(layer_data *my_data, VkDevice dev, shader_module const *fs, spirv_inst_iter entrypoint, RENDER_PASS_NODE const *rp, uint32_t subpass)
{
const std::vector<VkFormat> &color_formats = rp->subpassColorFormats[subpass];
std::map<uint32_t, interface_var> outputs;
std::map<uint32_t, interface_var> builtin_outputs;
bool pass = true;
/* TODO: dual source blend index (spv::DecIndex, zero if not provided) */
collect_interface_by_location(my_data, dev, fs, entrypoint, spv::StorageClassOutput, outputs, builtin_outputs, false);
auto it = outputs.begin();
uint32_t attachment = 0;
/* Walk attachment list and outputs together -- this is a little overpowered since attachments
* are currently dense, but the parallel with matching between shader stages is nice.
*/
while ((outputs.size() > 0 && it != outputs.end()) || attachment < color_formats.size()) {
if (attachment == color_formats.size() || ( it != outputs.end() && it->first < attachment)) {
if (log_msg(my_data->report_data, VK_DEBUG_REPORT_WARNING_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DEVICE_EXT, /*dev*/0, __LINE__, SHADER_CHECKER_OUTPUT_NOT_CONSUMED, "SC",
"FS writes to output location %d with no matching attachment", it->first)) {
pass = false;
}
it++;
}
else if (it == outputs.end() || it->first > attachment) {
if (log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DEVICE_EXT, /*dev*/0, __LINE__, SHADER_CHECKER_INPUT_NOT_PRODUCED, "SC",
"Attachment %d not written by FS", attachment)) {
pass = false;
}
attachment++;
}
else {
unsigned output_type = get_fundamental_type(fs, it->second.type_id);
unsigned att_type = get_format_type(color_formats[attachment]);
/* type checking */
if (att_type != FORMAT_TYPE_UNDEFINED && output_type != FORMAT_TYPE_UNDEFINED && att_type != output_type) {
char fs_type[1024];
describe_type(fs_type, fs, it->second.type_id);
if (log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DEVICE_EXT, /*dev*/0, __LINE__, SHADER_CHECKER_INTERFACE_TYPE_MISMATCH, "SC",
"Attachment %d of type `%s` does not match FS output type of `%s`",
attachment, string_VkFormat(color_formats[attachment]), fs_type)) {
pass = false;
}
}
/* OK! */
it++;
attachment++;
}
}
return pass;
}
/* For some analyses, we need to know about all ids referenced by the static call tree of a particular
* entrypoint. This is important for identifying the set of shader resources actually used by an entrypoint,
* for example.
* Note: we only explore parts of the image which might actually contain ids we care about for the above analyses.
* - NOT the shader input/output interfaces.
*
* TODO: The set of interesting opcodes here was determined by eyeballing the SPIRV spec. It might be worth
* converting parts of this to be generated from the machine-readable spec instead.
*/
static void
mark_accessible_ids(shader_module const *src, spirv_inst_iter entrypoint, std::unordered_set<uint32_t> &ids)
{
std::unordered_set<uint32_t> worklist;
worklist.insert(entrypoint.word(2));
while (!worklist.empty()) {
auto id_iter = worklist.begin();
auto id = *id_iter;
worklist.erase(id_iter);
auto insn = src->get_def(id);
if (insn == src->end()) {
/* id is something we didnt collect in build_def_index. that's OK -- we'll stumble
* across all kinds of things here that we may not care about. */
continue;
}
/* try to add to the output set */
if (!ids.insert(id).second) {
continue; /* if we already saw this id, we don't want to walk it again. */
}
switch (insn.opcode()) {
case spv::OpFunction:
/* scan whole body of the function, enlisting anything interesting */
while (++insn, insn.opcode() != spv::OpFunctionEnd) {
switch (insn.opcode()) {
case spv::OpLoad:
case spv::OpAtomicLoad:
case spv::OpAtomicExchange:
case spv::OpAtomicCompareExchange:
case spv::OpAtomicCompareExchangeWeak:
case spv::OpAtomicIIncrement:
case spv::OpAtomicIDecrement:
case spv::OpAtomicIAdd:
case spv::OpAtomicISub:
case spv::OpAtomicSMin:
case spv::OpAtomicUMin:
case spv::OpAtomicSMax:
case spv::OpAtomicUMax:
case spv::OpAtomicAnd:
case spv::OpAtomicOr:
case spv::OpAtomicXor:
worklist.insert(insn.word(3)); /* ptr */
break;
case spv::OpStore:
case spv::OpAtomicStore:
worklist.insert(insn.word(1)); /* ptr */
break;
case spv::OpAccessChain:
case spv::OpInBoundsAccessChain:
worklist.insert(insn.word(3)); /* base ptr */
break;
case spv::OpSampledImage:
case spv::OpImageSampleImplicitLod:
case spv::OpImageSampleExplicitLod:
case spv::OpImageSampleDrefImplicitLod:
case spv::OpImageSampleDrefExplicitLod:
case spv::OpImageSampleProjImplicitLod:
case spv::OpImageSampleProjExplicitLod:
case spv::OpImageSampleProjDrefImplicitLod:
case spv::OpImageSampleProjDrefExplicitLod:
case spv::OpImageFetch:
case spv::OpImageGather:
case spv::OpImageDrefGather:
case spv::OpImageRead:
case spv::OpImage:
case spv::OpImageQueryFormat:
case spv::OpImageQueryOrder:
case spv::OpImageQuerySizeLod:
case spv::OpImageQuerySize:
case spv::OpImageQueryLod:
case spv::OpImageQueryLevels:
case spv::OpImageQuerySamples:
case spv::OpImageSparseSampleImplicitLod:
case spv::OpImageSparseSampleExplicitLod:
case spv::OpImageSparseSampleDrefImplicitLod:
case spv::OpImageSparseSampleDrefExplicitLod:
case spv::OpImageSparseSampleProjImplicitLod:
case spv::OpImageSparseSampleProjExplicitLod:
case spv::OpImageSparseSampleProjDrefImplicitLod:
case spv::OpImageSparseSampleProjDrefExplicitLod:
case spv::OpImageSparseFetch:
case spv::OpImageSparseGather:
case spv::OpImageSparseDrefGather:
case spv::OpImageTexelPointer:
worklist.insert(insn.word(3)); /* image or sampled image */
break;
case spv::OpImageWrite:
worklist.insert(insn.word(1)); /* image -- different operand order to above */
break;
case spv::OpFunctionCall:
for (auto i = 3; i < insn.len(); i++) {
worklist.insert(insn.word(i)); /* fn itself, and all args */
}
break;
case spv::OpExtInst:
for (auto i = 5; i < insn.len(); i++) {
worklist.insert(insn.word(i)); /* operands to ext inst */
}
break;
}
}
break;
}
}
}
struct shader_stage_attributes {
char const * const name;
bool arrayed_input;
};
static shader_stage_attributes
shader_stage_attribs[] = {
{ "vertex shader", false },
{ "tessellation control shader", true },
{ "tessellation evaluation shader", false },
{ "geometry shader", true },
{ "fragment shader", false },
};
// For given pipelineLayout verify that the setLayout at slot.first
// has the requested binding at slot.second
static bool
has_descriptor_binding(layer_data* my_data,
vector<VkDescriptorSetLayout>* pipelineLayout,
std::pair<unsigned, unsigned> slot)
{
if (!pipelineLayout)
return false;
if (slot.first >= pipelineLayout->size())
return false;
const auto &bindingMap = my_data->descriptorSetLayoutMap[(*pipelineLayout)[slot.first]]
->bindingToIndexMap;
return (bindingMap.find(slot.second) != bindingMap.end());
}
static uint32_t get_shader_stage_id(VkShaderStageFlagBits stage)
{
uint32_t bit_pos = u_ffs(stage);
return bit_pos-1;
}
// Block of code at start here for managing/tracking Pipeline state that this layer cares about
static uint64_t g_drawCount[NUM_DRAW_TYPES] = {0, 0, 0, 0};
// TODO : Should be tracking lastBound per commandBuffer and when draws occur, report based on that cmd buffer lastBound
// Then need to synchronize the accesses based on cmd buffer so that if I'm reading state on one cmd buffer, updates
// to that same cmd buffer by separate thread are not changing state from underneath us
// Track the last cmd buffer touched by this thread
// prototype
static GLOBAL_CB_NODE* getCBNode(layer_data*, const VkCommandBuffer);
static VkBool32 hasDrawCmd(GLOBAL_CB_NODE* pCB)
{
for (uint32_t i=0; i<NUM_DRAW_TYPES; i++) {
if (pCB->drawCount[i])
return VK_TRUE;
}
return VK_FALSE;
}
// Check object status for selected flag state
static VkBool32 validate_status(layer_data* my_data, GLOBAL_CB_NODE* pNode, CBStatusFlags enable_mask, CBStatusFlags status_mask, CBStatusFlags status_flag, VkFlags msg_flags, DRAW_STATE_ERROR error_code, const char* fail_msg)
{
// If non-zero enable mask is present, check it against status but if enable_mask
// is 0 then no enable required so we should always just check status
if ((!enable_mask) || (enable_mask & pNode->status)) {
if ((pNode->status & status_mask) != status_flag) {
// TODO : How to pass dispatchable objects as srcObject? Here src obj should be cmd buffer
return log_msg(my_data->report_data, msg_flags, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, 0, __LINE__, error_code, "DS",
"CB object %#" PRIxLEAST64 ": %s", (uint64_t)(pNode->commandBuffer), fail_msg);
}
}
return VK_FALSE;
}
// Retrieve pipeline node ptr for given pipeline object
static PIPELINE_NODE* getPipeline(layer_data* my_data, const VkPipeline pipeline)
{
if (my_data->pipelineMap.find(pipeline) == my_data->pipelineMap.end()) {
return NULL;
}
return my_data->pipelineMap[pipeline];
}
// Return VK_TRUE if for a given PSO, the given state enum is dynamic, else return VK_FALSE
static VkBool32 isDynamic(const PIPELINE_NODE* pPipeline, const VkDynamicState state)
{
if (pPipeline && pPipeline->graphicsPipelineCI.pDynamicState) {
for (uint32_t i=0; i<pPipeline->graphicsPipelineCI.pDynamicState->dynamicStateCount; i++) {
if (state == pPipeline->graphicsPipelineCI.pDynamicState->pDynamicStates[i])
return VK_TRUE;
}
}
return VK_FALSE;
}
// Validate state stored as flags at time of draw call
static VkBool32 validate_draw_state_flags(layer_data* my_data, GLOBAL_CB_NODE* pCB, VkBool32 indexedDraw) {
VkBool32 result;
result = validate_status(my_data, pCB, CBSTATUS_NONE, CBSTATUS_VIEWPORT_SET, CBSTATUS_VIEWPORT_SET, VK_DEBUG_REPORT_ERROR_BIT_EXT, DRAWSTATE_VIEWPORT_NOT_BOUND, "Dynamic viewport state not set for this command buffer");
result |= validate_status(my_data, pCB, CBSTATUS_NONE, CBSTATUS_SCISSOR_SET, CBSTATUS_SCISSOR_SET, VK_DEBUG_REPORT_ERROR_BIT_EXT, DRAWSTATE_SCISSOR_NOT_BOUND, "Dynamic scissor state not set for this command buffer");
result |= validate_status(my_data, pCB, CBSTATUS_NONE, CBSTATUS_LINE_WIDTH_SET, CBSTATUS_LINE_WIDTH_SET, VK_DEBUG_REPORT_ERROR_BIT_EXT, DRAWSTATE_LINE_WIDTH_NOT_BOUND, "Dynamic line width state not set for this command buffer");
result |= validate_status(my_data, pCB, CBSTATUS_NONE, CBSTATUS_DEPTH_BIAS_SET, CBSTATUS_DEPTH_BIAS_SET, VK_DEBUG_REPORT_ERROR_BIT_EXT, DRAWSTATE_DEPTH_BIAS_NOT_BOUND, "Dynamic depth bias state not set for this command buffer");
result |= validate_status(my_data, pCB, CBSTATUS_COLOR_BLEND_WRITE_ENABLE, CBSTATUS_BLEND_SET, CBSTATUS_BLEND_SET, VK_DEBUG_REPORT_ERROR_BIT_EXT, DRAWSTATE_BLEND_NOT_BOUND, "Dynamic blend object state not set for this command buffer");
result |= validate_status(my_data, pCB, CBSTATUS_DEPTH_WRITE_ENABLE, CBSTATUS_DEPTH_BOUNDS_SET, CBSTATUS_DEPTH_BOUNDS_SET, VK_DEBUG_REPORT_ERROR_BIT_EXT, DRAWSTATE_DEPTH_BOUNDS_NOT_BOUND, "Dynamic depth bounds state not set for this command buffer");
result |= validate_status(my_data, pCB, CBSTATUS_STENCIL_TEST_ENABLE, CBSTATUS_STENCIL_READ_MASK_SET, CBSTATUS_STENCIL_READ_MASK_SET, VK_DEBUG_REPORT_ERROR_BIT_EXT, DRAWSTATE_STENCIL_NOT_BOUND, "Dynamic stencil read mask state not set for this command buffer");
result |= validate_status(my_data, pCB, CBSTATUS_STENCIL_TEST_ENABLE, CBSTATUS_STENCIL_WRITE_MASK_SET, CBSTATUS_STENCIL_WRITE_MASK_SET, VK_DEBUG_REPORT_ERROR_BIT_EXT, DRAWSTATE_STENCIL_NOT_BOUND, "Dynamic stencil write mask state not set for this command buffer");
result |= validate_status(my_data, pCB, CBSTATUS_STENCIL_TEST_ENABLE, CBSTATUS_STENCIL_REFERENCE_SET, CBSTATUS_STENCIL_REFERENCE_SET, VK_DEBUG_REPORT_ERROR_BIT_EXT, DRAWSTATE_STENCIL_NOT_BOUND, "Dynamic stencil reference state not set for this command buffer");
if (indexedDraw)
result |= validate_status(my_data, pCB, CBSTATUS_NONE, CBSTATUS_INDEX_BUFFER_BOUND, CBSTATUS_INDEX_BUFFER_BOUND, VK_DEBUG_REPORT_ERROR_BIT_EXT, DRAWSTATE_INDEX_BUFFER_NOT_BOUND, "Index buffer object not bound to this command buffer when Indexed Draw attempted");
return result;
}
// Verify attachment reference compatibility according to spec
// If one array is larger, treat missing elements of shorter array as VK_ATTACHMENT_UNUSED & other array much match this
// If both AttachmentReference arrays have requested index, check their corresponding AttachementDescriptions
// to make sure that format and samples counts match.
// If not, they are not compatible.
static bool attachment_references_compatible(const uint32_t index, const VkAttachmentReference* pPrimary, const uint32_t primaryCount, const VkAttachmentDescription* pPrimaryAttachments,
const VkAttachmentReference* pSecondary, const uint32_t secondaryCount, const VkAttachmentDescription* pSecondaryAttachments)
{
if (index >= primaryCount) { // Check secondary as if primary is VK_ATTACHMENT_UNUSED
if (VK_ATTACHMENT_UNUSED != pSecondary[index].attachment)
return false;
} else if (index >= secondaryCount) { // Check primary as if secondary is VK_ATTACHMENT_UNUSED
if (VK_ATTACHMENT_UNUSED != pPrimary[index].attachment)
return false;
} else { // format and sample count must match
if ((pPrimaryAttachments[pPrimary[index].attachment].format == pSecondaryAttachments[pSecondary[index].attachment].format) &&
(pPrimaryAttachments[pPrimary[index].attachment].samples == pSecondaryAttachments[pSecondary[index].attachment].samples))
return true;
}
// Format and sample counts didn't match
return false;
}
// For give primary and secondary RenderPass objects, verify that they're compatible
static bool verify_renderpass_compatibility(layer_data* my_data, const VkRenderPass primaryRP, const VkRenderPass secondaryRP, string& errorMsg)
{
stringstream errorStr;
if (my_data->renderPassMap.find(primaryRP) == my_data->renderPassMap.end()) {
errorStr << "invalid VkRenderPass (" << primaryRP << ")";
errorMsg = errorStr.str();
return false;
} else if (my_data->renderPassMap.find(secondaryRP) == my_data->renderPassMap.end()) {
errorStr << "invalid VkRenderPass (" << secondaryRP << ")";
errorMsg = errorStr.str();
return false;
}
// Trivial pass case is exact same RP
if (primaryRP == secondaryRP) {
return true;
}
const VkRenderPassCreateInfo* primaryRPCI = my_data->renderPassMap[primaryRP]->pCreateInfo;
const VkRenderPassCreateInfo* secondaryRPCI = my_data->renderPassMap[secondaryRP]->pCreateInfo;
if (primaryRPCI->subpassCount != secondaryRPCI->subpassCount) {
errorStr << "RenderPass for primary cmdBuffer has " << primaryRPCI->subpassCount << " subpasses but renderPass for secondary cmdBuffer has " << secondaryRPCI->subpassCount << " subpasses.";
errorMsg = errorStr.str();
return false;
}
uint32_t spIndex = 0;
for (spIndex = 0; spIndex < primaryRPCI->subpassCount; ++spIndex) {
// For each subpass, verify that corresponding color, input, resolve & depth/stencil attachment references are compatible
uint32_t primaryColorCount = primaryRPCI->pSubpasses[spIndex].colorAttachmentCount;
uint32_t secondaryColorCount = secondaryRPCI->pSubpasses[spIndex].colorAttachmentCount;
uint32_t colorMax = std::max(primaryColorCount, secondaryColorCount);
for (uint32_t cIdx = 0; cIdx < colorMax; ++cIdx) {
if (!attachment_references_compatible(cIdx, primaryRPCI->pSubpasses[spIndex].pColorAttachments, primaryColorCount, primaryRPCI->pAttachments,
secondaryRPCI->pSubpasses[spIndex].pColorAttachments, secondaryColorCount, secondaryRPCI->pAttachments)) {
errorStr << "color attachments at index " << cIdx << " of subpass index " << spIndex << " are not compatible.";
errorMsg = errorStr.str();
return false;
} else if (!attachment_references_compatible(cIdx, primaryRPCI->pSubpasses[spIndex].pResolveAttachments, primaryColorCount, primaryRPCI->pAttachments,
secondaryRPCI->pSubpasses[spIndex].pResolveAttachments, secondaryColorCount, secondaryRPCI->pAttachments)) {
errorStr << "resolve attachments at index " << cIdx << " of subpass index " << spIndex << " are not compatible.";
errorMsg = errorStr.str();
return false;
} else if (!attachment_references_compatible(cIdx, primaryRPCI->pSubpasses[spIndex].pDepthStencilAttachment, primaryColorCount, primaryRPCI->pAttachments,
secondaryRPCI->pSubpasses[spIndex].pDepthStencilAttachment, secondaryColorCount, secondaryRPCI->pAttachments)) {
errorStr << "depth/stencil attachments at index " << cIdx << " of subpass index " << spIndex << " are not compatible.";
errorMsg = errorStr.str();
return false;
}
}
uint32_t primaryInputCount = primaryRPCI->pSubpasses[spIndex].inputAttachmentCount;
uint32_t secondaryInputCount = secondaryRPCI->pSubpasses[spIndex].inputAttachmentCount;
uint32_t inputMax = std::max(primaryInputCount, secondaryInputCount);
for (uint32_t i = 0; i < inputMax; ++i) {
if (!attachment_references_compatible(i, primaryRPCI->pSubpasses[spIndex].pInputAttachments, primaryColorCount, primaryRPCI->pAttachments,
secondaryRPCI->pSubpasses[spIndex].pInputAttachments, secondaryColorCount, secondaryRPCI->pAttachments)) {
errorStr << "input attachments at index " << i << " of subpass index " << spIndex << " are not compatible.";
errorMsg = errorStr.str();
return false;
}
}
}
return true;
}
// For give SET_NODE, verify that its Set is compatible w/ the setLayout corresponding to pipelineLayout[layoutIndex]
static bool verify_set_layout_compatibility(layer_data* my_data, const SET_NODE* pSet, const VkPipelineLayout layout, const uint32_t layoutIndex, string& errorMsg)
{
stringstream errorStr;
if (my_data->pipelineLayoutMap.find(layout) == my_data->pipelineLayoutMap.end()) {
errorStr << "invalid VkPipelineLayout (" << layout << ")";
errorMsg = errorStr.str();
return false;
}
PIPELINE_LAYOUT_NODE pl = my_data->pipelineLayoutMap[layout];
if (layoutIndex >= pl.descriptorSetLayouts.size()) {
errorStr << "VkPipelineLayout (" << layout << ") only contains " << pl.descriptorSetLayouts.size() << " setLayouts corresponding to sets 0-" << pl.descriptorSetLayouts.size()-1 << ", but you're attempting to bind set to index " << layoutIndex;
errorMsg = errorStr.str();
return false;
}
// Get the specific setLayout from PipelineLayout that overlaps this set
LAYOUT_NODE* pLayoutNode = my_data->descriptorSetLayoutMap[pl.descriptorSetLayouts[layoutIndex]];
if (pLayoutNode->layout == pSet->pLayout->layout) { // trivial pass case
return true;
}
size_t descriptorCount = pLayoutNode->descriptorTypes.size();
if (descriptorCount != pSet->pLayout->descriptorTypes.size()) {
errorStr << "setLayout " << layoutIndex << " from pipelineLayout " << layout << " has " << descriptorCount << " descriptors, but corresponding set being bound has " << pSet->pLayout->descriptorTypes.size() << " descriptors.";
errorMsg = errorStr.str();
return false; // trivial fail case
}
// Now need to check set against corresponding pipelineLayout to verify compatibility
for (size_t i=0; i<descriptorCount; ++i) {
// Need to verify that layouts are identically defined
// TODO : Is below sufficient? Making sure that types & stageFlags match per descriptor
// do we also need to check immutable samplers?
if (pLayoutNode->descriptorTypes[i] != pSet->pLayout->descriptorTypes[i]) {
errorStr << "descriptor " << i << " for descriptorSet being bound is type '" << string_VkDescriptorType(pSet->pLayout->descriptorTypes[i]) << "' but corresponding descriptor from pipelineLayout is type '" << string_VkDescriptorType(pLayoutNode->descriptorTypes[i]) << "'";
errorMsg = errorStr.str();
return false;
}
if (pLayoutNode->stageFlags[i] != pSet->pLayout->stageFlags[i]) {
errorStr << "stageFlags " << i << " for descriptorSet being bound is " << pSet->pLayout->stageFlags[i] << "' but corresponding descriptor from pipelineLayout has stageFlags " << pLayoutNode->stageFlags[i];
errorMsg = errorStr.str();
return false;
}
}
return true;
}
// Validate that data for each specialization entry is fully contained within the buffer.
static VkBool32
validate_specialization_offsets(layer_data *my_data, VkPipelineShaderStageCreateInfo const *info)
{
VkBool32 pass = VK_TRUE;
VkSpecializationInfo const *spec = info->pSpecializationInfo;
if (spec) {
for (auto i = 0u; i < spec->mapEntryCount; i++) {
if (spec->pMapEntries[i].offset + spec->pMapEntries[i].size > spec->dataSize) {
if (log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DEVICE_EXT,
/*dev*/0, __LINE__, SHADER_CHECKER_BAD_SPECIALIZATION, "SC",
"Specialization entry %u (for constant id %u) references memory outside provided "
"specialization data (bytes %u.."
PRINTF_SIZE_T_SPECIFIER "; " PRINTF_SIZE_T_SPECIFIER " bytes provided)",
i, spec->pMapEntries[i].constantID,
spec->pMapEntries[i].offset,
spec->pMapEntries[i].offset + spec->pMapEntries[i].size - 1,
spec->dataSize)) {
pass = VK_FALSE;
}
}
}
}
return pass;
}
// Validate that the shaders used by the given pipeline
// As a side effect this function also records the sets that are actually used by the pipeline
static VkBool32
validate_pipeline_shaders(layer_data *my_data, VkDevice dev, PIPELINE_NODE* pPipeline)
{
VkGraphicsPipelineCreateInfo const *pCreateInfo = &pPipeline->graphicsPipelineCI;
/* We seem to allow pipeline stages to be specified out of order, so collect and identify them
* before trying to do anything more: */
int vertex_stage = get_shader_stage_id(VK_SHADER_STAGE_VERTEX_BIT);
int fragment_stage = get_shader_stage_id(VK_SHADER_STAGE_FRAGMENT_BIT);
shader_module *shaders[5];
memset(shaders, 0, sizeof(shaders));
spirv_inst_iter entrypoints[5];
memset(entrypoints, 0, sizeof(entrypoints));
RENDER_PASS_NODE const *rp = 0;
VkPipelineVertexInputStateCreateInfo const *vi = 0;
VkBool32 pass = VK_TRUE;
for (uint32_t i = 0; i < pCreateInfo->stageCount; i++) {
VkPipelineShaderStageCreateInfo const *pStage = &pCreateInfo->pStages[i];
if (pStage->sType == VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO) {
if ((pStage->stage & (VK_SHADER_STAGE_VERTEX_BIT | VK_SHADER_STAGE_GEOMETRY_BIT | VK_SHADER_STAGE_FRAGMENT_BIT
| VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT | VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT)) == 0) {
if (log_msg(my_data->report_data, VK_DEBUG_REPORT_WARNING_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DEVICE_EXT, /*dev*/0, __LINE__, SHADER_CHECKER_UNKNOWN_STAGE, "SC",
"Unknown shader stage %d", pStage->stage)) {
pass = VK_FALSE;
}
}
else {
pass = validate_specialization_offsets(my_data, pStage) && pass;
auto stage_id = get_shader_stage_id(pStage->stage);
shader_module *module = my_data->shaderModuleMap[pStage->module];
shaders[stage_id] = module;
/* find the entrypoint */
entrypoints[stage_id] = find_entrypoint(module, pStage->pName, pStage->stage);
if (entrypoints[stage_id] == module->end()) {
if (log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DEVICE_EXT, /*dev*/0, __LINE__, SHADER_CHECKER_MISSING_ENTRYPOINT, "SC",
"No entrypoint found named `%s` for stages %u", pStage->pName, pStage->stage)) {
pass = VK_FALSE;
}
}
/* mark accessible ids */
std::unordered_set<uint32_t> accessible_ids;
mark_accessible_ids(module, entrypoints[stage_id], accessible_ids);
/* validate descriptor set layout against what the entrypoint actually uses */
std::map<std::pair<unsigned, unsigned>, interface_var> descriptor_uses;
collect_interface_by_descriptor_slot(my_data, dev, module, spv::StorageClassUniform,
accessible_ids,
descriptor_uses);
auto layouts = pCreateInfo->layout != VK_NULL_HANDLE ?
&(my_data->pipelineLayoutMap[pCreateInfo->layout].descriptorSetLayouts) : nullptr;
for (auto it = descriptor_uses.begin(); it != descriptor_uses.end(); it++) {
// As a side-effect of this function, capture which sets are used by the pipeline
pPipeline->active_sets.insert(it->first.first);
/* find the matching binding */
auto found = has_descriptor_binding(my_data, layouts, it->first);
if (!found) {
char type_name[1024];
describe_type(type_name, module, it->second.type_id);
if (log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DEVICE_EXT, /*dev*/0, __LINE__,
SHADER_CHECKER_MISSING_DESCRIPTOR, "SC",
"Shader uses descriptor slot %u.%u (used as type `%s`) but not declared in pipeline layout",
it->first.first, it->first.second, type_name)) {
pass = VK_FALSE;
}
}
}
}
}
}
if (pCreateInfo->renderPass != VK_NULL_HANDLE)
rp = my_data->renderPassMap[pCreateInfo->renderPass];
vi = pCreateInfo->pVertexInputState;
if (vi) {
pass = validate_vi_consistency(my_data, dev, vi) && pass;
}
if (shaders[vertex_stage]) {
pass = validate_vi_against_vs_inputs(my_data, dev, vi, shaders[vertex_stage], entrypoints[vertex_stage]) && pass;
}
/* TODO: enforce rules about present combinations of shaders */
int producer = get_shader_stage_id(VK_SHADER_STAGE_VERTEX_BIT);
int consumer = get_shader_stage_id(VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT);
while (!shaders[producer] && producer != fragment_stage) {
producer++;
consumer++;
}
for (; producer != fragment_stage && consumer <= fragment_stage; consumer++) {
assert(shaders[producer]);
if (shaders[consumer]) {
pass = validate_interface_between_stages(my_data, dev,
shaders[producer], entrypoints[producer],
shader_stage_attribs[producer].name,
shaders[consumer], entrypoints[consumer],
shader_stage_attribs[consumer].name,
shader_stage_attribs[consumer].arrayed_input) && pass;
producer = consumer;
}
}
if (shaders[fragment_stage] && rp) {
pass = validate_fs_outputs_against_render_pass(my_data, dev, shaders[fragment_stage],
entrypoints[fragment_stage], rp, pCreateInfo->subpass) && pass;
}
return pass;
}
// Return Set node ptr for specified set or else NULL
static SET_NODE* getSetNode(layer_data* my_data, const VkDescriptorSet set)
{
if (my_data->setMap.find(set) == my_data->setMap.end()) {
return NULL;
}
return my_data->setMap[set];
}
// For the given command buffer, verify that for each set set in activeSetNodes
// that any dynamic descriptor in that set has a valid dynamic offset bound.
// To be valid, the dynamic offset combined with the offet and range from its
// descriptor update must not overflow the size of its buffer being updated
static VkBool32 validate_dynamic_offsets(layer_data* my_data, const GLOBAL_CB_NODE* pCB, const vector<SET_NODE*> activeSetNodes)
{
VkBool32 result = VK_FALSE;
VkWriteDescriptorSet* pWDS = NULL;
uint32_t dynOffsetIndex = 0;
VkDeviceSize bufferSize = 0;
for (auto set_node : activeSetNodes) {
for (uint32_t i=0; i < set_node->descriptorCount; ++i) {
switch (set_node->ppDescriptors[i]->sType) {
case VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET:
pWDS = (VkWriteDescriptorSet*)set_node->ppDescriptors[i];
if ((pWDS->descriptorType == VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC) ||
(pWDS->descriptorType == VK_DESCRIPTOR_TYPE_STORAGE_BUFFER_DYNAMIC)) {
for (uint32_t j=0; j<pWDS->descriptorCount; ++j) {
bufferSize = my_data->bufferMap[pWDS->pBufferInfo[j].buffer].create_info->size;
if (pWDS->pBufferInfo[j].range == VK_WHOLE_SIZE) {
if ((pCB->dynamicOffsets[dynOffsetIndex] +
pWDS->pBufferInfo[j].offset) > bufferSize) {
result |= log_msg(
my_data->report_data,
VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT,
(uint64_t)set_node->set, __LINE__,
DRAWSTATE_DYNAMIC_OFFSET_OVERFLOW, "DS",
"VkDescriptorSet (%#" PRIxLEAST64
") bound as set #%u has range of "
"VK_WHOLE_SIZE but dynamic offset %u "
"combined with offet %#" PRIxLEAST64
" oversteps its buffer (%#" PRIxLEAST64
") which has a size of %#" PRIxLEAST64 ".",
(uint64_t)set_node->set, i,
pCB->dynamicOffsets[dynOffsetIndex],
pWDS->pBufferInfo[j].offset,
(uint64_t)pWDS->pBufferInfo[j].buffer,
bufferSize);
}
} else if ((pCB->dynamicOffsets[dynOffsetIndex] +
pWDS->pBufferInfo[j].offset +
pWDS->pBufferInfo[j].range) >
bufferSize) {
result |= log_msg(
my_data->report_data,
VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT,
(uint64_t)set_node->set, __LINE__,
DRAWSTATE_DYNAMIC_OFFSET_OVERFLOW, "DS",
"VkDescriptorSet (%#" PRIxLEAST64
") bound as set #%u has dynamic offset %u. "
"Combined with offet %#" PRIxLEAST64
" and range %#" PRIxLEAST64
" from its update, this oversteps its buffer "
"(%#" PRIxLEAST64
") which has a size of %#" PRIxLEAST64 ".",
(uint64_t)set_node->set, i,
pCB->dynamicOffsets[dynOffsetIndex],
pWDS->pBufferInfo[j].offset,
pWDS->pBufferInfo[j].range,
(uint64_t)pWDS->pBufferInfo[j].buffer,
bufferSize);
}
dynOffsetIndex++;
i += j; // Advance i to end of this set of descriptors (++i at end of for loop will move 1 index past last of these descriptors)
}
}
break;
default: // Currently only shadowing Write update nodes so shouldn't get here
assert(0);
continue;
}
}
}
return result;
}
// Validate overall state at the time of a draw call
static VkBool32 validate_draw_state(layer_data* my_data, GLOBAL_CB_NODE* pCB, VkBool32 indexedDraw) {
// First check flag states
VkBool32 result = validate_draw_state_flags(my_data, pCB, indexedDraw);
PIPELINE_NODE* pPipe = getPipeline(my_data, pCB->lastBoundPipeline);
// Now complete other state checks
// TODO : Currently only performing next check if *something* was bound (non-zero last bound)
// There is probably a better way to gate when this check happens, and to know if something *should* have been bound
// We should have that check separately and then gate this check based on that check
if (pPipe) {
if (pCB->lastBoundPipelineLayout) {
string errorString;
// Need a vector (vs. std::set) of active Sets for dynamicOffset validation in case same set bound w/ different offsets
vector<SET_NODE*> activeSetNodes;
for (auto setIndex : pPipe->active_sets) {
// If valid set is not bound throw an error
if ((pCB->boundDescriptorSets.size() <= setIndex) || (!pCB->boundDescriptorSets[setIndex])) {
result |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_DESCRIPTOR_SET_NOT_BOUND, "DS",
"VkPipeline %#" PRIxLEAST64 " uses set #%u but that set is not bound.", (uint64_t)pPipe->pipeline, setIndex);
} else if (!verify_set_layout_compatibility(my_data, my_data->setMap[pCB->boundDescriptorSets[setIndex]], pPipe->graphicsPipelineCI.layout, setIndex, errorString)) {
// Set is bound but not compatible w/ overlapping pipelineLayout from PSO
VkDescriptorSet setHandle = my_data->setMap[pCB->boundDescriptorSets[setIndex]]->set;
result |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT, (uint64_t)setHandle, __LINE__, DRAWSTATE_PIPELINE_LAYOUTS_INCOMPATIBLE, "DS",
"VkDescriptorSet (%#" PRIxLEAST64 ") bound as set #%u is not compatible with overlapping VkPipelineLayout %#" PRIxLEAST64 " due to: %s",
(uint64_t)setHandle, setIndex, (uint64_t)pPipe->graphicsPipelineCI.layout, errorString.c_str());
} else { // Valid set is bound and layout compatible, validate that it's updated and verify any dynamic offsets
// Pull the set node
SET_NODE* pSet = my_data->setMap[pCB->boundDescriptorSets[setIndex]];
// Save vector of all active sets to verify dynamicOffsets below
activeSetNodes.push_back(pSet);
// Make sure set has been updated
if (!pSet->pUpdateStructs) {
result |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT, (uint64_t) pSet->set, __LINE__, DRAWSTATE_DESCRIPTOR_SET_NOT_UPDATED, "DS",
"DS %#" PRIxLEAST64 " bound but it was never updated. It is now being used to draw so this will result in undefined behavior.", (uint64_t) pSet->set);
}
}
}
// For each dynamic descriptor, make sure dynamic offset doesn't overstep buffer
if (!pCB->dynamicOffsets.empty())
result |= validate_dynamic_offsets(my_data, pCB, activeSetNodes);
}
// Verify Vtx binding
if (pPipe->vtxBindingCount > 0) {
VkPipelineVertexInputStateCreateInfo *vtxInCI = &pPipe->vertexInputCI;
for (uint32_t i = 0; i < vtxInCI->vertexBindingDescriptionCount; i++) {
if ((pCB->currentDrawData.buffers.size() < (i+1)) || (pCB->currentDrawData.buffers[i] == VK_NULL_HANDLE)) {
result |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_VTX_INDEX_OUT_OF_BOUNDS, "DS",
"The Pipeline State Object (%#" PRIxLEAST64 ") expects that this Command Buffer's vertex binding Index %d should be set via vkCmdBindVertexBuffers.",
(uint64_t)pCB->lastBoundPipeline, i);
}
}
} else {
if (!pCB->currentDrawData.buffers.empty()) {
result |= log_msg(my_data->report_data, VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_VTX_INDEX_OUT_OF_BOUNDS,
"DS", "Vertex buffers are bound to command buffer (%#" PRIxLEAST64 ") but no vertex buffers are attached to this Pipeline State Object (%#" PRIxLEAST64 ").",
(uint64_t)pCB->commandBuffer, (uint64_t)pCB->lastBoundPipeline);
}
}
// If Viewport or scissors are dynamic, verify that dynamic count matches PSO count
VkBool32 dynViewport = isDynamic(pPipe, VK_DYNAMIC_STATE_VIEWPORT);
VkBool32 dynScissor = isDynamic(pPipe, VK_DYNAMIC_STATE_SCISSOR);
if (dynViewport) {
if (pCB->viewports.size() != pPipe->graphicsPipelineCI.pViewportState->viewportCount) {
result |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_VIEWPORT_SCISSOR_MISMATCH, "DS",
"Dynamic viewportCount from vkCmdSetViewport() is " PRINTF_SIZE_T_SPECIFIER ", but PSO viewportCount is %u. These counts must match.", pCB->viewports.size(), pPipe->graphicsPipelineCI.pViewportState->viewportCount);
}
}
if (dynScissor) {
if (pCB->scissors.size() != pPipe->graphicsPipelineCI.pViewportState->scissorCount) {
result |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_VIEWPORT_SCISSOR_MISMATCH, "DS",
"Dynamic scissorCount from vkCmdSetScissor() is " PRINTF_SIZE_T_SPECIFIER ", but PSO scissorCount is %u. These counts must match.", pCB->scissors.size(), pPipe->graphicsPipelineCI.pViewportState->scissorCount);
}
}
}
return result;
}
// Verify that create state for a pipeline is valid
static VkBool32 verifyPipelineCreateState(layer_data* my_data, const VkDevice device, PIPELINE_NODE* pPipeline)
{
VkBool32 skipCall = VK_FALSE;
if (!validate_pipeline_shaders(my_data, device, pPipeline)) {
skipCall = VK_TRUE;
}
// VS is required
if (!(pPipeline->active_shaders & VK_SHADER_STAGE_VERTEX_BIT)) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_INVALID_PIPELINE_CREATE_STATE, "DS",
"Invalid Pipeline CreateInfo State: Vtx Shader required");
}
// Either both or neither TC/TE shaders should be defined
if (((pPipeline->active_shaders & VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT) == 0) !=
((pPipeline->active_shaders & VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT) == 0) ) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_INVALID_PIPELINE_CREATE_STATE, "DS",
"Invalid Pipeline CreateInfo State: TE and TC shaders must be included or excluded as a pair");
}
// Compute shaders should be specified independent of Gfx shaders
if ((pPipeline->active_shaders & VK_SHADER_STAGE_COMPUTE_BIT) &&
(pPipeline->active_shaders & (VK_SHADER_STAGE_VERTEX_BIT | VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT |
VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT | VK_SHADER_STAGE_GEOMETRY_BIT |
VK_SHADER_STAGE_FRAGMENT_BIT))) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_INVALID_PIPELINE_CREATE_STATE, "DS",
"Invalid Pipeline CreateInfo State: Do not specify Compute Shader for Gfx Pipeline");
}
// VK_PRIMITIVE_TOPOLOGY_PATCH_LIST primitive topology is only valid for tessellation pipelines.
// Mismatching primitive topology and tessellation fails graphics pipeline creation.
if (pPipeline->active_shaders & (VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT | VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT) &&
(pPipeline->iaStateCI.topology != VK_PRIMITIVE_TOPOLOGY_PATCH_LIST)) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_INVALID_PIPELINE_CREATE_STATE, "DS",
"Invalid Pipeline CreateInfo State: VK_PRIMITIVE_TOPOLOGY_PATCH_LIST must be set as IA topology for tessellation pipelines");
}
if (pPipeline->iaStateCI.topology == VK_PRIMITIVE_TOPOLOGY_PATCH_LIST) {
if (~pPipeline->active_shaders & VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_INVALID_PIPELINE_CREATE_STATE, "DS",
"Invalid Pipeline CreateInfo State: VK_PRIMITIVE_TOPOLOGY_PATCH_LIST primitive topology is only valid for tessellation pipelines");
}
if (!pPipeline->tessStateCI.patchControlPoints || (pPipeline->tessStateCI.patchControlPoints > 32)) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_INVALID_PIPELINE_CREATE_STATE, "DS",
"Invalid Pipeline CreateInfo State: VK_PRIMITIVE_TOPOLOGY_PATCH_LIST primitive topology used with patchControlPoints value %u."
" patchControlPoints should be >0 and <=32.", pPipeline->tessStateCI.patchControlPoints);
}
}
// Viewport state must be included and viewport and scissor counts should always match
// NOTE : Even if these are flagged as dynamic, counts need to be set correctly for shader compiler
if (!pPipeline->graphicsPipelineCI.pViewportState) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_VIEWPORT_SCISSOR_MISMATCH, "DS",
"Gfx Pipeline pViewportState is null. Even if viewport and scissors are dynamic PSO must include viewportCount and scissorCount in pViewportState.");
} else if (pPipeline->graphicsPipelineCI.pViewportState->scissorCount != pPipeline->graphicsPipelineCI.pViewportState->viewportCount) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_VIEWPORT_SCISSOR_MISMATCH, "DS",
"Gfx Pipeline viewport count (%u) must match scissor count (%u).", pPipeline->vpStateCI.viewportCount, pPipeline->vpStateCI.scissorCount);
} else {
// If viewport or scissor are not dynamic, then verify that data is appropriate for count
VkBool32 dynViewport = isDynamic(pPipeline, VK_DYNAMIC_STATE_VIEWPORT);
VkBool32 dynScissor = isDynamic(pPipeline, VK_DYNAMIC_STATE_SCISSOR);
if (!dynViewport) {
if (pPipeline->graphicsPipelineCI.pViewportState->viewportCount && !pPipeline->graphicsPipelineCI.pViewportState->pViewports) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_VIEWPORT_SCISSOR_MISMATCH, "DS",
"Gfx Pipeline viewportCount is %u, but pViewports is NULL. For non-zero viewportCount, you must either include pViewports data, or include viewport in pDynamicState and set it with vkCmdSetViewport().", pPipeline->graphicsPipelineCI.pViewportState->viewportCount);
}
}
if (!dynScissor) {
if (pPipeline->graphicsPipelineCI.pViewportState->scissorCount && !pPipeline->graphicsPipelineCI.pViewportState->pScissors) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_VIEWPORT_SCISSOR_MISMATCH, "DS",
"Gfx Pipeline scissorCount is %u, but pScissors is NULL. For non-zero scissorCount, you must either include pScissors data, or include scissor in pDynamicState and set it with vkCmdSetScissor().", pPipeline->graphicsPipelineCI.pViewportState->scissorCount);
}
}
}
return skipCall;
}
// Init the pipeline mapping info based on pipeline create info LL tree
// Threading note : Calls to this function should wrapped in mutex
// TODO : this should really just be in the constructor for PIPELINE_NODE
static PIPELINE_NODE* initGraphicsPipeline(layer_data* dev_data, const VkGraphicsPipelineCreateInfo* pCreateInfo, PIPELINE_NODE* pBasePipeline)
{
PIPELINE_NODE* pPipeline = new PIPELINE_NODE;
if (pBasePipeline) {
*pPipeline = *pBasePipeline;
}
// First init create info
memcpy(&pPipeline->graphicsPipelineCI, pCreateInfo, sizeof(VkGraphicsPipelineCreateInfo));
size_t bufferSize = 0;
const VkPipelineVertexInputStateCreateInfo* pVICI = NULL;
const VkPipelineColorBlendStateCreateInfo* pCBCI = NULL;
for (uint32_t i = 0; i < pCreateInfo->stageCount; i++) {
const VkPipelineShaderStageCreateInfo *pPSSCI = &pCreateInfo->pStages[i];
switch (pPSSCI->stage) {
case VK_SHADER_STAGE_VERTEX_BIT:
memcpy(&pPipeline->vsCI, pPSSCI, sizeof(VkPipelineShaderStageCreateInfo));
pPipeline->active_shaders |= VK_SHADER_STAGE_VERTEX_BIT;
break;
case VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT:
memcpy(&pPipeline->tcsCI, pPSSCI, sizeof(VkPipelineShaderStageCreateInfo));
pPipeline->active_shaders |= VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT;
break;
case VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT:
memcpy(&pPipeline->tesCI, pPSSCI, sizeof(VkPipelineShaderStageCreateInfo));
pPipeline->active_shaders |= VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT;
break;
case VK_SHADER_STAGE_GEOMETRY_BIT:
memcpy(&pPipeline->gsCI, pPSSCI, sizeof(VkPipelineShaderStageCreateInfo));
pPipeline->active_shaders |= VK_SHADER_STAGE_GEOMETRY_BIT;
break;
case VK_SHADER_STAGE_FRAGMENT_BIT:
memcpy(&pPipeline->fsCI, pPSSCI, sizeof(VkPipelineShaderStageCreateInfo));
pPipeline->active_shaders |= VK_SHADER_STAGE_FRAGMENT_BIT;
break;
case VK_SHADER_STAGE_COMPUTE_BIT:
// TODO : Flag error, CS is specified through VkComputePipelineCreateInfo
pPipeline->active_shaders |= VK_SHADER_STAGE_COMPUTE_BIT;
break;
default:
// TODO : Flag error
break;
}
}
// Copy over GraphicsPipelineCreateInfo structure embedded pointers
if (pCreateInfo->stageCount != 0) {
pPipeline->graphicsPipelineCI.pStages = new VkPipelineShaderStageCreateInfo[pCreateInfo->stageCount];
bufferSize = pCreateInfo->stageCount * sizeof(VkPipelineShaderStageCreateInfo);
memcpy((void*)pPipeline->graphicsPipelineCI.pStages, pCreateInfo->pStages, bufferSize);
}
if (pCreateInfo->pVertexInputState != NULL) {
memcpy((void*)&pPipeline->vertexInputCI, pCreateInfo->pVertexInputState , sizeof(VkPipelineVertexInputStateCreateInfo));
// Copy embedded ptrs
pVICI = pCreateInfo->pVertexInputState;
pPipeline->vtxBindingCount = pVICI->vertexBindingDescriptionCount;
if (pPipeline->vtxBindingCount) {
pPipeline->pVertexBindingDescriptions = new VkVertexInputBindingDescription[pPipeline->vtxBindingCount];
bufferSize = pPipeline->vtxBindingCount * sizeof(VkVertexInputBindingDescription);
memcpy((void*)pPipeline->pVertexBindingDescriptions, pVICI->pVertexBindingDescriptions, bufferSize);
}
pPipeline->vtxAttributeCount = pVICI->vertexAttributeDescriptionCount;
if (pPipeline->vtxAttributeCount) {
pPipeline->pVertexAttributeDescriptions = new VkVertexInputAttributeDescription[pPipeline->vtxAttributeCount];
bufferSize = pPipeline->vtxAttributeCount * sizeof(VkVertexInputAttributeDescription);
memcpy((void*)pPipeline->pVertexAttributeDescriptions, pVICI->pVertexAttributeDescriptions, bufferSize);
}
pPipeline->graphicsPipelineCI.pVertexInputState = &pPipeline->vertexInputCI;
}
if (pCreateInfo->pInputAssemblyState != NULL) {
memcpy((void*)&pPipeline->iaStateCI, pCreateInfo->pInputAssemblyState, sizeof(VkPipelineInputAssemblyStateCreateInfo));
pPipeline->graphicsPipelineCI.pInputAssemblyState = &pPipeline->iaStateCI;
}
if (pCreateInfo->pTessellationState != NULL) {
memcpy((void*)&pPipeline->tessStateCI, pCreateInfo->pTessellationState, sizeof(VkPipelineTessellationStateCreateInfo));
pPipeline->graphicsPipelineCI.pTessellationState = &pPipeline->tessStateCI;
}
if (pCreateInfo->pViewportState != NULL) {
memcpy((void*)&pPipeline->vpStateCI, pCreateInfo->pViewportState, sizeof(VkPipelineViewportStateCreateInfo));
pPipeline->graphicsPipelineCI.pViewportState = &pPipeline->vpStateCI;
}
if (pCreateInfo->pRasterizationState != NULL) {
memcpy((void*)&pPipeline->rsStateCI, pCreateInfo->pRasterizationState, sizeof(VkPipelineRasterizationStateCreateInfo));
pPipeline->graphicsPipelineCI.pRasterizationState = &pPipeline->rsStateCI;
}
if (pCreateInfo->pMultisampleState != NULL) {
memcpy((void*)&pPipeline->msStateCI, pCreateInfo->pMultisampleState, sizeof(VkPipelineMultisampleStateCreateInfo));
pPipeline->graphicsPipelineCI.pMultisampleState = &pPipeline->msStateCI;
}
if (pCreateInfo->pDepthStencilState != NULL) {
memcpy((void*)&pPipeline->dsStateCI, pCreateInfo->pDepthStencilState, sizeof(VkPipelineDepthStencilStateCreateInfo));
pPipeline->graphicsPipelineCI.pDepthStencilState = &pPipeline->dsStateCI;
}
if (pCreateInfo->pColorBlendState != NULL) {
memcpy((void*)&pPipeline->cbStateCI, pCreateInfo->pColorBlendState, sizeof(VkPipelineColorBlendStateCreateInfo));
// Copy embedded ptrs
pCBCI = pCreateInfo->pColorBlendState;
pPipeline->attachmentCount = pCBCI->attachmentCount;
if (pPipeline->attachmentCount) {
pPipeline->pAttachments = new VkPipelineColorBlendAttachmentState[pPipeline->attachmentCount];
bufferSize = pPipeline->attachmentCount * sizeof(VkPipelineColorBlendAttachmentState);
memcpy((void*)pPipeline->pAttachments, pCBCI->pAttachments, bufferSize);
}
pPipeline->graphicsPipelineCI.pColorBlendState = &pPipeline->cbStateCI;
}
if (pCreateInfo->pDynamicState != NULL) {
memcpy((void*)&pPipeline->dynStateCI, pCreateInfo->pDynamicState, sizeof(VkPipelineDynamicStateCreateInfo));
if (pPipeline->dynStateCI.dynamicStateCount) {
pPipeline->dynStateCI.pDynamicStates = new VkDynamicState[pPipeline->dynStateCI.dynamicStateCount];
bufferSize = pPipeline->dynStateCI.dynamicStateCount * sizeof(VkDynamicState);
memcpy((void*)pPipeline->dynStateCI.pDynamicStates, pCreateInfo->pDynamicState->pDynamicStates, bufferSize);
}
pPipeline->graphicsPipelineCI.pDynamicState = &pPipeline->dynStateCI;
}
pPipeline->active_sets.clear();
return pPipeline;
}
// Free the Pipeline nodes
static void deletePipelines(layer_data* my_data)
{
if (my_data->pipelineMap.size() <= 0)
return;
for (auto ii=my_data->pipelineMap.begin(); ii!=my_data->pipelineMap.end(); ++ii) {
if ((*ii).second->graphicsPipelineCI.stageCount != 0) {
delete[] (*ii).second->graphicsPipelineCI.pStages;
}
if ((*ii).second->pVertexBindingDescriptions) {
delete[] (*ii).second->pVertexBindingDescriptions;
}
if ((*ii).second->pVertexAttributeDescriptions) {
delete[] (*ii).second->pVertexAttributeDescriptions;
}
if ((*ii).second->pAttachments) {
delete[] (*ii).second->pAttachments;
}
if ((*ii).second->dynStateCI.dynamicStateCount != 0) {
delete[] (*ii).second->dynStateCI.pDynamicStates;
}
delete (*ii).second;
}
my_data->pipelineMap.clear();
}
// For given pipeline, return number of MSAA samples, or one if MSAA disabled
static VkSampleCountFlagBits getNumSamples(layer_data* my_data, const VkPipeline pipeline)
{
PIPELINE_NODE* pPipe = my_data->pipelineMap[pipeline];
if (VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO == pPipe->msStateCI.sType) {
return pPipe->msStateCI.rasterizationSamples;
}
return VK_SAMPLE_COUNT_1_BIT;
}
// Validate state related to the PSO
static VkBool32 validatePipelineState(layer_data* my_data, const GLOBAL_CB_NODE* pCB, const VkPipelineBindPoint pipelineBindPoint, const VkPipeline pipeline)
{
if (VK_PIPELINE_BIND_POINT_GRAPHICS == pipelineBindPoint) {
// Verify that any MSAA request in PSO matches sample# in bound FB
VkSampleCountFlagBits psoNumSamples = getNumSamples(my_data, pipeline);
if (pCB->activeRenderPass) {
const VkRenderPassCreateInfo* pRPCI = my_data->renderPassMap[pCB->activeRenderPass]->pCreateInfo;
const VkSubpassDescription* pSD = &pRPCI->pSubpasses[pCB->activeSubpass];
VkSampleCountFlagBits subpassNumSamples = (VkSampleCountFlagBits) 0;
uint32_t i;
for (i = 0; i < pSD->colorAttachmentCount; i++) {
VkSampleCountFlagBits samples;
if (pSD->pColorAttachments[i].attachment == VK_ATTACHMENT_UNUSED)
continue;
samples = pRPCI->pAttachments[pSD->pColorAttachments[i].attachment].samples;
if (subpassNumSamples == (VkSampleCountFlagBits) 0) {
subpassNumSamples = samples;
} else if (subpassNumSamples != samples) {
subpassNumSamples = (VkSampleCountFlagBits) -1;
break;
}
}
if (pSD->pDepthStencilAttachment && pSD->pDepthStencilAttachment->attachment != VK_ATTACHMENT_UNUSED) {
const VkSampleCountFlagBits samples = pRPCI->pAttachments[pSD->pDepthStencilAttachment->attachment].samples;
if (subpassNumSamples == (VkSampleCountFlagBits) 0)
subpassNumSamples = samples;
else if (subpassNumSamples != samples)
subpassNumSamples = (VkSampleCountFlagBits) -1;
}
if (psoNumSamples != subpassNumSamples) {
return log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_PIPELINE_EXT, (uint64_t) pipeline, __LINE__, DRAWSTATE_NUM_SAMPLES_MISMATCH, "DS",
"Num samples mismatch! Binding PSO (%#" PRIxLEAST64 ") with %u samples while current RenderPass (%#" PRIxLEAST64 ") w/ %u samples!",
(uint64_t) pipeline, psoNumSamples, (uint64_t) pCB->activeRenderPass, subpassNumSamples);
}
} else {
// TODO : I believe it's an error if we reach this point and don't have an activeRenderPass
// Verify and flag error as appropriate
}
// TODO : Add more checks here
} else {
// TODO : Validate non-gfx pipeline updates
}
return VK_FALSE;
}
// Block of code at start here specifically for managing/tracking DSs
// Return Pool node ptr for specified pool or else NULL
static DESCRIPTOR_POOL_NODE* getPoolNode(layer_data* my_data, const VkDescriptorPool pool)
{
if (my_data->descriptorPoolMap.find(pool) == my_data->descriptorPoolMap.end()) {
return NULL;
}
return my_data->descriptorPoolMap[pool];
}
static LAYOUT_NODE* getLayoutNode(layer_data* my_data, const VkDescriptorSetLayout layout) {
if (my_data->descriptorSetLayoutMap.find(layout) == my_data->descriptorSetLayoutMap.end()) {
return NULL;
}
return my_data->descriptorSetLayoutMap[layout];
}
// Return VK_FALSE if update struct is of valid type, otherwise flag error and return code from callback
static VkBool32 validUpdateStruct(layer_data* my_data, const VkDevice device, const GENERIC_HEADER* pUpdateStruct)
{
switch (pUpdateStruct->sType)
{
case VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET:
case VK_STRUCTURE_TYPE_COPY_DESCRIPTOR_SET:
return VK_FALSE;
default:
return log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_INVALID_UPDATE_STRUCT, "DS",
"Unexpected UPDATE struct of type %s (value %u) in vkUpdateDescriptors() struct tree", string_VkStructureType(pUpdateStruct->sType), pUpdateStruct->sType);
}
}
// Set count for given update struct in the last parameter
// Return value of skipCall, which is only VK_TRUE if error occurs and callback signals execution to cease
static uint32_t getUpdateCount(layer_data* my_data, const VkDevice device, const GENERIC_HEADER* pUpdateStruct)
{
switch (pUpdateStruct->sType)
{
case VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET:
return ((VkWriteDescriptorSet*)pUpdateStruct)->descriptorCount;
case VK_STRUCTURE_TYPE_COPY_DESCRIPTOR_SET:
// TODO : Need to understand this case better and make sure code is correct
return ((VkCopyDescriptorSet*)pUpdateStruct)->descriptorCount;
default:
return 0;
}
return 0;
}
// For given Layout Node and binding, return index where that binding begins
static uint32_t getBindingStartIndex(const LAYOUT_NODE* pLayout, const uint32_t binding)
{
uint32_t offsetIndex = 0;
for (uint32_t i = 0; i < pLayout->createInfo.bindingCount; i++) {
if (pLayout->createInfo.pBindings[i].binding == binding)
break;
offsetIndex += pLayout->createInfo.pBindings[i].descriptorCount;
}
return offsetIndex;
}
// For given layout node and binding, return last index that is updated
static uint32_t getBindingEndIndex(const LAYOUT_NODE* pLayout, const uint32_t binding)
{
uint32_t offsetIndex = 0;
for (uint32_t i = 0; i < pLayout->createInfo.bindingCount; i++) {
offsetIndex += pLayout->createInfo.pBindings[i].descriptorCount;
if (pLayout->createInfo.pBindings[i].binding == binding)
break;
}
return offsetIndex-1;
}
// For given layout and update, return the first overall index of the layout that is updated
static uint32_t getUpdateStartIndex(layer_data* my_data, const VkDevice device, const LAYOUT_NODE* pLayout, const uint32_t binding, const uint32_t arrayIndex, const GENERIC_HEADER* pUpdateStruct)
{
return getBindingStartIndex(pLayout, binding)+arrayIndex;
}
// For given layout and update, return the last overall index of the layout that is updated
static uint32_t getUpdateEndIndex(layer_data* my_data, const VkDevice device, const LAYOUT_NODE* pLayout, const uint32_t binding, const uint32_t arrayIndex, const GENERIC_HEADER* pUpdateStruct)
{
uint32_t count = getUpdateCount(my_data, device, pUpdateStruct);
return getBindingStartIndex(pLayout, binding)+arrayIndex+count-1;
}
// Verify that the descriptor type in the update struct matches what's expected by the layout
static VkBool32 validateUpdateConsistency(layer_data* my_data, const VkDevice device, const LAYOUT_NODE* pLayout, const GENERIC_HEADER* pUpdateStruct, uint32_t startIndex, uint32_t endIndex)
{
// First get actual type of update
VkBool32 skipCall = VK_FALSE;
VkDescriptorType actualType;
uint32_t i = 0;
switch (pUpdateStruct->sType)
{
case VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET:
actualType = ((VkWriteDescriptorSet*)pUpdateStruct)->descriptorType;
break;
case VK_STRUCTURE_TYPE_COPY_DESCRIPTOR_SET:
/* no need to validate */
return VK_FALSE;
break;
default:
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_INVALID_UPDATE_STRUCT, "DS",
"Unexpected UPDATE struct of type %s (value %u) in vkUpdateDescriptors() struct tree", string_VkStructureType(pUpdateStruct->sType), pUpdateStruct->sType);
}
if (VK_FALSE == skipCall) {
// Set first stageFlags as reference and verify that all other updates match it
VkShaderStageFlags refStageFlags = pLayout->stageFlags[startIndex];
for (i = startIndex; i <= endIndex; i++) {
if (pLayout->descriptorTypes[i] != actualType) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_DESCRIPTOR_TYPE_MISMATCH, "DS",
"Write descriptor update has descriptor type %s that does not match overlapping binding descriptor type of %s!",
string_VkDescriptorType(actualType), string_VkDescriptorType(pLayout->descriptorTypes[i]));
}
if (pLayout->stageFlags[i] != refStageFlags) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_DESCRIPTOR_STAGEFLAGS_MISMATCH, "DS",
"Write descriptor update has stageFlags %x that do not match overlapping binding descriptor stageFlags of %x!",
refStageFlags, pLayout->stageFlags[i]);
}
}
}
return skipCall;
}
// Determine the update type, allocate a new struct of that type, shadow the given pUpdate
// struct into the pNewNode param. Return VK_TRUE if error condition encountered and callback signals early exit.
// NOTE : Calls to this function should be wrapped in mutex
static VkBool32 shadowUpdateNode(layer_data* my_data, const VkDevice device, GENERIC_HEADER* pUpdate, GENERIC_HEADER** pNewNode)
{
VkBool32 skipCall = VK_FALSE;
VkWriteDescriptorSet* pWDS = NULL;
VkCopyDescriptorSet* pCDS = NULL;
switch (pUpdate->sType)
{
case VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET:
pWDS = new VkWriteDescriptorSet;
*pNewNode = (GENERIC_HEADER*)pWDS;
memcpy(pWDS, pUpdate, sizeof(VkWriteDescriptorSet));
switch (pWDS->descriptorType) {
case VK_DESCRIPTOR_TYPE_SAMPLER:
case VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER:
case VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE:
case VK_DESCRIPTOR_TYPE_STORAGE_IMAGE:
{
VkDescriptorImageInfo *info = new VkDescriptorImageInfo[pWDS->descriptorCount];
memcpy(info, pWDS->pImageInfo, pWDS->descriptorCount * sizeof(VkDescriptorImageInfo));
pWDS->pImageInfo = info;
}
break;
case VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER:
case VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER:
{
VkBufferView *info = new VkBufferView[pWDS->descriptorCount];
memcpy(info, pWDS->pTexelBufferView, pWDS->descriptorCount * sizeof(VkBufferView));
pWDS->pTexelBufferView = info;
}
break;
case VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER:
case VK_DESCRIPTOR_TYPE_STORAGE_BUFFER:
case VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC:
case VK_DESCRIPTOR_TYPE_STORAGE_BUFFER_DYNAMIC:
{
VkDescriptorBufferInfo *info = new VkDescriptorBufferInfo[pWDS->descriptorCount];
memcpy(info, pWDS->pBufferInfo, pWDS->descriptorCount * sizeof(VkDescriptorBufferInfo));
pWDS->pBufferInfo = info;
}
break;
default:
return VK_ERROR_VALIDATION_FAILED_EXT;
break;
}
break;
case VK_STRUCTURE_TYPE_COPY_DESCRIPTOR_SET:
pCDS = new VkCopyDescriptorSet;
*pNewNode = (GENERIC_HEADER*)pCDS;
memcpy(pCDS, pUpdate, sizeof(VkCopyDescriptorSet));
break;
default:
if (log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_INVALID_UPDATE_STRUCT, "DS",
"Unexpected UPDATE struct of type %s (value %u) in vkUpdateDescriptors() struct tree", string_VkStructureType(pUpdate->sType), pUpdate->sType))
return VK_TRUE;
}
// Make sure that pNext for the end of shadow copy is NULL
(*pNewNode)->pNext = NULL;
return skipCall;
}
// Verify that given sampler is valid
static VkBool32 validateSampler(const layer_data* my_data, const VkSampler* pSampler, const VkBool32 immutable)
{
VkBool32 skipCall = VK_FALSE;
auto sampIt = my_data->sampleMap.find(*pSampler);
if (sampIt == my_data->sampleMap.end()) {
if (!immutable) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_SAMPLER_EXT, (uint64_t) *pSampler, __LINE__, DRAWSTATE_SAMPLER_DESCRIPTOR_ERROR, "DS",
"vkUpdateDescriptorSets: Attempt to update descriptor with invalid sampler %#" PRIxLEAST64, (uint64_t) *pSampler);
} else { // immutable
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_SAMPLER_EXT, (uint64_t) *pSampler, __LINE__, DRAWSTATE_SAMPLER_DESCRIPTOR_ERROR, "DS",
"vkUpdateDescriptorSets: Attempt to update descriptor whose binding has an invalid immutable sampler %#" PRIxLEAST64, (uint64_t) *pSampler);
}
} else {
// TODO : Any further checks we want to do on the sampler?
}
return skipCall;
}
// Set the layout on the global level
void SetLayout(layer_data *my_data, ImageSubresourcePair imgpair,
const VkImageLayout &layout) {
VkImage &image = imgpair.image;
// TODO (mlentine): Maybe set format if new? Not used atm.
my_data->imageLayoutMap[imgpair].layout = layout;
// TODO (mlentine): Maybe make vector a set?
auto subresource =
std::find(my_data->imageSubresourceMap[image].begin(),
my_data->imageSubresourceMap[image].end(), imgpair);
if (subresource == my_data->imageSubresourceMap[image].end()) {
my_data->imageSubresourceMap[image].push_back(imgpair);
}
}
void SetLayout(layer_data *my_data, VkImage image,
const VkImageLayout &layout) {
ImageSubresourcePair imgpair = {image, false, VkImageSubresource()};
SetLayout(my_data, imgpair, layout);
}
void SetLayout(layer_data *my_data, VkImage image, VkImageSubresource range,
const VkImageLayout &layout) {
ImageSubresourcePair imgpair = {image, true, range};
SetLayout(my_data, imgpair, layout);
}
// Set the layout on the cmdbuf level
void SetLayout(GLOBAL_CB_NODE *pCB, VkImage image, ImageSubresourcePair imgpair,
const IMAGE_CMD_BUF_NODE &node) {
pCB->imageLayoutMap[imgpair] = node;
// TODO (mlentine): Maybe make vector a set?
auto subresource =
std::find(pCB->imageSubresourceMap[image].begin(),
pCB->imageSubresourceMap[image].end(), imgpair);
if (subresource == pCB->imageSubresourceMap[image].end()) {
pCB->imageSubresourceMap[image].push_back(imgpair);
}
}
void SetLayout(GLOBAL_CB_NODE *pCB, VkImage image, ImageSubresourcePair imgpair,
const VkImageLayout &layout) {
pCB->imageLayoutMap[imgpair].layout = layout;
// TODO (mlentine): Maybe make vector a set?
assert(std::find(pCB->imageSubresourceMap[image].begin(),
pCB->imageSubresourceMap[image].end(),
imgpair) != pCB->imageSubresourceMap[image].end());
}
void SetLayout(GLOBAL_CB_NODE *pCB, VkImage image,
const IMAGE_CMD_BUF_NODE &node) {
ImageSubresourcePair imgpair = {image, false, VkImageSubresource()};
SetLayout(pCB, image, imgpair, node);
}
void SetLayout(GLOBAL_CB_NODE *pCB, VkImage image, VkImageSubresource range,
const IMAGE_CMD_BUF_NODE &node) {
ImageSubresourcePair imgpair = {image, true, range};
SetLayout(pCB, image, imgpair, node);
}
void SetLayout(GLOBAL_CB_NODE *pCB, VkImage image,
const VkImageLayout &layout) {
ImageSubresourcePair imgpair = {image, false, VkImageSubresource()};
SetLayout(pCB, image, imgpair, layout);
}
void SetLayout(GLOBAL_CB_NODE *pCB, VkImage image, VkImageSubresource range,
const VkImageLayout &layout) {
ImageSubresourcePair imgpair = {image, true, range};
SetLayout(pCB, image, imgpair, layout);
}
void SetLayout(const layer_data *dev_data, GLOBAL_CB_NODE *pCB,
VkImageView imageView, const VkImageLayout &layout) {
auto image_view_data = dev_data->imageViewMap.find(imageView);
assert(image_view_data != dev_data->imageViewMap.end());
const VkImage &image = image_view_data->second->image;
const VkImageSubresourceRange &subRange =
image_view_data->second->subresourceRange;
// TODO: Do not iterate over every possibility - consolidate where possible
for (uint32_t j = 0; j < subRange.levelCount; j++) {
uint32_t level = subRange.baseMipLevel + j;
for (uint32_t k = 0; k < subRange.layerCount; k++) {
uint32_t layer = subRange.baseArrayLayer + k;
VkImageSubresource sub = {subRange.aspectMask, level, layer};
SetLayout(pCB, image, sub, layout);
}
}
}
// find layout(s) on the cmd buf level
bool FindLayout(const GLOBAL_CB_NODE *pCB, VkImage image,
VkImageSubresource range, IMAGE_CMD_BUF_NODE &node) {
ImageSubresourcePair imgpair = {image, true, range};
auto imgsubIt = pCB->imageLayoutMap.find(imgpair);
if (imgsubIt == pCB->imageLayoutMap.end()) {
imgpair = {image, false, VkImageSubresource()};
imgsubIt = pCB->imageLayoutMap.find(imgpair);
if (imgsubIt == pCB->imageLayoutMap.end())
return false;
}
node = imgsubIt->second;
return true;
}
// find layout(s) on the global level
bool FindLayout(const layer_data *my_data, ImageSubresourcePair imgpair,
VkImageLayout &layout) {
auto imgsubIt = my_data->imageLayoutMap.find(imgpair);
if (imgsubIt == my_data->imageLayoutMap.end()) {
imgpair = {imgpair.image, false, VkImageSubresource()};
imgsubIt = my_data->imageLayoutMap.find(imgpair);
if(imgsubIt == my_data->imageLayoutMap.end()) return false;
}
layout = imgsubIt->second.layout;
return true;
}
bool FindLayout(const layer_data *my_data, VkImage image,
VkImageSubresource range, VkImageLayout &layout) {
ImageSubresourcePair imgpair = {image, true, range};
return FindLayout(my_data, imgpair, layout);
}
bool FindLayouts(const layer_data *my_data, VkImage image,
std::vector<VkImageLayout> &layouts) {
auto sub_data = my_data->imageSubresourceMap.find(image);
if (sub_data == my_data->imageSubresourceMap.end())
return false;
auto imgIt = my_data->imageMap.find(image);
if (imgIt == my_data->imageMap.end())
return false;
bool ignoreGlobal = false;
// TODO: Make this robust for >1 aspect mask. Now it will just say ignore
// potential errors in this case.
if (sub_data->second.size() >=
(imgIt->second->arrayLayers * imgIt->second->mipLevels + 1)) {
ignoreGlobal = true;
}
for (auto imgsubpair : sub_data->second) {
if (ignoreGlobal && !imgsubpair.hasSubresource)
continue;
auto img_data = my_data->imageLayoutMap.find(imgsubpair);
if (img_data != my_data->imageLayoutMap.end()) {
layouts.push_back(img_data->second.layout);
}
}
return true;
}
// Verify that given imageView is valid
static VkBool32 validateImageView(const layer_data* my_data, const VkImageView* pImageView, const VkImageLayout imageLayout)
{
VkBool32 skipCall = VK_FALSE;
auto ivIt = my_data->imageViewMap.find(*pImageView);
if (ivIt == my_data->imageViewMap.end()) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_IMAGE_VIEW_EXT, (uint64_t) *pImageView, __LINE__, DRAWSTATE_IMAGEVIEW_DESCRIPTOR_ERROR, "DS",
"vkUpdateDescriptorSets: Attempt to update descriptor with invalid imageView %#" PRIxLEAST64, (uint64_t) *pImageView);
} else {
// Validate that imageLayout is compatible with aspectMask and image format
VkImageAspectFlags aspectMask = ivIt->second->subresourceRange.aspectMask;
VkImage image = ivIt->second->image;
// TODO : Check here in case we have a bad image
auto imgIt = my_data->imageMap.find(image);
if (imgIt == my_data->imageMap.end()) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_IMAGE_EXT, (uint64_t) image, __LINE__, DRAWSTATE_IMAGEVIEW_DESCRIPTOR_ERROR, "DS",
"vkUpdateDescriptorSets: Attempt to update descriptor with invalid image %#" PRIxLEAST64 " in imageView %#" PRIxLEAST64, (uint64_t) image, (uint64_t) *pImageView);
} else {
VkFormat format = (*imgIt).second->format;
VkBool32 ds = vk_format_is_depth_or_stencil(format);
switch (imageLayout) {
case VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL:
// Only Color bit must be set
if ((aspectMask & VK_IMAGE_ASPECT_COLOR_BIT) != VK_IMAGE_ASPECT_COLOR_BIT) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_IMAGE_VIEW_EXT, (uint64_t) *pImageView, __LINE__,
DRAWSTATE_INVALID_IMAGE_ASPECT, "DS", "vkUpdateDescriptorSets: Updating descriptor with layout VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL and imageView %#" PRIxLEAST64 ""
" that does not have VK_IMAGE_ASPECT_COLOR_BIT set.", (uint64_t) *pImageView);
}
// format must NOT be DS
if (ds) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_IMAGE_VIEW_EXT, (uint64_t) *pImageView, __LINE__,
DRAWSTATE_IMAGEVIEW_DESCRIPTOR_ERROR, "DS", "vkUpdateDescriptorSets: Updating descriptor with layout VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL and imageView %#" PRIxLEAST64 ""
" but the image format is %s which is not a color format.", (uint64_t) *pImageView, string_VkFormat(format));
}
break;
case VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL:
case VK_IMAGE_LAYOUT_DEPTH_STENCIL_READ_ONLY_OPTIMAL:
// Depth or stencil bit must be set, but both must NOT be set
if (aspectMask & VK_IMAGE_ASPECT_DEPTH_BIT) {
if (aspectMask & VK_IMAGE_ASPECT_STENCIL_BIT) {
// both must NOT be set
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_IMAGE_VIEW_EXT, (uint64_t) *pImageView, __LINE__,
DRAWSTATE_INVALID_IMAGE_ASPECT, "DS", "vkUpdateDescriptorSets: Updating descriptor with imageView %#" PRIxLEAST64 ""
" that has both STENCIL and DEPTH aspects set", (uint64_t) *pImageView);
}
} else if (!(aspectMask & VK_IMAGE_ASPECT_STENCIL_BIT)) {
// Neither were set
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_IMAGE_VIEW_EXT, (uint64_t) *pImageView, __LINE__,
DRAWSTATE_INVALID_IMAGE_ASPECT, "DS", "vkUpdateDescriptorSets: Updating descriptor with layout %s and imageView %#" PRIxLEAST64 ""
" that does not have STENCIL or DEPTH aspect set.", string_VkImageLayout(imageLayout), (uint64_t) *pImageView);
}
// format must be DS
if (!ds) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_IMAGE_VIEW_EXT, (uint64_t) *pImageView, __LINE__,
DRAWSTATE_IMAGEVIEW_DESCRIPTOR_ERROR, "DS", "vkUpdateDescriptorSets: Updating descriptor with layout %s and imageView %#" PRIxLEAST64 ""
" but the image format is %s which is not a depth/stencil format.", string_VkImageLayout(imageLayout), (uint64_t) *pImageView, string_VkFormat(format));
}
break;
default:
// anything to check for other layouts?
break;
}
}
}
return skipCall;
}
// Verify that given bufferView is valid
static VkBool32 validateBufferView(const layer_data* my_data, const VkBufferView* pBufferView)
{
VkBool32 skipCall = VK_FALSE;
auto sampIt = my_data->bufferViewMap.find(*pBufferView);
if (sampIt == my_data->bufferViewMap.end()) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_BUFFER_VIEW_EXT, (uint64_t) *pBufferView, __LINE__, DRAWSTATE_BUFFERVIEW_DESCRIPTOR_ERROR, "DS",
"vkUpdateDescriptorSets: Attempt to update descriptor with invalid bufferView %#" PRIxLEAST64, (uint64_t) *pBufferView);
} else {
// TODO : Any further checks we want to do on the bufferView?
}
return skipCall;
}
// Verify that given bufferInfo is valid
static VkBool32 validateBufferInfo(const layer_data* my_data, const VkDescriptorBufferInfo* pBufferInfo)
{
VkBool32 skipCall = VK_FALSE;
auto sampIt = my_data->bufferMap.find(pBufferInfo->buffer);
if (sampIt == my_data->bufferMap.end()) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_BUFFER_EXT, (uint64_t) pBufferInfo->buffer, __LINE__, DRAWSTATE_BUFFERINFO_DESCRIPTOR_ERROR, "DS",
"vkUpdateDescriptorSets: Attempt to update descriptor where bufferInfo has invalid buffer %#" PRIxLEAST64, (uint64_t) pBufferInfo->buffer);
} else {
// TODO : Any further checks we want to do on the bufferView?
}
return skipCall;
}
static VkBool32 validateUpdateContents(const layer_data* my_data, const VkWriteDescriptorSet *pWDS, const VkDescriptorSetLayoutBinding* pLayoutBinding)
{
VkBool32 skipCall = VK_FALSE;
// First verify that for the given Descriptor type, the correct DescriptorInfo data is supplied
const VkSampler* pSampler = NULL;
VkBool32 immutable = VK_FALSE;
uint32_t i = 0;
// For given update type, verify that update contents are correct
switch (pWDS->descriptorType) {
case VK_DESCRIPTOR_TYPE_SAMPLER:
for (i=0; i<pWDS->descriptorCount; ++i) {
skipCall |= validateSampler(my_data, &(pWDS->pImageInfo[i].sampler), immutable);
}
break;
case VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER:
for (i=0; i<pWDS->descriptorCount; ++i) {
if (NULL == pLayoutBinding->pImmutableSamplers) {
pSampler = &(pWDS->pImageInfo[i].sampler);
if (immutable) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_SAMPLER_EXT, (uint64_t) *pSampler, __LINE__, DRAWSTATE_INCONSISTENT_IMMUTABLE_SAMPLER_UPDATE, "DS",
"vkUpdateDescriptorSets: Update #%u is not an immutable sampler %#" PRIxLEAST64 ", but previous update(s) from this "
"VkWriteDescriptorSet struct used an immutable sampler. All updates from a single struct must either "
"use immutable or non-immutable samplers.", i, (uint64_t) *pSampler);
}
} else {
if (i>0 && !immutable) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_SAMPLER_EXT, (uint64_t) *pSampler, __LINE__, DRAWSTATE_INCONSISTENT_IMMUTABLE_SAMPLER_UPDATE, "DS",
"vkUpdateDescriptorSets: Update #%u is an immutable sampler, but previous update(s) from this "
"VkWriteDescriptorSet struct used a non-immutable sampler. All updates from a single struct must either "
"use immutable or non-immutable samplers.", i);
}
immutable = VK_TRUE;
pSampler = &(pLayoutBinding->pImmutableSamplers[i]);
}
skipCall |= validateSampler(my_data, pSampler, immutable);
}
// Intentionally fall through here to also validate image stuff
case VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE:
case VK_DESCRIPTOR_TYPE_STORAGE_IMAGE:
case VK_DESCRIPTOR_TYPE_INPUT_ATTACHMENT:
for (i=0; i<pWDS->descriptorCount; ++i) {
skipCall |= validateImageView(my_data, &(pWDS->pImageInfo[i].imageView), pWDS->pImageInfo[i].imageLayout);
}
break;
case VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER:
case VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER:
for (i=0; i<pWDS->descriptorCount; ++i) {
skipCall |= validateBufferView(my_data, &(pWDS->pTexelBufferView[i]));
}
break;
case VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER:
case VK_DESCRIPTOR_TYPE_STORAGE_BUFFER:
case VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC:
case VK_DESCRIPTOR_TYPE_STORAGE_BUFFER_DYNAMIC:
for (i=0; i<pWDS->descriptorCount; ++i) {
skipCall |= validateBufferInfo(my_data, &(pWDS->pBufferInfo[i]));
}
break;
default:
break;
}
return skipCall;
}
// Validate that given set is valid and that it's not being used by an in-flight CmdBuffer
// func_str is the name of the calling function
// Return VK_FALSE if no errors occur
// Return VK_TRUE if validation error occurs and callback returns VK_TRUE (to skip upcoming API call down the chain)
VkBool32 validateIdleDescriptorSet(const layer_data* my_data, VkDescriptorSet set, std::string func_str) {
VkBool32 skip_call = VK_FALSE;
auto set_node = my_data->setMap.find(set);
if (set_node == my_data->setMap.end()) {
skip_call |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT, (uint64_t)(set), __LINE__, DRAWSTATE_DOUBLE_DESTROY, "DS",
"Cannot call %s() on descriptor set %" PRIxLEAST64 " that has not been allocated.", func_str.c_str(), (uint64_t)(set));
} else {
if (set_node->second->in_use.load()) {
skip_call |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT, (uint64_t)(set), __LINE__, DRAWSTATE_OBJECT_INUSE, "DS",
"Cannot call %s() on descriptor set %" PRIxLEAST64 " that is in use by a command buffer.", func_str.c_str(), (uint64_t)(set));
}
}
return skip_call;
}
static void invalidateBoundCmdBuffers(layer_data* dev_data, const SET_NODE* pSet)
{
// Flag any CBs this set is bound to as INVALID
for (auto cb : pSet->boundCmdBuffers) {
auto cb_node = dev_data->commandBufferMap.find(cb);
if (cb_node != dev_data->commandBufferMap.end()) {
cb_node->second->state = CB_INVALID;
}
}
}
// update DS mappings based on write and copy update arrays
static VkBool32 dsUpdate(layer_data* my_data, VkDevice device, uint32_t descriptorWriteCount, const VkWriteDescriptorSet* pWDS, uint32_t descriptorCopyCount, const VkCopyDescriptorSet* pCDS)
{
VkBool32 skipCall = VK_FALSE;
LAYOUT_NODE* pLayout = NULL;
VkDescriptorSetLayoutCreateInfo* pLayoutCI = NULL;
// Validate Write updates
uint32_t i = 0;
for (i=0; i < descriptorWriteCount; i++) {
VkDescriptorSet ds = pWDS[i].dstSet;
SET_NODE* pSet = my_data->setMap[ds];
// Set being updated cannot be in-flight
if ((skipCall = validateIdleDescriptorSet(my_data, ds, "VkUpdateDescriptorSets")) == VK_TRUE)
return skipCall;
// If set is bound to any cmdBuffers, mark them invalid
invalidateBoundCmdBuffers(my_data, pSet);
GENERIC_HEADER* pUpdate = (GENERIC_HEADER*) &pWDS[i];
pLayout = pSet->pLayout;
// First verify valid update struct
if ((skipCall = validUpdateStruct(my_data, device, pUpdate)) == VK_TRUE) {
break;
}
uint32_t binding = 0, endIndex = 0;
binding = pWDS[i].dstBinding;
auto bindingToIndex = pLayout->bindingToIndexMap.find(binding);
// Make sure that layout being updated has the binding being updated
if (bindingToIndex == pLayout->bindingToIndexMap.end()) {
skipCall |= log_msg(
my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT, (uint64_t)(ds),
__LINE__, DRAWSTATE_INVALID_UPDATE_INDEX, "DS",
"Descriptor Set %" PRIu64 " does not have binding to match "
"update binding %u for update type "
"%s!",
(uint64_t)(ds), binding,
string_VkStructureType(pUpdate->sType));
} else {
// Next verify that update falls within size of given binding
endIndex = getUpdateEndIndex(my_data, device, pLayout, binding, pWDS[i].dstArrayElement, pUpdate);
if (getBindingEndIndex(pLayout, binding) < endIndex) {
pLayoutCI = &pLayout->createInfo;
string DSstr = vk_print_vkdescriptorsetlayoutcreateinfo(pLayoutCI, "{DS} ");
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT, (uint64_t)(ds), __LINE__, DRAWSTATE_DESCRIPTOR_UPDATE_OUT_OF_BOUNDS, "DS",
"Descriptor update type of %s is out of bounds for matching binding %u in Layout w/ CI:\n%s!", string_VkStructureType(pUpdate->sType), binding, DSstr.c_str());
} else { // TODO : should we skip update on a type mismatch or force it?
uint32_t startIndex;
startIndex =
getUpdateStartIndex(my_data, device, pLayout, binding,
pWDS[i].dstArrayElement, pUpdate);
// Layout bindings match w/ update, now verify that update type
// & stageFlags are the same for entire update
if ((skipCall = validateUpdateConsistency(
my_data, device, pLayout, pUpdate, startIndex,
endIndex)) == VK_FALSE) {
// The update is within bounds and consistent, but need to
// make sure contents make sense as well
if ((skipCall = validateUpdateContents(
my_data, &pWDS[i],
&pLayout->createInfo.pBindings[bindingToIndex->second])) ==
VK_FALSE) {
// Update is good. Save the update info
// Create new update struct for this set's shadow copy
GENERIC_HEADER* pNewNode = NULL;
skipCall |= shadowUpdateNode(my_data, device, pUpdate, &pNewNode);
if (NULL == pNewNode) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT, (uint64_t)(ds), __LINE__, DRAWSTATE_OUT_OF_MEMORY, "DS",
"Out of memory while attempting to allocate UPDATE struct in vkUpdateDescriptors()");
} else {
// Insert shadow node into LL of updates for this set
pNewNode->pNext = pSet->pUpdateStructs;
pSet->pUpdateStructs = pNewNode;
// Now update appropriate descriptor(s) to point to new Update node
for (uint32_t j = startIndex; j <= endIndex; j++) {
assert(j<pSet->descriptorCount);
pSet->ppDescriptors[j] = pNewNode;
}
}
}
}
}
}
}
// Now validate copy updates
for (i=0; i < descriptorCopyCount; ++i) {
SET_NODE *pSrcSet = NULL, *pDstSet = NULL;
LAYOUT_NODE *pSrcLayout = NULL, *pDstLayout = NULL;
uint32_t srcStartIndex = 0, srcEndIndex = 0, dstStartIndex = 0, dstEndIndex = 0;
// For each copy make sure that update falls within given layout and that types match
pSrcSet = my_data->setMap[pCDS[i].srcSet];
pDstSet = my_data->setMap[pCDS[i].dstSet];
// Set being updated cannot be in-flight
if ((skipCall = validateIdleDescriptorSet(my_data, pDstSet->set, "VkUpdateDescriptorSets")) == VK_TRUE)
return skipCall;
invalidateBoundCmdBuffers(my_data, pDstSet);
pSrcLayout = pSrcSet->pLayout;
pDstLayout = pDstSet->pLayout;
// Validate that src binding is valid for src set layout
if (pSrcLayout->bindingToIndexMap.find(pCDS[i].srcBinding) ==
pSrcLayout->bindingToIndexMap.end()) {
skipCall |=
log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT,
(uint64_t)pSrcSet->set, __LINE__,
DRAWSTATE_INVALID_UPDATE_INDEX,
"DS", "Copy descriptor update %u has srcBinding %u "
"which is out of bounds for underlying SetLayout "
"%#" PRIxLEAST64 " which only has bindings 0-%u.",
i, pCDS[i].srcBinding, (uint64_t)pSrcLayout->layout,
pSrcLayout->createInfo.bindingCount - 1);
} else if (pDstLayout->bindingToIndexMap.find(pCDS[i].dstBinding) ==
pDstLayout->bindingToIndexMap.end()) {
skipCall |=
log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT,
(uint64_t)pDstSet->set, __LINE__,
DRAWSTATE_INVALID_UPDATE_INDEX,
"DS", "Copy descriptor update %u has dstBinding %u "
"which is out of bounds for underlying SetLayout "
"%#" PRIxLEAST64 " which only has bindings 0-%u.",
i, pCDS[i].dstBinding, (uint64_t)pDstLayout->layout,
pDstLayout->createInfo.bindingCount - 1);
} else {
// Proceed with validation. Bindings are ok, but make sure update is within bounds of given layout
srcEndIndex = getUpdateEndIndex(my_data, device, pSrcLayout, pCDS[i].srcBinding, pCDS[i].srcArrayElement, (const GENERIC_HEADER*)&(pCDS[i]));
dstEndIndex = getUpdateEndIndex(my_data, device, pDstLayout, pCDS[i].dstBinding, pCDS[i].dstArrayElement, (const GENERIC_HEADER*)&(pCDS[i]));
if (getBindingEndIndex(pSrcLayout, pCDS[i].srcBinding) < srcEndIndex) {
pLayoutCI = &pSrcLayout->createInfo;
string DSstr = vk_print_vkdescriptorsetlayoutcreateinfo(pLayoutCI, "{DS} ");
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT, (uint64_t) pSrcSet->set, __LINE__, DRAWSTATE_DESCRIPTOR_UPDATE_OUT_OF_BOUNDS, "DS",
"Copy descriptor src update is out of bounds for matching binding %u in Layout w/ CI:\n%s!", pCDS[i].srcBinding, DSstr.c_str());
} else if (getBindingEndIndex(pDstLayout, pCDS[i].dstBinding) < dstEndIndex) {
pLayoutCI = &pDstLayout->createInfo;
string DSstr = vk_print_vkdescriptorsetlayoutcreateinfo(pLayoutCI, "{DS} ");
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT, (uint64_t) pDstSet->set, __LINE__, DRAWSTATE_DESCRIPTOR_UPDATE_OUT_OF_BOUNDS, "DS",
"Copy descriptor dest update is out of bounds for matching binding %u in Layout w/ CI:\n%s!", pCDS[i].dstBinding, DSstr.c_str());
} else {
srcStartIndex = getUpdateStartIndex(my_data, device, pSrcLayout, pCDS[i].srcBinding, pCDS[i].srcArrayElement, (const GENERIC_HEADER*)&(pCDS[i]));
dstStartIndex = getUpdateStartIndex(my_data, device, pDstLayout, pCDS[i].dstBinding, pCDS[i].dstArrayElement, (const GENERIC_HEADER*)&(pCDS[i]));
for (uint32_t j=0; j<pCDS[i].descriptorCount; ++j) {
// For copy just make sure that the types match and then perform the update
if (pSrcLayout->descriptorTypes[srcStartIndex+j] != pDstLayout->descriptorTypes[dstStartIndex+j]) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_DESCRIPTOR_TYPE_MISMATCH, "DS",
"Copy descriptor update index %u, update count #%u, has src update descriptor type %s that does not match overlapping dest descriptor type of %s!",
i, j+1, string_VkDescriptorType(pSrcLayout->descriptorTypes[srcStartIndex+j]), string_VkDescriptorType(pDstLayout->descriptorTypes[dstStartIndex+j]));
} else {
// point dst descriptor at corresponding src descriptor
// TODO : This may be a hole. I believe copy should be its own copy,
// otherwise a subsequent write update to src will incorrectly affect the copy
pDstSet->ppDescriptors[j+dstStartIndex] = pSrcSet->ppDescriptors[j+srcStartIndex];
}
}
}
}
}
return skipCall;
}
// Verify that given pool has descriptors that are being requested for allocation
static VkBool32 validate_descriptor_availability_in_pool(layer_data* dev_data, DESCRIPTOR_POOL_NODE* pPoolNode, uint32_t count, const VkDescriptorSetLayout* pSetLayouts)
{
VkBool32 skipCall = VK_FALSE;
uint32_t i = 0, j = 0;
for (i=0; i<count; ++i) {
LAYOUT_NODE* pLayout = getLayoutNode(dev_data, pSetLayouts[i]);
if (NULL == pLayout) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_LAYOUT_EXT, (uint64_t) pSetLayouts[i], __LINE__, DRAWSTATE_INVALID_LAYOUT, "DS",
"Unable to find set layout node for layout %#" PRIxLEAST64 " specified in vkAllocateDescriptorSets() call", (uint64_t) pSetLayouts[i]);
} else {
uint32_t typeIndex = 0, poolSizeCount = 0;
for (j=0; j<pLayout->createInfo.bindingCount; ++j) {
typeIndex = static_cast<uint32_t>(pLayout->createInfo.pBindings[j].descriptorType);
poolSizeCount = pLayout->createInfo.pBindings[j].descriptorCount;
if (poolSizeCount > pPoolNode->availableDescriptorTypeCount[typeIndex]) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_LAYOUT_EXT, (uint64_t) pLayout->layout, __LINE__, DRAWSTATE_DESCRIPTOR_POOL_EMPTY, "DS",
"Unable to allocate %u descriptors of type %s from pool %#" PRIxLEAST64 ". This pool only has %u descriptors of this type remaining.",
poolSizeCount, string_VkDescriptorType(pLayout->createInfo.pBindings[j].descriptorType), (uint64_t) pPoolNode->pool, pPoolNode->availableDescriptorTypeCount[typeIndex]);
} else { // Decrement available descriptors of this type
pPoolNode->availableDescriptorTypeCount[typeIndex] -= poolSizeCount;
}
}
}
}
return skipCall;
}
// Free the shadowed update node for this Set
// NOTE : Calls to this function should be wrapped in mutex
static void freeShadowUpdateTree(SET_NODE* pSet)
{
GENERIC_HEADER* pShadowUpdate = pSet->pUpdateStructs;
pSet->pUpdateStructs = NULL;
GENERIC_HEADER* pFreeUpdate = pShadowUpdate;
// Clear the descriptor mappings as they will now be invalid
memset(pSet->ppDescriptors, 0, pSet->descriptorCount*sizeof(GENERIC_HEADER*));
while(pShadowUpdate) {
pFreeUpdate = pShadowUpdate;
pShadowUpdate = (GENERIC_HEADER*)pShadowUpdate->pNext;
VkWriteDescriptorSet * pWDS = NULL;
switch (pFreeUpdate->sType)
{
case VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET:
pWDS = (VkWriteDescriptorSet*)pFreeUpdate;
switch (pWDS->descriptorType) {
case VK_DESCRIPTOR_TYPE_SAMPLER:
case VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER:
case VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE:
case VK_DESCRIPTOR_TYPE_STORAGE_IMAGE:
{
delete[] pWDS->pImageInfo;
}
break;
case VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER:
case VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER:
{
delete[] pWDS->pTexelBufferView;
}
break;
case VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER:
case VK_DESCRIPTOR_TYPE_STORAGE_BUFFER:
case VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC:
case VK_DESCRIPTOR_TYPE_STORAGE_BUFFER_DYNAMIC:
{
delete[] pWDS->pBufferInfo;
}
break;
default:
break;
}
break;
case VK_STRUCTURE_TYPE_COPY_DESCRIPTOR_SET:
break;
default:
assert(0);
break;
}
delete pFreeUpdate;
}
}
// Free all DS Pools including their Sets & related sub-structs
// NOTE : Calls to this function should be wrapped in mutex
static void deletePools(layer_data* my_data)
{
if (my_data->descriptorPoolMap.size() <= 0)
return;
for (auto ii=my_data->descriptorPoolMap.begin(); ii!=my_data->descriptorPoolMap.end(); ++ii) {
SET_NODE* pSet = (*ii).second->pSets;
SET_NODE* pFreeSet = pSet;
while (pSet) {
pFreeSet = pSet;
pSet = pSet->pNext;
// Freeing layouts handled in deleteLayouts() function
// Free Update shadow struct tree
freeShadowUpdateTree(pFreeSet);
if (pFreeSet->ppDescriptors) {
delete[] pFreeSet->ppDescriptors;
}
delete pFreeSet;
}
delete (*ii).second;
}
my_data->descriptorPoolMap.clear();
}
// WARN : Once deleteLayouts() called, any layout ptrs in Pool/Set data structure will be invalid
// NOTE : Calls to this function should be wrapped in mutex
static void deleteLayouts(layer_data* my_data)
{
if (my_data->descriptorSetLayoutMap.size() <= 0)
return;
for (auto ii=my_data->descriptorSetLayoutMap.begin(); ii!=my_data->descriptorSetLayoutMap.end(); ++ii) {
LAYOUT_NODE* pLayout = (*ii).second;
if (pLayout->createInfo.pBindings) {
for (uint32_t i=0; i<pLayout->createInfo.bindingCount; i++) {
if (pLayout->createInfo.pBindings[i].pImmutableSamplers)
delete[] pLayout->createInfo.pBindings[i].pImmutableSamplers;
}
delete[] pLayout->createInfo.pBindings;
}
delete pLayout;
}
my_data->descriptorSetLayoutMap.clear();
}
// Currently clearing a set is removing all previous updates to that set
// TODO : Validate if this is correct clearing behavior
static void clearDescriptorSet(layer_data* my_data, VkDescriptorSet set)
{
SET_NODE* pSet = getSetNode(my_data, set);
if (!pSet) {
// TODO : Return error
} else {
freeShadowUpdateTree(pSet);
}
}
static void clearDescriptorPool(layer_data* my_data, const VkDevice device, const VkDescriptorPool pool, VkDescriptorPoolResetFlags flags)
{
DESCRIPTOR_POOL_NODE* pPool = getPoolNode(my_data, pool);
if (!pPool) {
log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_POOL_EXT, (uint64_t) pool, __LINE__, DRAWSTATE_INVALID_POOL, "DS",
"Unable to find pool node for pool %#" PRIxLEAST64 " specified in vkResetDescriptorPool() call", (uint64_t) pool);
} else {
// TODO: validate flags
// For every set off of this pool, clear it
SET_NODE* pSet = pPool->pSets;
while (pSet) {
clearDescriptorSet(my_data, pSet->set);
pSet = pSet->pNext;
}
// Reset available count to max count for this pool
for (uint32_t i=0; i<pPool->availableDescriptorTypeCount.size(); ++i) {
pPool->availableDescriptorTypeCount[i] = pPool->maxDescriptorTypeCount[i];
}
}
}
// For given CB object, fetch associated CB Node from map
static GLOBAL_CB_NODE* getCBNode(layer_data* my_data, const VkCommandBuffer cb)
{
if (my_data->commandBufferMap.count(cb) == 0) {
// TODO : How to pass cb as srcObj here?
log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, 0, __LINE__, DRAWSTATE_INVALID_COMMAND_BUFFER, "DS",
"Attempt to use CommandBuffer %#" PRIxLEAST64 " that doesn't exist!", (uint64_t)(cb));
return NULL;
}
return my_data->commandBufferMap[cb];
}
// Free all CB Nodes
// NOTE : Calls to this function should be wrapped in mutex
static void deleteCommandBuffers(layer_data* my_data)
{
if (my_data->commandBufferMap.size() <= 0) {
return;
}
for (auto ii=my_data->commandBufferMap.begin(); ii!=my_data->commandBufferMap.end(); ++ii) {
delete (*ii).second;
}
my_data->commandBufferMap.clear();
}
static VkBool32 report_error_no_cb_begin(const layer_data* dev_data, const VkCommandBuffer cb, const char* caller_name)
{
return log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT,
(uint64_t)cb, __LINE__, DRAWSTATE_NO_BEGIN_COMMAND_BUFFER, "DS",
"You must call vkBeginCommandBuffer() before this call to %s", caller_name);
}
VkBool32 validateCmdsInCmdBuffer(const layer_data* dev_data, const GLOBAL_CB_NODE* pCB, const CMD_TYPE cmd_type) {
if (!pCB->activeRenderPass) return VK_FALSE;
VkBool32 skip_call = VK_FALSE;
if (pCB->activeSubpassContents == VK_SUBPASS_CONTENTS_SECONDARY_COMMAND_BUFFERS && cmd_type != CMD_EXECUTECOMMANDS) {
skip_call |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__,
DRAWSTATE_INVALID_COMMAND_BUFFER, "DS", "Commands cannot be called in a subpass using secondary command buffers.");
} else if (pCB->activeSubpassContents == VK_SUBPASS_CONTENTS_INLINE && cmd_type == CMD_EXECUTECOMMANDS) {
skip_call |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__,
DRAWSTATE_INVALID_COMMAND_BUFFER, "DS", "vkCmdExecuteCommands() cannot be called in a subpass using inline commands.");
}
return skip_call;
}
static bool checkGraphicsBit(const layer_data* my_data, VkQueueFlags flags, const char* name) {
if (!(flags & VK_QUEUE_GRAPHICS_BIT))
return log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__,
DRAWSTATE_INVALID_COMMAND_BUFFER, "DS", "Cannot call %s on a command buffer allocated from a pool without graphics capabilities.", name);
return false;
}
static bool checkComputeBit(const layer_data* my_data, VkQueueFlags flags, const char* name) {
if (!(flags & VK_QUEUE_COMPUTE_BIT))
return log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__,
DRAWSTATE_INVALID_COMMAND_BUFFER, "DS", "Cannot call %s on a command buffer allocated from a pool without compute capabilities.", name);
return false;
}
static bool checkGraphicsOrComputeBit(const layer_data* my_data, VkQueueFlags flags, const char* name) {
if (!((flags & VK_QUEUE_GRAPHICS_BIT) || (flags & VK_QUEUE_COMPUTE_BIT)))
return log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__,
DRAWSTATE_INVALID_COMMAND_BUFFER, "DS", "Cannot call %s on a command buffer allocated from a pool without graphics capabilities.", name);
return false;
}
// Add specified CMD to the CmdBuffer in given pCB, flagging errors if CB is not
// in the recording state or if there's an issue with the Cmd ordering
static VkBool32 addCmd(const layer_data* my_data, GLOBAL_CB_NODE* pCB, const CMD_TYPE cmd, const char* caller_name)
{
VkBool32 skipCall = VK_FALSE;
auto pool_data = my_data->commandPoolMap.find(pCB->createInfo.commandPool);
if (pool_data != my_data->commandPoolMap.end()) {
VkQueueFlags flags = my_data->physDevProperties.queue_family_properties[pool_data->second.queueFamilyIndex].queueFlags;
switch (cmd)
{
case CMD_BINDPIPELINE:
case CMD_BINDPIPELINEDELTA:
case CMD_BINDDESCRIPTORSETS:
case CMD_FILLBUFFER:
case CMD_CLEARCOLORIMAGE:
case CMD_SETEVENT:
case CMD_RESETEVENT:
case CMD_WAITEVENTS:
case CMD_BEGINQUERY:
case CMD_ENDQUERY:
case CMD_RESETQUERYPOOL:
case CMD_COPYQUERYPOOLRESULTS:
case CMD_WRITETIMESTAMP:
skipCall |= checkGraphicsOrComputeBit(my_data, flags, cmdTypeToString(cmd).c_str());
break;
case CMD_SETVIEWPORTSTATE:
case CMD_SETSCISSORSTATE:
case CMD_SETLINEWIDTHSTATE:
case CMD_SETDEPTHBIASSTATE:
case CMD_SETBLENDSTATE:
case CMD_SETDEPTHBOUNDSSTATE:
case CMD_SETSTENCILREADMASKSTATE:
case CMD_SETSTENCILWRITEMASKSTATE:
case CMD_SETSTENCILREFERENCESTATE:
case CMD_BINDINDEXBUFFER:
case CMD_BINDVERTEXBUFFER:
case CMD_DRAW:
case CMD_DRAWINDEXED:
case CMD_DRAWINDIRECT:
case CMD_DRAWINDEXEDINDIRECT:
case CMD_BLITIMAGE:
case CMD_CLEARATTACHMENTS:
case CMD_CLEARDEPTHSTENCILIMAGE:
case CMD_RESOLVEIMAGE:
case CMD_BEGINRENDERPASS:
case CMD_NEXTSUBPASS:
case CMD_ENDRENDERPASS:
skipCall |= checkGraphicsBit(my_data, flags, cmdTypeToString(cmd).c_str());
break;
case CMD_DISPATCH:
case CMD_DISPATCHINDIRECT:
skipCall |= checkComputeBit(my_data, flags, cmdTypeToString(cmd).c_str());
break;
case CMD_COPYBUFFER:
case CMD_COPYIMAGE:
case CMD_COPYBUFFERTOIMAGE:
case CMD_COPYIMAGETOBUFFER:
case CMD_CLONEIMAGEDATA:
case CMD_UPDATEBUFFER:
case CMD_PIPELINEBARRIER:
case CMD_EXECUTECOMMANDS:
break;
default:
break;
}
}
if (pCB->state != CB_RECORDING) {
skipCall |= report_error_no_cb_begin(my_data, pCB->commandBuffer, caller_name);
skipCall |= validateCmdsInCmdBuffer(my_data, pCB, cmd);
CMD_NODE cmdNode = {};
// init cmd node and append to end of cmd LL
cmdNode.cmdNumber = ++pCB->numCmds;
cmdNode.type = cmd;
pCB->cmds.push_back(cmdNode);
}
return skipCall;
}
// Reset the command buffer state
// Maintain the createInfo and set state to CB_NEW, but clear all other state
static void resetCB(layer_data* my_data, const VkCommandBuffer cb)
{
GLOBAL_CB_NODE* pCB = my_data->commandBufferMap[cb];
if (pCB) {
pCB->cmds.clear();
// Reset CB state (note that createInfo is not cleared)
pCB->commandBuffer = cb;
memset(&pCB->beginInfo, 0, sizeof(VkCommandBufferBeginInfo));
memset(&pCB->inheritanceInfo, 0, sizeof(VkCommandBufferInheritanceInfo));
pCB->fence = 0;
pCB->numCmds = 0;
memset(pCB->drawCount, 0, NUM_DRAW_TYPES * sizeof(uint64_t));
pCB->state = CB_NEW;
pCB->submitCount = 0;
pCB->status = 0;
pCB->lastBoundPipeline = 0;
pCB->lastVtxBinding = 0;
pCB->boundVtxBuffers.clear();
pCB->viewports.clear();
pCB->scissors.clear();
pCB->lineWidth = 0;
pCB->depthBiasConstantFactor = 0;
pCB->depthBiasClamp = 0;
pCB->depthBiasSlopeFactor = 0;
memset(pCB->blendConstants, 0, 4 * sizeof(float));
pCB->minDepthBounds = 0;
pCB->maxDepthBounds = 0;
memset(&pCB->front, 0, sizeof(stencil_data));
memset(&pCB->back, 0, sizeof(stencil_data));
pCB->lastBoundDescriptorSet = 0;
pCB->lastBoundPipelineLayout = 0;
memset(&pCB->activeRenderPassBeginInfo, 0, sizeof(pCB->activeRenderPassBeginInfo));
pCB->activeRenderPass = 0;
pCB->activeSubpassContents = VK_SUBPASS_CONTENTS_INLINE;
pCB->activeSubpass = 0;
pCB->framebuffer = 0;
// Before clearing uniqueBoundSets, remove this CB off of its boundCBs
for (auto set : pCB->uniqueBoundSets) {
auto set_node = my_data->setMap.find(set);
if (set_node != my_data->setMap.end()) {
set_node->second->boundCmdBuffers.erase(pCB->commandBuffer);
}
}
pCB->uniqueBoundSets.clear();
pCB->destroyedSets.clear();
pCB->updatedSets.clear();
pCB->boundDescriptorSets.clear();
pCB->waitedEvents.clear();
pCB->semaphores.clear();
pCB->events.clear();
pCB->waitedEventsBeforeQueryReset.clear();
pCB->queryToStateMap.clear();
pCB->activeQueries.clear();
pCB->startedQueries.clear();
pCB->imageLayoutMap.clear();
pCB->eventToStageMap.clear();
pCB->drawData.clear();
pCB->currentDrawData.buffers.clear();
pCB->primaryCommandBuffer = VK_NULL_HANDLE;
pCB->secondaryCommandBuffers.clear();
pCB->dynamicOffsets.clear();
}
}
// Set PSO-related status bits for CB, including dynamic state set via PSO
static void set_cb_pso_status(GLOBAL_CB_NODE* pCB, const PIPELINE_NODE* pPipe)
{
for (uint32_t i = 0; i < pPipe->cbStateCI.attachmentCount; i++) {
if (0 != pPipe->pAttachments[i].colorWriteMask) {
pCB->status |= CBSTATUS_COLOR_BLEND_WRITE_ENABLE;
}
}
if (pPipe->dsStateCI.depthWriteEnable) {
pCB->status |= CBSTATUS_DEPTH_WRITE_ENABLE;
}
if (pPipe->dsStateCI.stencilTestEnable) {
pCB->status |= CBSTATUS_STENCIL_TEST_ENABLE;
}
// Account for any dynamic state not set via this PSO
if (!pPipe->dynStateCI.dynamicStateCount) { // All state is static
pCB->status = CBSTATUS_ALL;
} else {
// First consider all state on
// Then unset any state that's noted as dynamic in PSO
// Finally OR that into CB statemask
CBStatusFlags psoDynStateMask = CBSTATUS_ALL;
for (uint32_t i=0; i < pPipe->dynStateCI.dynamicStateCount; i++) {
switch (pPipe->dynStateCI.pDynamicStates[i]) {
case VK_DYNAMIC_STATE_VIEWPORT:
psoDynStateMask &= ~CBSTATUS_VIEWPORT_SET;
break;
case VK_DYNAMIC_STATE_SCISSOR:
psoDynStateMask &= ~CBSTATUS_SCISSOR_SET;
break;
case VK_DYNAMIC_STATE_LINE_WIDTH:
psoDynStateMask &= ~CBSTATUS_LINE_WIDTH_SET;
break;
case VK_DYNAMIC_STATE_DEPTH_BIAS:
psoDynStateMask &= ~CBSTATUS_DEPTH_BIAS_SET;
break;
case VK_DYNAMIC_STATE_BLEND_CONSTANTS:
psoDynStateMask &= ~CBSTATUS_BLEND_SET;
break;
case VK_DYNAMIC_STATE_DEPTH_BOUNDS:
psoDynStateMask &= ~CBSTATUS_DEPTH_BOUNDS_SET;
break;
case VK_DYNAMIC_STATE_STENCIL_COMPARE_MASK:
psoDynStateMask &= ~CBSTATUS_STENCIL_READ_MASK_SET;
break;
case VK_DYNAMIC_STATE_STENCIL_WRITE_MASK:
psoDynStateMask &= ~CBSTATUS_STENCIL_WRITE_MASK_SET;
break;
case VK_DYNAMIC_STATE_STENCIL_REFERENCE:
psoDynStateMask &= ~CBSTATUS_STENCIL_REFERENCE_SET;
break;
default:
// TODO : Flag error here
break;
}
}
pCB->status |= psoDynStateMask;
}
}
// Print the last bound Gfx Pipeline
static VkBool32 printPipeline(layer_data* my_data, const VkCommandBuffer cb)
{
VkBool32 skipCall = VK_FALSE;
GLOBAL_CB_NODE* pCB = getCBNode(my_data, cb);
if (pCB) {
PIPELINE_NODE *pPipeTrav = getPipeline(my_data, pCB->lastBoundPipeline);
if (!pPipeTrav) {
// nothing to print
} else {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_NONE, "DS",
"%s", vk_print_vkgraphicspipelinecreateinfo(&pPipeTrav->graphicsPipelineCI, "{DS}").c_str());
}
}
return skipCall;
}
// Print details of DS config to stdout
static VkBool32 printDSConfig(layer_data* my_data, const VkCommandBuffer cb)
{
VkBool32 skipCall = VK_FALSE;
GLOBAL_CB_NODE* pCB = getCBNode(my_data, cb);
if (pCB && pCB->lastBoundDescriptorSet) {
SET_NODE* pSet = getSetNode(my_data, pCB->lastBoundDescriptorSet);
DESCRIPTOR_POOL_NODE* pPool = getPoolNode(my_data, pSet->pool);
// Print out pool details
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_NONE, "DS",
"Details for pool %#" PRIxLEAST64 ".", (uint64_t) pPool->pool);
string poolStr = vk_print_vkdescriptorpoolcreateinfo(&pPool->createInfo, " ");
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_NONE, "DS",
"%s", poolStr.c_str());
// Print out set details
char prefix[10];
uint32_t index = 0;
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_NONE, "DS",
"Details for descriptor set %#" PRIxLEAST64 ".", (uint64_t) pSet->set);
LAYOUT_NODE* pLayout = pSet->pLayout;
// Print layout details
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_NONE, "DS",
"Layout #%u, (object %#" PRIxLEAST64 ") for DS %#" PRIxLEAST64 ".", index+1, (uint64_t)(pLayout->layout), (uint64_t)(pSet->set));
sprintf(prefix, " [L%u] ", index);
string DSLstr = vk_print_vkdescriptorsetlayoutcreateinfo(&pLayout->createInfo, prefix).c_str();
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_NONE, "DS",
"%s", DSLstr.c_str());
index++;
GENERIC_HEADER* pUpdate = pSet->pUpdateStructs;
if (pUpdate) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_NONE, "DS",
"Update Chain [UC] for descriptor set %#" PRIxLEAST64 ":", (uint64_t) pSet->set);
sprintf(prefix, " [UC] ");
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_NONE, "DS",
"%s", dynamic_display(pUpdate, prefix).c_str());
// TODO : If there is a "view" associated with this update, print CI for that view
} else {
if (0 != pSet->descriptorCount) {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_NONE, "DS",
"No Update Chain for descriptor set %#" PRIxLEAST64 " which has %u descriptors (vkUpdateDescriptors has not been called)", (uint64_t) pSet->set, pSet->descriptorCount);
} else {
skipCall |= log_msg(my_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_NONE, "DS",
"FYI: No descriptors in descriptor set %#" PRIxLEAST64 ".", (uint64_t) pSet->set);
}
}
}
return skipCall;
}
static void printCB(layer_data* my_data, const VkCommandBuffer cb)
{
GLOBAL_CB_NODE* pCB = getCBNode(my_data, cb);
if (pCB && pCB->cmds.size() > 0) {
log_msg(my_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_NONE, "DS",
"Cmds in CB %p", (void*)cb);
vector<CMD_NODE> cmds = pCB->cmds;
for (auto ii=cmds.begin(); ii!=cmds.end(); ++ii) {
// TODO : Need to pass cb as srcObj here
log_msg(my_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, 0, __LINE__, DRAWSTATE_NONE, "DS",
" CMD#%" PRIu64 ": %s", (*ii).cmdNumber, cmdTypeToString((*ii).type).c_str());
}
} else {
// Nothing to print
}
}
static VkBool32 synchAndPrintDSConfig(layer_data* my_data, const VkCommandBuffer cb)
{
VkBool32 skipCall = VK_FALSE;
if (!(my_data->report_data->active_flags & VK_DEBUG_REPORT_INFORMATION_BIT_EXT)) {
return skipCall;
}
skipCall |= printDSConfig(my_data, cb);
skipCall |= printPipeline(my_data, cb);
return skipCall;
}
// Flags validation error if the associated call is made inside a render pass. The apiName
// routine should ONLY be called outside a render pass.
static VkBool32 insideRenderPass(const layer_data* my_data, GLOBAL_CB_NODE *pCB, const char *apiName)
{
VkBool32 inside = VK_FALSE;
if (pCB->activeRenderPass) {
inside = log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT,
(uint64_t)pCB->commandBuffer, __LINE__, DRAWSTATE_INVALID_RENDERPASS_CMD, "DS",
"%s: It is invalid to issue this call inside an active render pass (%#" PRIxLEAST64 ")",
apiName, (uint64_t) pCB->activeRenderPass);
}
return inside;
}
// Flags validation error if the associated call is made outside a render pass. The apiName
// routine should ONLY be called inside a render pass.
static VkBool32 outsideRenderPass(const layer_data* my_data, GLOBAL_CB_NODE *pCB, const char *apiName)
{
VkBool32 outside = VK_FALSE;
if (((pCB->createInfo.level == VK_COMMAND_BUFFER_LEVEL_PRIMARY) &&
(!pCB->activeRenderPass)) ||
((pCB->createInfo.level == VK_COMMAND_BUFFER_LEVEL_SECONDARY) &&
(!pCB->activeRenderPass) &&
!(pCB->beginInfo.flags & VK_COMMAND_BUFFER_USAGE_RENDER_PASS_CONTINUE_BIT))) {
outside = log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT,
(uint64_t)pCB->commandBuffer, __LINE__, DRAWSTATE_NO_ACTIVE_RENDERPASS, "DS",
"%s: This call must be issued inside an active render pass.", apiName);
}
return outside;
}
static void init_draw_state(layer_data *my_data, const VkAllocationCallbacks *pAllocator)
{
uint32_t report_flags = 0;
uint32_t debug_action = 0;
FILE *log_output = NULL;
const char *option_str;
VkDebugReportCallbackEXT callback;
// initialize DrawState options
report_flags = getLayerOptionFlags("DrawStateReportFlags", 0);
getLayerOptionEnum("DrawStateDebugAction", (uint32_t *) &debug_action);
if (debug_action & VK_DBG_LAYER_ACTION_LOG_MSG)
{
option_str = getLayerOption("DrawStateLogFilename");
log_output = getLayerLogOutput(option_str, "DrawState");
VkDebugReportCallbackCreateInfoEXT dbgInfo;
memset(&dbgInfo, 0, sizeof(dbgInfo));
dbgInfo.sType = VK_STRUCTURE_TYPE_DEBUG_REPORT_CREATE_INFO_EXT;
dbgInfo.pfnCallback = log_callback;
dbgInfo.pUserData = log_output;
dbgInfo.flags = report_flags;
layer_create_msg_callback(my_data->report_data, &dbgInfo, pAllocator, &callback);
my_data->logging_callback.push_back(callback);
}
if (debug_action & VK_DBG_LAYER_ACTION_DEBUG_OUTPUT) {
VkDebugReportCallbackCreateInfoEXT dbgInfo;
memset(&dbgInfo, 0, sizeof(dbgInfo));
dbgInfo.sType = VK_STRUCTURE_TYPE_DEBUG_REPORT_CREATE_INFO_EXT;
dbgInfo.pfnCallback = win32_debug_output_msg;
dbgInfo.pUserData = log_output;
dbgInfo.flags = report_flags;
layer_create_msg_callback(my_data->report_data, &dbgInfo, pAllocator, &callback);
my_data->logging_callback.push_back(callback);
}
if (!globalLockInitialized)
{
loader_platform_thread_create_mutex(&globalLock);
globalLockInitialized = 1;
}
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkCreateInstance(const VkInstanceCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkInstance* pInstance)
{
VkLayerInstanceCreateInfo *chain_info = get_chain_info(pCreateInfo, VK_LAYER_LINK_INFO);
assert(chain_info->u.pLayerInfo);
PFN_vkGetInstanceProcAddr fpGetInstanceProcAddr = chain_info->u.pLayerInfo->pfnNextGetInstanceProcAddr;
PFN_vkCreateInstance fpCreateInstance = (PFN_vkCreateInstance) fpGetInstanceProcAddr(NULL, "vkCreateInstance");
if (fpCreateInstance == NULL) {
return VK_ERROR_INITIALIZATION_FAILED;
}
// Advance the link info for the next element on the chain
chain_info->u.pLayerInfo = chain_info->u.pLayerInfo->pNext;
VkResult result = fpCreateInstance(pCreateInfo, pAllocator, pInstance);
if (result != VK_SUCCESS)
return result;
// TBD: Need any locking this early, in case this function is called at the
// same time by more than one thread?
layer_data *my_data = get_my_data_ptr(get_dispatch_key(*pInstance), layer_data_map);
my_data->instance_dispatch_table = new VkLayerInstanceDispatchTable;
layer_init_instance_dispatch_table(*pInstance, my_data->instance_dispatch_table, fpGetInstanceProcAddr);
my_data->report_data = debug_report_create_instance(
my_data->instance_dispatch_table,
*pInstance,
pCreateInfo->enabledExtensionCount,
pCreateInfo->ppEnabledExtensionNames);
init_draw_state(my_data, pAllocator);
return result;
}
/* hook DestroyInstance to remove tableInstanceMap entry */
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkDestroyInstance(VkInstance instance, const VkAllocationCallbacks* pAllocator)
{
// TODOSC : Shouldn't need any customization here
dispatch_key key = get_dispatch_key(instance);
// TBD: Need any locking this early, in case this function is called at the
// same time by more than one thread?
layer_data *my_data = get_my_data_ptr(key, layer_data_map);
VkLayerInstanceDispatchTable *pTable = my_data->instance_dispatch_table;
pTable->DestroyInstance(instance, pAllocator);
loader_platform_thread_lock_mutex(&globalLock);
// Clean up logging callback, if any
while (my_data->logging_callback.size() > 0) {
VkDebugReportCallbackEXT callback = my_data->logging_callback.back();
layer_destroy_msg_callback(my_data->report_data, callback, pAllocator);
my_data->logging_callback.pop_back();
}
layer_debug_report_destroy_instance(my_data->report_data);
delete my_data->instance_dispatch_table;
layer_data_map.erase(key);
// TODO : Potential race here with separate threads creating/destroying instance
if (layer_data_map.empty()) {
// Release mutex when destroying last instance.
loader_platform_thread_unlock_mutex(&globalLock);
loader_platform_thread_delete_mutex(&globalLock);
globalLockInitialized = 0;
} else {
loader_platform_thread_unlock_mutex(&globalLock);
}
}
static void createDeviceRegisterExtensions(const VkDeviceCreateInfo* pCreateInfo, VkDevice device)
{
uint32_t i;
// TBD: Need any locking, in case this function is called at the same time
// by more than one thread?
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
dev_data->device_extensions.debug_marker_enabled = false;
dev_data->device_extensions.wsi_enabled = false;
VkLayerDispatchTable *pDisp = dev_data->device_dispatch_table;
PFN_vkGetDeviceProcAddr gpa = pDisp->GetDeviceProcAddr;
pDisp->CreateSwapchainKHR = (PFN_vkCreateSwapchainKHR) gpa(device, "vkCreateSwapchainKHR");
pDisp->DestroySwapchainKHR = (PFN_vkDestroySwapchainKHR) gpa(device, "vkDestroySwapchainKHR");
pDisp->GetSwapchainImagesKHR = (PFN_vkGetSwapchainImagesKHR) gpa(device, "vkGetSwapchainImagesKHR");
pDisp->AcquireNextImageKHR = (PFN_vkAcquireNextImageKHR) gpa(device, "vkAcquireNextImageKHR");
pDisp->QueuePresentKHR = (PFN_vkQueuePresentKHR) gpa(device, "vkQueuePresentKHR");
for (i = 0; i < pCreateInfo->enabledExtensionCount; i++) {
if (strcmp(pCreateInfo->ppEnabledExtensionNames[i], VK_KHR_SWAPCHAIN_EXTENSION_NAME) == 0) {
dev_data->device_extensions.wsi_enabled = true;
}
if (strcmp(pCreateInfo->ppEnabledExtensionNames[i], DEBUG_MARKER_EXTENSION_NAME) == 0) {
/* Found a matching extension name, mark it enabled and init dispatch table*/
dev_data->device_extensions.debug_marker_enabled = true;
initDebugMarkerTable(device);
}
}
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkCreateDevice(VkPhysicalDevice gpu, const VkDeviceCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkDevice* pDevice)
{
VkLayerDeviceCreateInfo *chain_info = get_chain_info(pCreateInfo, VK_LAYER_LINK_INFO);
assert(chain_info->u.pLayerInfo);
PFN_vkGetInstanceProcAddr fpGetInstanceProcAddr = chain_info->u.pLayerInfo->pfnNextGetInstanceProcAddr;
PFN_vkGetDeviceProcAddr fpGetDeviceProcAddr = chain_info->u.pLayerInfo->pfnNextGetDeviceProcAddr;
PFN_vkCreateDevice fpCreateDevice = (PFN_vkCreateDevice) fpGetInstanceProcAddr(NULL, "vkCreateDevice");
if (fpCreateDevice == NULL) {
return VK_ERROR_INITIALIZATION_FAILED;
}
// Advance the link info for the next element on the chain
chain_info->u.pLayerInfo = chain_info->u.pLayerInfo->pNext;
VkResult result = fpCreateDevice(gpu, pCreateInfo, pAllocator, pDevice);
if (result != VK_SUCCESS) {
return result;
}
loader_platform_thread_lock_mutex(&globalLock);
layer_data *my_instance_data = get_my_data_ptr(get_dispatch_key(gpu), layer_data_map);
layer_data *my_device_data = get_my_data_ptr(get_dispatch_key(*pDevice), layer_data_map);
// Setup device dispatch table
my_device_data->device_dispatch_table = new VkLayerDispatchTable;
layer_init_device_dispatch_table(*pDevice, my_device_data->device_dispatch_table, fpGetDeviceProcAddr);
my_device_data->report_data = layer_debug_report_create_device(my_instance_data->report_data, *pDevice);
createDeviceRegisterExtensions(pCreateInfo, *pDevice);
// Get physical device limits for this device
my_instance_data->instance_dispatch_table->GetPhysicalDeviceProperties(gpu, &(my_device_data->physDevProperties.properties));
my_instance_data->instance_dispatch_table->GetPhysicalDeviceFeatures(gpu, &(my_device_data->physDevProperties.features));
uint32_t count;
my_instance_data->instance_dispatch_table->GetPhysicalDeviceQueueFamilyProperties(gpu, &count, nullptr);
my_device_data->physDevProperties.queue_family_properties.resize(count);
my_instance_data->instance_dispatch_table->GetPhysicalDeviceQueueFamilyProperties(gpu, &count, &my_device_data->physDevProperties.queue_family_properties[0]);
// TODO: device limits should make sure these are compatible
if (pCreateInfo->pEnabledFeatures) {
my_device_data->physDevProperties.features =
*pCreateInfo->pEnabledFeatures;
} else {
my_instance_data->instance_dispatch_table->GetPhysicalDeviceFeatures(
gpu, &my_device_data->physDevProperties.features);
}
loader_platform_thread_unlock_mutex(&globalLock);
return result;
}
// prototype
static void deleteRenderPasses(layer_data*);
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkDestroyDevice(VkDevice device, const VkAllocationCallbacks* pAllocator)
{
// TODOSC : Shouldn't need any customization here
dispatch_key key = get_dispatch_key(device);
layer_data* dev_data = get_my_data_ptr(key, layer_data_map);
// Free all the memory
loader_platform_thread_lock_mutex(&globalLock);
deletePipelines(dev_data);
deleteRenderPasses(dev_data);
deleteCommandBuffers(dev_data);
deletePools(dev_data);
deleteLayouts(dev_data);
dev_data->imageViewMap.clear();
dev_data->imageMap.clear();
dev_data->bufferViewMap.clear();
dev_data->bufferMap.clear();
loader_platform_thread_unlock_mutex(&globalLock);
dev_data->device_dispatch_table->DestroyDevice(device, pAllocator);
tableDebugMarkerMap.erase(key);
delete dev_data->device_dispatch_table;
layer_data_map.erase(key);
}
static const VkExtensionProperties instance_extensions[] = {
{
VK_EXT_DEBUG_REPORT_EXTENSION_NAME,
VK_EXT_DEBUG_REPORT_SPEC_VERSION
}
};
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkEnumerateInstanceExtensionProperties(
const char *pLayerName,
uint32_t *pCount,
VkExtensionProperties* pProperties)
{
return util_GetExtensionProperties(1, instance_extensions, pCount, pProperties);
}
static const VkLayerProperties ds_global_layers[] = {
{
"VK_LAYER_LUNARG_draw_state",
VK_API_VERSION,
VK_MAKE_VERSION(0, 1, 0),
"Validation layer: draw_state",
}
};
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkEnumerateInstanceLayerProperties(
uint32_t *pCount,
VkLayerProperties* pProperties)
{
return util_GetLayerProperties(ARRAY_SIZE(ds_global_layers),
ds_global_layers,
pCount, pProperties);
}
static const VkExtensionProperties ds_device_extensions[] = {
{
DEBUG_MARKER_EXTENSION_NAME,
VK_MAKE_VERSION(0, 1, 0),
}
};
static const VkLayerProperties ds_device_layers[] = {
{
"VK_LAYER_LUNARG_draw_state",
VK_API_VERSION,
VK_MAKE_VERSION(0, 1, 0),
"Validation layer: draw_state",
}
};
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkEnumerateDeviceExtensionProperties(
VkPhysicalDevice physicalDevice,
const char* pLayerName,
uint32_t* pCount,
VkExtensionProperties* pProperties)
{
// DrawState does not have any physical device extensions
if (pLayerName == NULL) {
dispatch_key key = get_dispatch_key(physicalDevice);
layer_data *my_data = get_my_data_ptr(key, layer_data_map);
return my_data->instance_dispatch_table->EnumerateDeviceExtensionProperties(
physicalDevice,
NULL,
pCount,
pProperties);
} else {
return util_GetExtensionProperties(ARRAY_SIZE(ds_device_extensions),
ds_device_extensions,
pCount, pProperties);
}
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkEnumerateDeviceLayerProperties(
VkPhysicalDevice physicalDevice,
uint32_t* pCount,
VkLayerProperties* pProperties)
{
/* DrawState physical device layers are the same as global */
return util_GetLayerProperties(ARRAY_SIZE(ds_device_layers), ds_device_layers,
pCount, pProperties);
}
// This validates that the initial layout specified in the command buffer for
// the IMAGE is the same
// as the global IMAGE layout
VkBool32 ValidateCmdBufImageLayouts(VkCommandBuffer cmdBuffer) {
VkBool32 skip_call = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(cmdBuffer), layer_data_map);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, cmdBuffer);
for (auto cb_image_data : pCB->imageLayoutMap) {
VkImageLayout imageLayout;
if (!FindLayout(dev_data, cb_image_data.first, imageLayout)) {
skip_call |= log_msg(
dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, 0, __LINE__,
DRAWSTATE_INVALID_IMAGE_LAYOUT, "DS",
"Cannot submit cmd buffer using deleted image %" PRIu64 ".",
reinterpret_cast<const uint64_t &>(cb_image_data.first));
} else {
if (imageLayout != cb_image_data.second.initialLayout) {
skip_call |= log_msg(
dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, 0, __LINE__,
DRAWSTATE_INVALID_IMAGE_LAYOUT, "DS",
"Cannot submit cmd buffer using image with layout %s when "
"first use is %s.",
string_VkImageLayout(imageLayout),
string_VkImageLayout(cb_image_data.second.initialLayout));
}
SetLayout(dev_data, cb_image_data.first, cb_image_data.second.layout);
}
}
return skip_call;
}
// Track which resources are in-flight by atomically incrementing their "in_use" count
VkBool32 validateAndIncrementResources(layer_data* my_data, GLOBAL_CB_NODE* pCB) {
VkBool32 skip_call = VK_FALSE;
for (auto drawDataElement : pCB->drawData) {
for (auto buffer : drawDataElement.buffers) {
auto buffer_data = my_data->bufferMap.find(buffer);
if (buffer_data == my_data->bufferMap.end()) {
skip_call |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_BUFFER_EXT, (uint64_t)(buffer), __LINE__, DRAWSTATE_INVALID_BUFFER, "DS",
"Cannot submit cmd buffer using deleted buffer %" PRIu64 ".", (uint64_t)(buffer));
} else {
buffer_data->second.in_use.fetch_add(1);
}
}
}
for (auto set : pCB->uniqueBoundSets) {
auto setNode = my_data->setMap.find(set);
if (setNode == my_data->setMap.end()) {
skip_call |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT, (uint64_t)(set), __LINE__, DRAWSTATE_INVALID_DESCRIPTOR_SET, "DS",
"Cannot submit cmd buffer using deleted descriptor set %" PRIu64 ".", (uint64_t)(set));
} else {
setNode->second->in_use.fetch_add(1);
}
}
for (auto semaphore : pCB->semaphores) {
auto semaphoreNode = my_data->semaphoreMap.find(semaphore);
if (semaphoreNode == my_data->semaphoreMap.end()) {
skip_call |= log_msg(
my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT,
reinterpret_cast<uint64_t &>(semaphore), __LINE__,
DRAWSTATE_INVALID_SEMAPHORE, "DS",
"Cannot submit cmd buffer using deleted semaphore %" PRIu64 ".",
reinterpret_cast<uint64_t &>(semaphore));
} else {
semaphoreNode->second.in_use.fetch_add(1);
}
}
for (auto event : pCB->events) {
auto eventNode = my_data->eventMap.find(event);
if (eventNode == my_data->eventMap.end()) {
skip_call |= log_msg(
my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT,
reinterpret_cast<uint64_t &>(event), __LINE__,
DRAWSTATE_INVALID_EVENT, "DS",
"Cannot submit cmd buffer using deleted event %" PRIu64 ".",
reinterpret_cast<uint64_t &>(event));
} else {
eventNode->second.in_use.fetch_add(1);
}
}
return skip_call;
}
void decrementResources(layer_data* my_data, VkCommandBuffer cmdBuffer) {
GLOBAL_CB_NODE* pCB = getCBNode(my_data, cmdBuffer);
for (auto drawDataElement : pCB->drawData) {
for (auto buffer : drawDataElement.buffers) {
auto buffer_data = my_data->bufferMap.find(buffer);
if (buffer_data != my_data->bufferMap.end()) {
buffer_data->second.in_use.fetch_sub(1);
}
}
}
for (auto set : pCB->uniqueBoundSets) {
auto setNode = my_data->setMap.find(set);
if (setNode != my_data->setMap.end()) {
setNode->second->in_use.fetch_sub(1);
}
}
for (auto semaphore : pCB->semaphores) {
auto semaphoreNode = my_data->semaphoreMap.find(semaphore);
if (semaphoreNode != my_data->semaphoreMap.end()) {
semaphoreNode->second.in_use.fetch_sub(1);
}
}
for (auto event : pCB->events) {
auto eventNode = my_data->eventMap.find(event);
if (eventNode != my_data->eventMap.end()) {
eventNode->second.in_use.fetch_sub(1);
}
}
for (auto queryStatePair : pCB->queryToStateMap) {
my_data->queryToStateMap[queryStatePair.first] = queryStatePair.second;
}
for (auto eventStagePair : pCB->eventToStageMap) {
my_data->eventMap[eventStagePair.first].stageMask =
eventStagePair.second;
}
}
void decrementResources(layer_data* my_data, uint32_t fenceCount, const VkFence* pFences) {
for (uint32_t i = 0; i < fenceCount; ++i) {
auto fence_data = my_data->fenceMap.find(pFences[i]);
if (fence_data == my_data->fenceMap.end() || !fence_data->second.needsSignaled) return;
fence_data->second.needsSignaled = false;
fence_data->second.in_use.fetch_sub(1);
if (fence_data->second.priorFence != VK_NULL_HANDLE) {
decrementResources(my_data, 1, &fence_data->second.priorFence);
}
for (auto cmdBuffer : fence_data->second.cmdBuffers) {
decrementResources(my_data, cmdBuffer);
}
}
}
void decrementResources(layer_data* my_data, VkQueue queue) {
auto queue_data = my_data->queueMap.find(queue);
if (queue_data != my_data->queueMap.end()) {
for (auto cmdBuffer : queue_data->second.untrackedCmdBuffers) {
decrementResources(my_data, cmdBuffer);
}
queue_data->second.untrackedCmdBuffers.clear();
decrementResources(my_data, 1, &queue_data->second.priorFence);
}
}
void trackCommandBuffers(layer_data* my_data, VkQueue queue, uint32_t cmdBufferCount, const VkCommandBuffer* pCmdBuffers, VkFence fence) {
auto queue_data = my_data->queueMap.find(queue);
if (fence != VK_NULL_HANDLE) {
VkFence priorFence = VK_NULL_HANDLE;
auto fence_data = my_data->fenceMap.find(fence);
if (fence_data == my_data->fenceMap.end()) {
return;
}
if (queue_data != my_data->queueMap.end()) {
priorFence = queue_data->second.priorFence;
queue_data->second.priorFence = fence;
for (auto cmdBuffer : queue_data->second.untrackedCmdBuffers) {
fence_data->second.cmdBuffers.push_back(cmdBuffer);
}
queue_data->second.untrackedCmdBuffers.clear();
}
fence_data->second.cmdBuffers.clear();
fence_data->second.priorFence = priorFence;
fence_data->second.needsSignaled = true;
fence_data->second.queue = queue;
fence_data->second.in_use.fetch_add(1);
for (uint32_t i = 0; i < cmdBufferCount; ++i) {
for (auto secondaryCmdBuffer :
my_data->commandBufferMap[pCmdBuffers[i]]
->secondaryCommandBuffers) {
fence_data->second.cmdBuffers.push_back(
secondaryCmdBuffer);
}
fence_data->second.cmdBuffers.push_back(pCmdBuffers[i]);
}
} else {
if (queue_data != my_data->queueMap.end()) {
for (uint32_t i = 0; i < cmdBufferCount; ++i) {
for (auto secondaryCmdBuffer : my_data->commandBufferMap[pCmdBuffers[i]]->secondaryCommandBuffers) {
queue_data->second.untrackedCmdBuffers.push_back(secondaryCmdBuffer);
}
queue_data->second.untrackedCmdBuffers.push_back(pCmdBuffers[i]);
}
}
}
if (queue_data != my_data->queueMap.end()) {
for (uint32_t i = 0; i < cmdBufferCount; ++i) {
// Add cmdBuffers to both the global set and queue set
for (auto secondaryCmdBuffer : my_data->commandBufferMap[pCmdBuffers[i]]->secondaryCommandBuffers) {
my_data->globalInFlightCmdBuffers.insert(secondaryCmdBuffer);
queue_data->second.inFlightCmdBuffers.insert(secondaryCmdBuffer);
}
my_data->globalInFlightCmdBuffers.insert(pCmdBuffers[i]);
queue_data->second.inFlightCmdBuffers.insert(pCmdBuffers[i]);
}
}
}
bool validateCommandBufferSimultaneousUse(layer_data *dev_data,
GLOBAL_CB_NODE *pCB) {
bool skip_call = false;
if (dev_data->globalInFlightCmdBuffers.count(pCB->commandBuffer) &&
!(pCB->beginInfo.flags &
VK_COMMAND_BUFFER_USAGE_SIMULTANEOUS_USE_BIT)) {
skip_call |= log_msg(
dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, 0, __LINE__,
DRAWSTATE_INVALID_FENCE, "DS",
"Command Buffer %#" PRIx64 " is already in use and is not marked "
"for simultaneous use.",
reinterpret_cast<uint64_t>(pCB->commandBuffer));
}
return skip_call;
}
static bool validateCommandBufferState(layer_data *dev_data,
GLOBAL_CB_NODE *pCB) {
bool skipCall = false;
// Validate that cmd buffers have been updated
if (CB_RECORDED != pCB->state) {
if (CB_INVALID == pCB->state) {
// Inform app of reason CB invalid
if (!pCB->destroyedSets.empty()) {
std::stringstream set_string;
for (auto set : pCB->destroyedSets) {
set_string << " " << set;
}
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, (uint64_t)(pCB->commandBuffer), __LINE__, DRAWSTATE_INVALID_COMMAND_BUFFER, "DS",
"You are submitting command buffer %#" PRIxLEAST64 " that is invalid because it had the following bound descriptor set(s) destroyed: %s", (uint64_t)(pCB->commandBuffer), set_string.str().c_str());
}
if (!pCB->updatedSets.empty()) {
std::stringstream set_string;
for (auto set : pCB->updatedSets) {
set_string << " " << set;
}
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, (uint64_t)(pCB->commandBuffer), __LINE__, DRAWSTATE_INVALID_COMMAND_BUFFER, "DS",
"You are submitting command buffer %#" PRIxLEAST64 " that is invalid because it had the following bound descriptor set(s) updated: %s", (uint64_t)(pCB->commandBuffer), set_string.str().c_str());
}
} else { // Flag error for using CB w/o vkEndCommandBuffer() called
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, (uint64_t)(pCB->commandBuffer), __LINE__, DRAWSTATE_NO_END_COMMAND_BUFFER, "DS",
"You must call vkEndCommandBuffer() on CB %#" PRIxLEAST64 " before this call to vkQueueSubmit()!", (uint64_t)(pCB->commandBuffer));
}
}
return skipCall;
}
static VkBool32 validatePrimaryCommandBufferState(layer_data *dev_data,
GLOBAL_CB_NODE *pCB) {
// Track in-use for resources off of primary and any secondary CBs
VkBool32 skipCall = validateAndIncrementResources(dev_data, pCB);
if (!pCB->secondaryCommandBuffers.empty()) {
for (auto secondaryCmdBuffer : pCB->secondaryCommandBuffers) {
skipCall |= validateAndIncrementResources(
dev_data, dev_data->commandBufferMap[secondaryCmdBuffer]);
GLOBAL_CB_NODE* pSubCB = getCBNode(dev_data, secondaryCmdBuffer);
if (pSubCB->primaryCommandBuffer != pCB->commandBuffer) {
log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, 0,
__LINE__,
DRAWSTATE_COMMAND_BUFFER_SINGLE_SUBMIT_VIOLATION, "DS",
"CB %#" PRIxLEAST64
" was submitted with secondary buffer %#" PRIxLEAST64
" but that buffer has subsequently been bound to "
"primary cmd buffer %#" PRIxLEAST64 ".",
reinterpret_cast<uint64_t>(pCB->commandBuffer),
reinterpret_cast<uint64_t>(secondaryCmdBuffer),
reinterpret_cast<uint64_t>(
pSubCB->primaryCommandBuffer));
}
}
}
// TODO : Verify if this also needs to be checked for secondary command
// buffers. If so, this block of code can move to
// validateCommandBufferState() function. vulkan GL106 filed to clarify
if ((pCB->beginInfo.flags & VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT) &&
(pCB->submitCount > 1)) {
skipCall |=
log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, 0, __LINE__,
DRAWSTATE_COMMAND_BUFFER_SINGLE_SUBMIT_VIOLATION, "DS",
"CB %#" PRIxLEAST64
" was begun w/ VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT "
"set, but has been submitted %#" PRIxLEAST64 " times.",
(uint64_t)(pCB->commandBuffer), pCB->submitCount);
}
skipCall |= validateCommandBufferState(dev_data, pCB);
// If USAGE_SIMULTANEOUS_USE_BIT not set then CB cannot already be executing
// on device
skipCall |= validateCommandBufferSimultaneousUse(dev_data, pCB);
return skipCall;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkQueueSubmit(VkQueue queue, uint32_t submitCount, const VkSubmitInfo* pSubmits, VkFence fence)
{
VkBool32 skipCall = VK_FALSE;
GLOBAL_CB_NODE* pCB = NULL;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(queue), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
for (uint32_t submit_idx = 0; submit_idx < submitCount; submit_idx++) {
const VkSubmitInfo *submit = &pSubmits[submit_idx];
vector<VkSemaphore> semaphoreList;
for (uint32_t i=0; i < submit->waitSemaphoreCount; ++i) {
semaphoreList.push_back(submit->pWaitSemaphores[i]);
if (dev_data->semaphoreMap[submit->pWaitSemaphores[i]].signaled) {
dev_data->semaphoreMap[submit->pWaitSemaphores[i]].signaled = 0;
} else {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, 0, __LINE__, DRAWSTATE_QUEUE_FORWARD_PROGRESS, "DS",
"Queue %#" PRIx64 " is waiting on semaphore %#" PRIx64 " that has no way to be signaled.",
(uint64_t)(queue), (uint64_t)(submit->pWaitSemaphores[i]));
}
}
for (uint32_t i=0; i < submit->signalSemaphoreCount; ++i) {
semaphoreList.push_back(submit->pSignalSemaphores[i]);
dev_data->semaphoreMap[submit->pSignalSemaphores[i]].signaled = 1;
}
for (uint32_t i=0; i < submit->commandBufferCount; i++) {
skipCall |= ValidateCmdBufImageLayouts(submit->pCommandBuffers[i]);
pCB = getCBNode(dev_data, submit->pCommandBuffers[i]);
pCB->semaphores = semaphoreList;
pCB->submitCount++; // increment submit count
skipCall |= validatePrimaryCommandBufferState(dev_data, pCB);
}
if ((fence != VK_NULL_HANDLE) && dev_data->fenceMap[fence].in_use.load()) {
skipCall |= log_msg(
dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_FENCE_EXT,
(uint64_t)(fence), __LINE__,
DRAWSTATE_INVALID_FENCE, "DS",
"Fence %#" PRIx64 " is already in use by another submission.",
(uint64_t)(fence));
}
trackCommandBuffers(dev_data, queue, submit->commandBufferCount,
submit->pCommandBuffers, fence);
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
return dev_data->device_dispatch_table->QueueSubmit(queue, submitCount, pSubmits, fence);
return VK_ERROR_VALIDATION_FAILED_EXT;
}
// Note: This function assumes that the global lock is held by the calling
// thread.
VkBool32 cleanInFlightCmdBuffer(layer_data* my_data, VkCommandBuffer cmdBuffer) {
VkBool32 skip_call = VK_FALSE;
GLOBAL_CB_NODE* pCB = getCBNode(my_data, cmdBuffer);
if (pCB) {
for (auto queryEventsPair : pCB->waitedEventsBeforeQueryReset) {
for (auto event : queryEventsPair.second) {
if (my_data->eventMap[event].needsSignaled) {
skip_call |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_QUERY_POOL_EXT, 0, 0, DRAWSTATE_INVALID_QUERY, "DS",
"Cannot get query results on queryPool %" PRIu64 " with index %d which was guarded by unsignaled event %" PRIu64 ".",
(uint64_t)(queryEventsPair.first.pool), queryEventsPair.first.index, (uint64_t)(event));
}
}
}
}
return skip_call;
}
// Remove given cmd_buffer from the global inFlight set.
// Also, if given queue is valid, then remove the cmd_buffer from that queues
// inFlightCmdBuffer set. Finally, check all other queues and if given cmd_buffer
// is still in flight on another queue, add it back into the global set.
// Note: This function assumes that the global lock is held by the calling
// thread.
static inline void removeInFlightCmdBuffer(layer_data* dev_data, VkCommandBuffer cmd_buffer, VkQueue queue)
{
// Pull it off of global list initially, but if we find it in any other queue list, add it back in
dev_data->globalInFlightCmdBuffers.erase(cmd_buffer);
if (dev_data->queueMap.find(queue) != dev_data->queueMap.end()) {
dev_data->queueMap[queue].inFlightCmdBuffers.erase(cmd_buffer);
for (auto q : dev_data->queues) {
if ((q != queue) && (dev_data->queueMap[q].inFlightCmdBuffers.find(cmd_buffer) != dev_data->queueMap[q].inFlightCmdBuffers.end())) {
dev_data->globalInFlightCmdBuffers.insert(cmd_buffer);
break;
}
}
}
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkWaitForFences(VkDevice device, uint32_t fenceCount, const VkFence* pFences, VkBool32 waitAll, uint64_t timeout)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->WaitForFences(device, fenceCount, pFences, waitAll, timeout);
VkBool32 skip_call = VK_FALSE;
loader_platform_thread_lock_mutex(&globalLock);
if (result == VK_SUCCESS) {
// When we know that all fences are complete we can clean/remove their CBs
if (waitAll || fenceCount == 1) {
for (uint32_t i = 0; i < fenceCount; ++i) {
VkQueue fence_queue = dev_data->fenceMap[pFences[i]].queue;
for (auto cmdBuffer : dev_data->fenceMap[pFences[i]].cmdBuffers) {
skip_call |= cleanInFlightCmdBuffer(dev_data, cmdBuffer);
removeInFlightCmdBuffer(dev_data, cmdBuffer, fence_queue);
}
}
decrementResources(dev_data, fenceCount, pFences);
}
// NOTE : Alternate case not handled here is when some fences have completed. In
// this case for app to guarantee which fences completed it will have to call
// vkGetFenceStatus() at which point we'll clean/remove their CBs if complete.
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE != skip_call)
return VK_ERROR_VALIDATION_FAILED_EXT;
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkGetFenceStatus(VkDevice device, VkFence fence)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->GetFenceStatus(device, fence);
VkBool32 skip_call = VK_FALSE;
loader_platform_thread_lock_mutex(&globalLock);
if (result == VK_SUCCESS) {
auto fence_queue = dev_data->fenceMap[fence].queue;
for (auto cmdBuffer : dev_data->fenceMap[fence].cmdBuffers) {
skip_call |= cleanInFlightCmdBuffer(dev_data, cmdBuffer);
removeInFlightCmdBuffer(dev_data, cmdBuffer, fence_queue);
}
decrementResources(dev_data, 1, &fence);
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE != skip_call)
return VK_ERROR_VALIDATION_FAILED_EXT;
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkGetDeviceQueue(VkDevice device, uint32_t queueFamilyIndex, uint32_t queueIndex, VkQueue* pQueue)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
dev_data->device_dispatch_table->GetDeviceQueue(device, queueFamilyIndex, queueIndex, pQueue);
dev_data->queues.push_back(*pQueue);
dev_data->queueMap[*pQueue].device = device;
loader_platform_thread_unlock_mutex(&globalLock);
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkQueueWaitIdle(VkQueue queue)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(queue), layer_data_map);
decrementResources(dev_data, queue);
VkBool32 skip_call = VK_FALSE;
loader_platform_thread_lock_mutex(&globalLock);
// Iterate over local set since we erase set members as we go in for loop
auto local_cb_set = dev_data->queueMap[queue].inFlightCmdBuffers;
for (auto cmdBuffer : local_cb_set) {
skip_call |= cleanInFlightCmdBuffer(dev_data, cmdBuffer);
removeInFlightCmdBuffer(dev_data, cmdBuffer, queue);
}
dev_data->queueMap[queue].inFlightCmdBuffers.clear();
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE != skip_call)
return VK_ERROR_VALIDATION_FAILED_EXT;
return dev_data->device_dispatch_table->QueueWaitIdle(queue);
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkDeviceWaitIdle(VkDevice device)
{
VkBool32 skip_call = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
for (auto queue : dev_data->queues) {
decrementResources(dev_data, queue);
if (dev_data->queueMap.find(queue) != dev_data->queueMap.end()) {
// Clear all of the queue inFlightCmdBuffers (global set cleared below)
dev_data->queueMap[queue].inFlightCmdBuffers.clear();
}
}
for (auto cmdBuffer : dev_data->globalInFlightCmdBuffers) {
skip_call |= cleanInFlightCmdBuffer(dev_data, cmdBuffer);
}
dev_data->globalInFlightCmdBuffers.clear();
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE != skip_call)
return VK_ERROR_VALIDATION_FAILED_EXT;
return dev_data->device_dispatch_table->DeviceWaitIdle(device);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkDestroyFence(VkDevice device, VkFence fence, const VkAllocationCallbacks* pAllocator)
{
layer_data *dev_data =
get_my_data_ptr(get_dispatch_key(device), layer_data_map);
bool skipCall = false;
loader_platform_thread_lock_mutex(&globalLock);
if (dev_data->fenceMap[fence].in_use.load()) {
skipCall |=
log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_FENCE_EXT, (uint64_t)(fence),
__LINE__, DRAWSTATE_INVALID_FENCE, "DS",
"Fence %#" PRIx64 " is in use by a command buffer.",
(uint64_t)(fence));
}
loader_platform_thread_unlock_mutex(&globalLock);
if (!skipCall)
dev_data->device_dispatch_table->DestroyFence(device, fence,
pAllocator);
// TODO : Clean up any internal data structures using this obj.
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkDestroySemaphore(VkDevice device, VkSemaphore semaphore, const VkAllocationCallbacks* pAllocator)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
dev_data->device_dispatch_table->DestroySemaphore(device, semaphore, pAllocator);
loader_platform_thread_lock_mutex(&globalLock);
if (dev_data->semaphoreMap[semaphore].in_use.load()) {
log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_SEMAPHORE_EXT,
reinterpret_cast<uint64_t &>(semaphore), __LINE__,
DRAWSTATE_INVALID_SEMAPHORE, "DS",
"Cannot delete semaphore %" PRIx64 " which is in use.",
reinterpret_cast<uint64_t &>(semaphore));
}
dev_data->semaphoreMap.erase(semaphore);
loader_platform_thread_unlock_mutex(&globalLock);
// TODO : Clean up any internal data structures using this obj.
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkDestroyEvent(VkDevice device, VkEvent event, const VkAllocationCallbacks* pAllocator)
{
layer_data *dev_data =
get_my_data_ptr(get_dispatch_key(device), layer_data_map);
bool skip_call = false;
loader_platform_thread_lock_mutex(&globalLock);
auto event_data = dev_data->eventMap.find(event);
if (event_data != dev_data->eventMap.end() &&
event_data->second.in_use.load()) {
skip_call |=
log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT,
reinterpret_cast<uint64_t &>(event), __LINE__,
DRAWSTATE_INVALID_EVENT, "DS",
"Cannot delete event %" PRIu64
" which is in use by a command buffer.",
reinterpret_cast<uint64_t &>(event));
}
loader_platform_thread_unlock_mutex(&globalLock);
if (!skip_call)
dev_data->device_dispatch_table->DestroyEvent(device, event,
pAllocator);
// TODO : Clean up any internal data structures using this obj.
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkDestroyQueryPool(VkDevice device, VkQueryPool queryPool, const VkAllocationCallbacks* pAllocator)
{
get_my_data_ptr(get_dispatch_key(device), layer_data_map)->device_dispatch_table->DestroyQueryPool(device, queryPool, pAllocator);
// TODO : Clean up any internal data structures using this obj.
}
VKAPI_ATTR VkResult VKAPI_CALL vkGetQueryPoolResults(VkDevice device, VkQueryPool queryPool, uint32_t firstQuery, uint32_t queryCount,
size_t dataSize, void* pData, VkDeviceSize stride, VkQueryResultFlags flags) {
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
unordered_map<QueryObject, vector<VkCommandBuffer>> queriesInFlight;
GLOBAL_CB_NODE* pCB = nullptr;
loader_platform_thread_lock_mutex(&globalLock);
for (auto cmdBuffer : dev_data->globalInFlightCmdBuffers) {
pCB = getCBNode(dev_data, cmdBuffer);
for (auto queryStatePair : pCB->queryToStateMap) {
queriesInFlight[queryStatePair.first].push_back(cmdBuffer);
}
}
VkBool32 skip_call = VK_FALSE;
for (uint32_t i = 0; i < queryCount; ++i) {
QueryObject query = {queryPool, firstQuery + i};
auto queryElement = queriesInFlight.find(query);
auto queryToStateElement = dev_data->queryToStateMap.find(query);
if (queryToStateElement != dev_data->queryToStateMap.end()) {
}
// Available and in flight
if(queryElement != queriesInFlight.end() && queryToStateElement != dev_data->queryToStateMap.end() && queryToStateElement->second) {
for (auto cmdBuffer : queryElement->second) {
pCB = getCBNode(dev_data, cmdBuffer);
auto queryEventElement = pCB->waitedEventsBeforeQueryReset.find(query);
if (queryEventElement == pCB->waitedEventsBeforeQueryReset.end()) {
skip_call |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_QUERY_POOL_EXT, 0, __LINE__,
DRAWSTATE_INVALID_QUERY, "DS", "Cannot get query results on queryPool %" PRIu64 " with index %d which is in flight.",
(uint64_t)(queryPool), firstQuery + i);
} else {
for (auto event : queryEventElement->second) {
dev_data->eventMap[event].needsSignaled = true;
}
}
}
// Unavailable and in flight
} else if (queryElement != queriesInFlight.end() && queryToStateElement != dev_data->queryToStateMap.end() && !queryToStateElement->second) {
// TODO : Can there be the same query in use by multiple command buffers in flight?
bool make_available = false;
for (auto cmdBuffer : queryElement->second) {
pCB = getCBNode(dev_data, cmdBuffer);
make_available |= pCB->queryToStateMap[query];
}
if (!(((flags & VK_QUERY_RESULT_PARTIAL_BIT) || (flags & VK_QUERY_RESULT_WAIT_BIT)) && make_available)) {
skip_call |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_QUERY_POOL_EXT, 0, __LINE__, DRAWSTATE_INVALID_QUERY, "DS",
"Cannot get query results on queryPool %" PRIu64 " with index %d which is unavailable.",
(uint64_t)(queryPool), firstQuery + i);
}
// Unavailable
} else if (queryToStateElement != dev_data->queryToStateMap.end() && !queryToStateElement->second) {
skip_call |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_QUERY_POOL_EXT, 0, __LINE__, DRAWSTATE_INVALID_QUERY, "DS",
"Cannot get query results on queryPool %" PRIu64 " with index %d which is unavailable.",
(uint64_t)(queryPool), firstQuery + i);
// Unitialized
} else if (queryToStateElement == dev_data->queryToStateMap.end()) {
skip_call |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_QUERY_POOL_EXT, 0, __LINE__, DRAWSTATE_INVALID_QUERY, "DS",
"Cannot get query results on queryPool %" PRIu64 " with index %d which is uninitialized.",
(uint64_t)(queryPool), firstQuery + i);
}
}
loader_platform_thread_unlock_mutex(&globalLock);
if (skip_call)
return VK_ERROR_VALIDATION_FAILED_EXT;
return dev_data->device_dispatch_table->GetQueryPoolResults(device, queryPool, firstQuery, queryCount, dataSize, pData, stride, flags);
}
VkBool32 validateIdleBuffer(const layer_data* my_data, VkBuffer buffer) {
VkBool32 skip_call = VK_FALSE;
auto buffer_data = my_data->bufferMap.find(buffer);
if (buffer_data == my_data->bufferMap.end()) {
skip_call |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_BUFFER_EXT, (uint64_t)(buffer), __LINE__, DRAWSTATE_DOUBLE_DESTROY, "DS",
"Cannot free buffer %" PRIxLEAST64 " that has not been allocated.", (uint64_t)(buffer));
} else {
if (buffer_data->second.in_use.load()) {
skip_call |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_BUFFER_EXT, (uint64_t)(buffer), __LINE__, DRAWSTATE_OBJECT_INUSE, "DS",
"Cannot free buffer %" PRIxLEAST64 " that is in use by a command buffer.", (uint64_t)(buffer));
}
}
return skip_call;
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkDestroyBuffer(VkDevice device, VkBuffer buffer, const VkAllocationCallbacks* pAllocator)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
if (!validateIdleBuffer(dev_data, buffer)) {
loader_platform_thread_unlock_mutex(&globalLock);
dev_data->device_dispatch_table->DestroyBuffer(device, buffer, pAllocator);
loader_platform_thread_lock_mutex(&globalLock);
}
dev_data->bufferMap.erase(buffer);
loader_platform_thread_unlock_mutex(&globalLock);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkDestroyBufferView(VkDevice device, VkBufferView bufferView, const VkAllocationCallbacks* pAllocator)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
dev_data->device_dispatch_table->DestroyBufferView(device, bufferView, pAllocator);
loader_platform_thread_lock_mutex(&globalLock);
dev_data->bufferViewMap.erase(bufferView);
loader_platform_thread_unlock_mutex(&globalLock);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkDestroyImage(VkDevice device, VkImage image, const VkAllocationCallbacks* pAllocator)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
dev_data->device_dispatch_table->DestroyImage(device, image, pAllocator);
loader_platform_thread_lock_mutex(&globalLock);
dev_data->imageMap.erase(image);
loader_platform_thread_unlock_mutex(&globalLock);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkDestroyImageView(VkDevice device, VkImageView imageView, const VkAllocationCallbacks* pAllocator)
{
get_my_data_ptr(get_dispatch_key(device), layer_data_map)->device_dispatch_table->DestroyImageView(device, imageView, pAllocator);
// TODO : Clean up any internal data structures using this obj.
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkDestroyShaderModule(VkDevice device, VkShaderModule shaderModule, const VkAllocationCallbacks* pAllocator)
{
get_my_data_ptr(get_dispatch_key(device), layer_data_map)->device_dispatch_table->DestroyShaderModule(device, shaderModule, pAllocator);
// TODO : Clean up any internal data structures using this obj.
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkDestroyPipeline(VkDevice device, VkPipeline pipeline, const VkAllocationCallbacks* pAllocator)
{
get_my_data_ptr(get_dispatch_key(device), layer_data_map)->device_dispatch_table->DestroyPipeline(device, pipeline, pAllocator);
// TODO : Clean up any internal data structures using this obj.
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkDestroyPipelineLayout(VkDevice device, VkPipelineLayout pipelineLayout, const VkAllocationCallbacks* pAllocator)
{
get_my_data_ptr(get_dispatch_key(device), layer_data_map)->device_dispatch_table->DestroyPipelineLayout(device, pipelineLayout, pAllocator);
// TODO : Clean up any internal data structures using this obj.
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkDestroySampler(VkDevice device, VkSampler sampler, const VkAllocationCallbacks* pAllocator)
{
get_my_data_ptr(get_dispatch_key(device), layer_data_map)->device_dispatch_table->DestroySampler(device, sampler, pAllocator);
// TODO : Clean up any internal data structures using this obj.
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkDestroyDescriptorSetLayout(VkDevice device, VkDescriptorSetLayout descriptorSetLayout, const VkAllocationCallbacks* pAllocator)
{
get_my_data_ptr(get_dispatch_key(device), layer_data_map)->device_dispatch_table->DestroyDescriptorSetLayout(device, descriptorSetLayout, pAllocator);
// TODO : Clean up any internal data structures using this obj.
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkDestroyDescriptorPool(VkDevice device, VkDescriptorPool descriptorPool, const VkAllocationCallbacks* pAllocator)
{
get_my_data_ptr(get_dispatch_key(device), layer_data_map)->device_dispatch_table->DestroyDescriptorPool(device, descriptorPool, pAllocator);
// TODO : Clean up any internal data structures using this obj.
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkFreeCommandBuffers(VkDevice device, VkCommandPool commandPool, uint32_t count, const VkCommandBuffer *pCommandBuffers)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
bool skip_call = false;
loader_platform_thread_lock_mutex(&globalLock);
for (uint32_t i = 0; i < count; i++) {
if (dev_data->globalInFlightCmdBuffers.count(pCommandBuffers[i])) {
skip_call |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT,
reinterpret_cast<uint64_t>(pCommandBuffers[i]), __LINE__, DRAWSTATE_INVALID_COMMAND_BUFFER_RESET, "DS",
"Attempt to free command buffer (%#" PRIxLEAST64 ") which is in use.", reinterpret_cast<uint64_t>(pCommandBuffers[i]));
}
// Delete CB information structure, and remove from commandBufferMap
auto cb = dev_data->commandBufferMap.find(pCommandBuffers[i]);
if (cb != dev_data->commandBufferMap.end()) {
// reset prior to delete for data clean-up
resetCB(dev_data, (*cb).second->commandBuffer);
delete (*cb).second;
dev_data->commandBufferMap.erase(cb);
}
// Remove commandBuffer reference from commandPoolMap
dev_data->commandPoolMap[commandPool].commandBuffers.remove(pCommandBuffers[i]);
}
loader_platform_thread_unlock_mutex(&globalLock);
if (!skip_call)
dev_data->device_dispatch_table->FreeCommandBuffers(device, commandPool, count, pCommandBuffers);
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkCreateCommandPool(VkDevice device, const VkCommandPoolCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkCommandPool* pCommandPool)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->CreateCommandPool(device, pCreateInfo, pAllocator, pCommandPool);
if (VK_SUCCESS == result) {
loader_platform_thread_lock_mutex(&globalLock);
dev_data->commandPoolMap[*pCommandPool].createFlags = pCreateInfo->flags;
dev_data->commandPoolMap[*pCommandPool].queueFamilyIndex = pCreateInfo->queueFamilyIndex;
loader_platform_thread_unlock_mutex(&globalLock);
}
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkCreateQueryPool(
VkDevice device, const VkQueryPoolCreateInfo *pCreateInfo,
const VkAllocationCallbacks *pAllocator, VkQueryPool *pQueryPool) {
layer_data *dev_data =
get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->CreateQueryPool(
device, pCreateInfo, pAllocator, pQueryPool);
if (result == VK_SUCCESS) {
loader_platform_thread_lock_mutex(&globalLock);
dev_data->queryPoolMap[*pQueryPool].createInfo = *pCreateInfo;
loader_platform_thread_unlock_mutex(&globalLock);
}
return result;
}
VkBool32 validateCommandBuffersNotInUse(const layer_data* dev_data, VkCommandPool commandPool) {
VkBool32 skipCall = VK_FALSE;
auto pool_data = dev_data->commandPoolMap.find(commandPool);
if (pool_data != dev_data->commandPoolMap.end()) {
for (auto cmdBuffer : pool_data->second.commandBuffers) {
if (dev_data->globalInFlightCmdBuffers.count(cmdBuffer)) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_POOL_EXT, (uint64_t)(commandPool),
__LINE__, DRAWSTATE_OBJECT_INUSE, "DS", "Cannot reset command pool %" PRIx64 " when allocated command buffer %" PRIx64 " is in use.",
(uint64_t)(commandPool), (uint64_t)(cmdBuffer));
}
}
}
return skipCall;
}
// Destroy commandPool along with all of the commandBuffers allocated from that pool
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkDestroyCommandPool(VkDevice device, VkCommandPool commandPool, const VkAllocationCallbacks* pAllocator)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
// Must remove cmdpool from cmdpoolmap, after removing all cmdbuffers in its list from the commandPoolMap
if (dev_data->commandPoolMap.find(commandPool) != dev_data->commandPoolMap.end()) {
for (auto poolCb = dev_data->commandPoolMap[commandPool].commandBuffers.begin(); poolCb != dev_data->commandPoolMap[commandPool].commandBuffers.end();) {
auto del_cb = dev_data->commandBufferMap.find(*poolCb);
delete (*del_cb).second; // delete CB info structure
dev_data->commandBufferMap.erase(del_cb); // Remove this command buffer from cbMap
poolCb = dev_data->commandPoolMap[commandPool].commandBuffers.erase(poolCb); // Remove CB reference from commandPoolMap's list
}
}
dev_data->commandPoolMap.erase(commandPool);
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_TRUE == validateCommandBuffersNotInUse(dev_data, commandPool))
return;
dev_data->device_dispatch_table->DestroyCommandPool(device, commandPool, pAllocator);
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkResetCommandPool(
VkDevice device,
VkCommandPool commandPool,
VkCommandPoolResetFlags flags)
{
layer_data *dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = VK_ERROR_VALIDATION_FAILED_EXT;
if (VK_TRUE == validateCommandBuffersNotInUse(dev_data, commandPool))
return VK_ERROR_VALIDATION_FAILED_EXT;
result = dev_data->device_dispatch_table->ResetCommandPool(device, commandPool, flags);
// Reset all of the CBs allocated from this pool
if (VK_SUCCESS == result) {
loader_platform_thread_lock_mutex(&globalLock);
auto it = dev_data->commandPoolMap[commandPool].commandBuffers.begin();
while (it != dev_data->commandPoolMap[commandPool].commandBuffers.end()) {
resetCB(dev_data, (*it));
++it;
}
loader_platform_thread_unlock_mutex(&globalLock);
}
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL
vkResetFences(VkDevice device, uint32_t fenceCount, const VkFence *pFences) {
layer_data *dev_data =
get_my_data_ptr(get_dispatch_key(device), layer_data_map);
bool skipCall = false;
loader_platform_thread_lock_mutex(&globalLock);
for (uint32_t i = 0; i < fenceCount; ++i) {
if (dev_data->fenceMap[pFences[i]].in_use.load()) {
skipCall |=
log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_FENCE_EXT,
reinterpret_cast<const uint64_t &>(pFences[i]),
__LINE__, DRAWSTATE_INVALID_FENCE, "DS",
"Fence %#" PRIx64 " is in use by a command buffer.",
reinterpret_cast<const uint64_t &>(pFences[i]));
}
}
loader_platform_thread_unlock_mutex(&globalLock);
VkResult result = VK_ERROR_VALIDATION_FAILED_EXT;
if (!skipCall)
result = dev_data->device_dispatch_table->ResetFences(
device, fenceCount, pFences);
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkDestroyFramebuffer(VkDevice device, VkFramebuffer framebuffer, const VkAllocationCallbacks* pAllocator)
{
get_my_data_ptr(get_dispatch_key(device), layer_data_map)->device_dispatch_table->DestroyFramebuffer(device, framebuffer, pAllocator);
// TODO : Clean up any internal data structures using this obj.
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkDestroyRenderPass(VkDevice device, VkRenderPass renderPass, const VkAllocationCallbacks* pAllocator)
{
get_my_data_ptr(get_dispatch_key(device), layer_data_map)->device_dispatch_table->DestroyRenderPass(device, renderPass, pAllocator);
// TODO : Clean up any internal data structures using this obj.
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkCreateBuffer(VkDevice device, const VkBufferCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkBuffer* pBuffer)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->CreateBuffer(device, pCreateInfo, pAllocator, pBuffer);
if (VK_SUCCESS == result) {
loader_platform_thread_lock_mutex(&globalLock);
// TODO : This doesn't create deep copy of pQueueFamilyIndices so need to fix that if/when we want that data to be valid
dev_data->bufferMap[*pBuffer].create_info = unique_ptr<VkBufferCreateInfo>(new VkBufferCreateInfo(*pCreateInfo));
dev_data->bufferMap[*pBuffer].in_use.store(0);
loader_platform_thread_unlock_mutex(&globalLock);
}
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkCreateBufferView(VkDevice device, const VkBufferViewCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkBufferView* pView)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->CreateBufferView(device, pCreateInfo, pAllocator, pView);
if (VK_SUCCESS == result) {
loader_platform_thread_lock_mutex(&globalLock);
dev_data->bufferViewMap[*pView] = unique_ptr<VkBufferViewCreateInfo>(new VkBufferViewCreateInfo(*pCreateInfo));
loader_platform_thread_unlock_mutex(&globalLock);
}
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkCreateImage(VkDevice device, const VkImageCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkImage* pImage)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->CreateImage(device, pCreateInfo, pAllocator, pImage);
if (VK_SUCCESS == result) {
IMAGE_NODE image_node;
image_node.layout = pCreateInfo->initialLayout;
image_node.format = pCreateInfo->format;
loader_platform_thread_lock_mutex(&globalLock);
dev_data->imageMap[*pImage] = unique_ptr<VkImageCreateInfo>(new VkImageCreateInfo(*pCreateInfo));
ImageSubresourcePair subpair = {*pImage, false, VkImageSubresource()};
dev_data->imageSubresourceMap[*pImage].push_back(subpair);
dev_data->imageLayoutMap[subpair] = image_node;
loader_platform_thread_unlock_mutex(&globalLock);
}
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkCreateImageView(VkDevice device, const VkImageViewCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkImageView* pView)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->CreateImageView(device, pCreateInfo, pAllocator, pView);
if (VK_SUCCESS == result) {
loader_platform_thread_lock_mutex(&globalLock);
dev_data->imageViewMap[*pView] = unique_ptr<VkImageViewCreateInfo>(new VkImageViewCreateInfo(*pCreateInfo));
loader_platform_thread_unlock_mutex(&globalLock);
}
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL
vkCreateFence(VkDevice device, const VkFenceCreateInfo* pCreateInfo,
const VkAllocationCallbacks* pAllocator, VkFence* pFence) {
layer_data *dev_data =
get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->CreateFence(
device, pCreateInfo, pAllocator, pFence);
if (VK_SUCCESS == result) {
loader_platform_thread_lock_mutex(&globalLock);
dev_data->fenceMap[*pFence].in_use.store(0);
loader_platform_thread_unlock_mutex(&globalLock);
}
return result;
}
// TODO handle pipeline caches
VKAPI_ATTR VkResult VKAPI_CALL
vkCreatePipelineCache(VkDevice device,
const VkPipelineCacheCreateInfo *pCreateInfo,
const VkAllocationCallbacks *pAllocator,
VkPipelineCache *pPipelineCache) {
layer_data *dev_data =
get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->CreatePipelineCache(
device, pCreateInfo, pAllocator, pPipelineCache);
return result;
}
VKAPI_ATTR void VKAPI_CALL vkDestroyPipelineCache(
VkDevice device,
VkPipelineCache pipelineCache,
const VkAllocationCallbacks* pAllocator)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
dev_data->device_dispatch_table->DestroyPipelineCache(device, pipelineCache, pAllocator);
}
VKAPI_ATTR VkResult VKAPI_CALL vkGetPipelineCacheData(
VkDevice device,
VkPipelineCache pipelineCache,
size_t* pDataSize,
void* pData)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->GetPipelineCacheData(device, pipelineCache, pDataSize, pData);
return result;
}
VKAPI_ATTR VkResult VKAPI_CALL vkMergePipelineCaches(
VkDevice device,
VkPipelineCache dstCache,
uint32_t srcCacheCount,
const VkPipelineCache* pSrcCaches)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->MergePipelineCaches(device, dstCache, srcCacheCount, pSrcCaches);
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkCreateGraphicsPipelines(
VkDevice device,
VkPipelineCache pipelineCache,
uint32_t count,
const VkGraphicsPipelineCreateInfo *pCreateInfos,
const VkAllocationCallbacks *pAllocator,
VkPipeline *pPipelines)
{
VkResult result = VK_SUCCESS;
//TODO What to do with pipelineCache?
// The order of operations here is a little convoluted but gets the job done
// 1. Pipeline create state is first shadowed into PIPELINE_NODE struct
// 2. Create state is then validated (which uses flags setup during shadowing)
// 3. If everything looks good, we'll then create the pipeline and add NODE to pipelineMap
VkBool32 skipCall = VK_FALSE;
// TODO : Improve this data struct w/ unique_ptrs so cleanup below is automatic
vector<PIPELINE_NODE*> pPipeNode(count);
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
uint32_t i=0;
loader_platform_thread_lock_mutex(&globalLock);
for (i=0; i<count; i++) {
pPipeNode[i] = initGraphicsPipeline(dev_data, &pCreateInfos[i], NULL);
skipCall |= verifyPipelineCreateState(dev_data, device, pPipeNode[i]);
}
if (VK_FALSE == skipCall) {
loader_platform_thread_unlock_mutex(&globalLock);
result = dev_data->device_dispatch_table->CreateGraphicsPipelines(device,
pipelineCache, count, pCreateInfos, pAllocator, pPipelines);
loader_platform_thread_lock_mutex(&globalLock);
for (i=0; i<count; i++) {
pPipeNode[i]->pipeline = pPipelines[i];
dev_data->pipelineMap[pPipeNode[i]->pipeline] = pPipeNode[i];
}
loader_platform_thread_unlock_mutex(&globalLock);
} else {
for (i=0; i<count; i++) {
if (pPipeNode[i]) {
// If we allocated a pipeNode, need to clean it up here
delete[] pPipeNode[i]->pVertexBindingDescriptions;
delete[] pPipeNode[i]->pVertexAttributeDescriptions;
delete[] pPipeNode[i]->pAttachments;
delete pPipeNode[i];
}
}
loader_platform_thread_unlock_mutex(&globalLock);
return VK_ERROR_VALIDATION_FAILED_EXT;
}
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkCreateComputePipelines(
VkDevice device,
VkPipelineCache pipelineCache,
uint32_t count,
const VkComputePipelineCreateInfo *pCreateInfos,
const VkAllocationCallbacks *pAllocator,
VkPipeline *pPipelines)
{
VkResult result = VK_SUCCESS;
VkBool32 skipCall = VK_FALSE;
// TODO : Improve this data struct w/ unique_ptrs so cleanup below is automatic
vector<PIPELINE_NODE*> pPipeNode(count);
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
uint32_t i=0;
loader_platform_thread_lock_mutex(&globalLock);
for (i=0; i<count; i++) {
// TODO: Verify compute stage bits
// Create and initialize internal tracking data structure
pPipeNode[i] = new PIPELINE_NODE;
memcpy(&pPipeNode[i]->computePipelineCI, (const void*)&pCreateInfos[i], sizeof(VkComputePipelineCreateInfo));
// TODO: Add Compute Pipeline Verification
// skipCall |= verifyPipelineCreateState(dev_data, device, pPipeNode[i]);
}
if (VK_FALSE == skipCall) {
loader_platform_thread_unlock_mutex(&globalLock);
result = dev_data->device_dispatch_table->CreateComputePipelines(device, pipelineCache, count, pCreateInfos, pAllocator, pPipelines);
loader_platform_thread_lock_mutex(&globalLock);
for (i=0; i<count; i++) {
pPipeNode[i]->pipeline = pPipelines[i];
dev_data->pipelineMap[pPipeNode[i]->pipeline] = pPipeNode[i];
}
loader_platform_thread_unlock_mutex(&globalLock);
} else {
for (i=0; i<count; i++) {
if (pPipeNode[i]) {
// Clean up any locally allocated data structures
delete pPipeNode[i];
}
}
loader_platform_thread_unlock_mutex(&globalLock);
return VK_ERROR_VALIDATION_FAILED_EXT;
}
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkCreateSampler(VkDevice device, const VkSamplerCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkSampler* pSampler)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->CreateSampler(device, pCreateInfo, pAllocator, pSampler);
if (VK_SUCCESS == result) {
loader_platform_thread_lock_mutex(&globalLock);
dev_data->sampleMap[*pSampler] = unique_ptr<SAMPLER_NODE>(new SAMPLER_NODE(pSampler, pCreateInfo));
loader_platform_thread_unlock_mutex(&globalLock);
}
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkCreateDescriptorSetLayout(VkDevice device, const VkDescriptorSetLayoutCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkDescriptorSetLayout* pSetLayout)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->CreateDescriptorSetLayout(device, pCreateInfo, pAllocator, pSetLayout);
if (VK_SUCCESS == result) {
// TODOSC : Capture layout bindings set
LAYOUT_NODE* pNewNode = new LAYOUT_NODE;
if (NULL == pNewNode) {
if (log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_LAYOUT_EXT, (uint64_t) *pSetLayout, __LINE__, DRAWSTATE_OUT_OF_MEMORY, "DS",
"Out of memory while attempting to allocate LAYOUT_NODE in vkCreateDescriptorSetLayout()"))
return VK_ERROR_VALIDATION_FAILED_EXT;
}
memcpy((void*)&pNewNode->createInfo, pCreateInfo, sizeof(VkDescriptorSetLayoutCreateInfo));
pNewNode->createInfo.pBindings = new VkDescriptorSetLayoutBinding[pCreateInfo->bindingCount];
memcpy((void*)pNewNode->createInfo.pBindings, pCreateInfo->pBindings, sizeof(VkDescriptorSetLayoutBinding)*pCreateInfo->bindingCount);
// g++ does not like reserve with size 0
if (pCreateInfo->bindingCount)
pNewNode->bindingToIndexMap.reserve(pCreateInfo->bindingCount);
uint32_t totalCount = 0;
for (uint32_t i = 0; i < pCreateInfo->bindingCount; i++) {
if (!pNewNode->bindingToIndexMap.emplace(pCreateInfo->pBindings[i].binding, i).second) {
if (log_msg(
dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_LAYOUT_EXT,
(uint64_t)*pSetLayout, __LINE__,
DRAWSTATE_INVALID_LAYOUT, "DS",
"duplicated binding number in "
"VkDescriptorSetLayoutBinding"))
return VK_ERROR_VALIDATION_FAILED_EXT;
} else {
pNewNode->bindingToIndexMap[pCreateInfo->pBindings[i].binding] = i;
}
totalCount += pCreateInfo->pBindings[i].descriptorCount;
if (pCreateInfo->pBindings[i].pImmutableSamplers) {
VkSampler** ppIS = (VkSampler**)&pNewNode->createInfo.pBindings[i].pImmutableSamplers;
*ppIS = new VkSampler[pCreateInfo->pBindings[i].descriptorCount];
memcpy(*ppIS, pCreateInfo->pBindings[i].pImmutableSamplers, pCreateInfo->pBindings[i].descriptorCount*sizeof(VkSampler));
}
}
pNewNode->layout = *pSetLayout;
pNewNode->startIndex = 0;
if (totalCount > 0) {
pNewNode->descriptorTypes.resize(totalCount);
pNewNode->stageFlags.resize(totalCount);
uint32_t offset = 0;
uint32_t j = 0;
VkDescriptorType dType;
for (uint32_t i=0; i<pCreateInfo->bindingCount; i++) {
dType = pCreateInfo->pBindings[i].descriptorType;
for (j = 0; j < pCreateInfo->pBindings[i].descriptorCount; j++) {
pNewNode->descriptorTypes[offset + j] = dType;
pNewNode->stageFlags[offset + j] = pCreateInfo->pBindings[i].stageFlags;
if ((dType == VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC) ||
(dType == VK_DESCRIPTOR_TYPE_STORAGE_BUFFER_DYNAMIC)) {
pNewNode->dynamicDescriptorCount++;
}
}
offset += j;
}
pNewNode->endIndex = pNewNode->startIndex + totalCount - 1;
} else { // no descriptors
pNewNode->endIndex = 0;
}
// Put new node at Head of global Layer list
loader_platform_thread_lock_mutex(&globalLock);
dev_data->descriptorSetLayoutMap[*pSetLayout] = pNewNode;
loader_platform_thread_unlock_mutex(&globalLock);
}
return result;
}
VKAPI_ATTR VkResult VKAPI_CALL vkCreatePipelineLayout(VkDevice device, const VkPipelineLayoutCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkPipelineLayout* pPipelineLayout)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->CreatePipelineLayout(device, pCreateInfo, pAllocator, pPipelineLayout);
if (VK_SUCCESS == result) {
loader_platform_thread_lock_mutex(&globalLock);
// TODOSC : Merge capture of the setLayouts per pipeline
PIPELINE_LAYOUT_NODE& plNode = dev_data->pipelineLayoutMap[*pPipelineLayout];
plNode.descriptorSetLayouts.resize(pCreateInfo->setLayoutCount);
uint32_t i = 0;
for (i=0; i<pCreateInfo->setLayoutCount; ++i) {
plNode.descriptorSetLayouts[i] = pCreateInfo->pSetLayouts[i];
}
plNode.pushConstantRanges.resize(pCreateInfo->pushConstantRangeCount);
for (i=0; i<pCreateInfo->pushConstantRangeCount; ++i) {
plNode.pushConstantRanges[i] = pCreateInfo->pPushConstantRanges[i];
}
loader_platform_thread_unlock_mutex(&globalLock);
}
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkCreateDescriptorPool(VkDevice device, const VkDescriptorPoolCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkDescriptorPool* pDescriptorPool)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->CreateDescriptorPool(device, pCreateInfo, pAllocator, pDescriptorPool);
if (VK_SUCCESS == result) {
// Insert this pool into Global Pool LL at head
if (log_msg(dev_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_POOL_EXT, (uint64_t) *pDescriptorPool, __LINE__, DRAWSTATE_OUT_OF_MEMORY, "DS",
"Created Descriptor Pool %#" PRIxLEAST64, (uint64_t) *pDescriptorPool))
return VK_ERROR_VALIDATION_FAILED_EXT;
DESCRIPTOR_POOL_NODE* pNewNode = new DESCRIPTOR_POOL_NODE(*pDescriptorPool, pCreateInfo);
if (NULL == pNewNode) {
if (log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_POOL_EXT, (uint64_t) *pDescriptorPool, __LINE__, DRAWSTATE_OUT_OF_MEMORY, "DS",
"Out of memory while attempting to allocate DESCRIPTOR_POOL_NODE in vkCreateDescriptorPool()"))
return VK_ERROR_VALIDATION_FAILED_EXT;
} else {
loader_platform_thread_lock_mutex(&globalLock);
dev_data->descriptorPoolMap[*pDescriptorPool] = pNewNode;
loader_platform_thread_unlock_mutex(&globalLock);
}
} else {
// Need to do anything if pool create fails?
}
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkResetDescriptorPool(VkDevice device, VkDescriptorPool descriptorPool, VkDescriptorPoolResetFlags flags)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->ResetDescriptorPool(device, descriptorPool, flags);
if (VK_SUCCESS == result) {
loader_platform_thread_lock_mutex(&globalLock);
clearDescriptorPool(dev_data, device, descriptorPool, flags);
loader_platform_thread_unlock_mutex(&globalLock);
}
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkAllocateDescriptorSets(VkDevice device, const VkDescriptorSetAllocateInfo* pAllocateInfo, VkDescriptorSet* pDescriptorSets)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
// Verify that requested descriptorSets are available in pool
DESCRIPTOR_POOL_NODE *pPoolNode = getPoolNode(dev_data, pAllocateInfo->descriptorPool);
if (!pPoolNode) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_POOL_EXT, (uint64_t) pAllocateInfo->descriptorPool, __LINE__, DRAWSTATE_INVALID_POOL, "DS",
"Unable to find pool node for pool %#" PRIxLEAST64 " specified in vkAllocateDescriptorSets() call", (uint64_t) pAllocateInfo->descriptorPool);
} else { // Make sure pool has all the available descriptors before calling down chain
skipCall |= validate_descriptor_availability_in_pool(dev_data, pPoolNode, pAllocateInfo->descriptorSetCount, pAllocateInfo->pSetLayouts);
}
loader_platform_thread_unlock_mutex(&globalLock);
if (skipCall)
return VK_ERROR_VALIDATION_FAILED_EXT;
VkResult result = dev_data->device_dispatch_table->AllocateDescriptorSets(device, pAllocateInfo, pDescriptorSets);
if (VK_SUCCESS == result) {
loader_platform_thread_lock_mutex(&globalLock);
DESCRIPTOR_POOL_NODE *pPoolNode = getPoolNode(dev_data, pAllocateInfo->descriptorPool);
if (pPoolNode) {
if (pAllocateInfo->descriptorSetCount == 0) {
log_msg(dev_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT, pAllocateInfo->descriptorSetCount, __LINE__, DRAWSTATE_NONE, "DS",
"AllocateDescriptorSets called with 0 count");
}
for (uint32_t i = 0; i < pAllocateInfo->descriptorSetCount; i++) {
log_msg(dev_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT, (uint64_t) pDescriptorSets[i], __LINE__, DRAWSTATE_NONE, "DS",
"Created Descriptor Set %#" PRIxLEAST64, (uint64_t) pDescriptorSets[i]);
// Create new set node and add to head of pool nodes
SET_NODE* pNewNode = new SET_NODE;
if (NULL == pNewNode) {
if (log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT, (uint64_t) pDescriptorSets[i], __LINE__, DRAWSTATE_OUT_OF_MEMORY, "DS",
"Out of memory while attempting to allocate SET_NODE in vkAllocateDescriptorSets()"))
return VK_ERROR_VALIDATION_FAILED_EXT;
} else {
// TODO : Pool should store a total count of each type of Descriptor available
// When descriptors are allocated, decrement the count and validate here
// that the count doesn't go below 0. One reset/free need to bump count back up.
// Insert set at head of Set LL for this pool
pNewNode->pNext = pPoolNode->pSets;
pNewNode->in_use.store(0);
pPoolNode->pSets = pNewNode;
LAYOUT_NODE* pLayout = getLayoutNode(dev_data, pAllocateInfo->pSetLayouts[i]);
if (NULL == pLayout) {
if (log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_LAYOUT_EXT, (uint64_t) pAllocateInfo->pSetLayouts[i], __LINE__, DRAWSTATE_INVALID_LAYOUT, "DS",
"Unable to find set layout node for layout %#" PRIxLEAST64 " specified in vkAllocateDescriptorSets() call", (uint64_t) pAllocateInfo->pSetLayouts[i]))
return VK_ERROR_VALIDATION_FAILED_EXT;
}
pNewNode->pLayout = pLayout;
pNewNode->pool = pAllocateInfo->descriptorPool;
pNewNode->set = pDescriptorSets[i];
pNewNode->descriptorCount = (pLayout->createInfo.bindingCount != 0) ? pLayout->endIndex + 1 : 0;
if (pNewNode->descriptorCount) {
size_t descriptorArraySize = sizeof(GENERIC_HEADER*)*pNewNode->descriptorCount;
pNewNode->ppDescriptors = new GENERIC_HEADER*[descriptorArraySize];
memset(pNewNode->ppDescriptors, 0, descriptorArraySize);
}
dev_data->setMap[pDescriptorSets[i]] = pNewNode;
}
}
}
loader_platform_thread_unlock_mutex(&globalLock);
}
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkFreeDescriptorSets(VkDevice device, VkDescriptorPool descriptorPool, uint32_t count, const VkDescriptorSet* pDescriptorSets)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
// Make sure that no sets being destroyed are in-flight
loader_platform_thread_lock_mutex(&globalLock);
for (uint32_t i=0; i<count; ++i)
skipCall |= validateIdleDescriptorSet(dev_data, pDescriptorSets[i], "vkFreeDesriptorSets");
DESCRIPTOR_POOL_NODE *pPoolNode = getPoolNode(dev_data, descriptorPool);
if (pPoolNode && !(VK_DESCRIPTOR_POOL_CREATE_FREE_DESCRIPTOR_SET_BIT & pPoolNode->createInfo.flags)) {
// Can't Free from a NON_FREE pool
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DEVICE_EXT, (uint64_t)device, __LINE__, DRAWSTATE_CANT_FREE_FROM_NON_FREE_POOL, "DS",
"It is invalid to call vkFreeDescriptorSets() with a pool created without setting VK_DESCRIPTOR_POOL_CREATE_FREE_DESCRIPTOR_SET_BIT.");
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE != skipCall)
return VK_ERROR_VALIDATION_FAILED_EXT;
VkResult result = dev_data->device_dispatch_table->FreeDescriptorSets(device, descriptorPool, count, pDescriptorSets);
if (VK_SUCCESS == result) {
// For each freed descriptor add it back into the pool as available
loader_platform_thread_lock_mutex(&globalLock);
for (uint32_t i=0; i<count; ++i) {
SET_NODE* pSet = dev_data->setMap[pDescriptorSets[i]]; // getSetNode() without locking
invalidateBoundCmdBuffers(dev_data, pSet);
LAYOUT_NODE* pLayout = pSet->pLayout;
uint32_t typeIndex = 0, poolSizeCount = 0;
for (uint32_t j=0; j<pLayout->createInfo.bindingCount; ++j) {
typeIndex = static_cast<uint32_t>(pLayout->createInfo.pBindings[j].descriptorType);
poolSizeCount = pLayout->createInfo.pBindings[j].descriptorCount;
pPoolNode->availableDescriptorTypeCount[typeIndex] += poolSizeCount;
}
}
loader_platform_thread_unlock_mutex(&globalLock);
}
// TODO : Any other clean-up or book-keeping to do here?
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkUpdateDescriptorSets(VkDevice device, uint32_t descriptorWriteCount, const VkWriteDescriptorSet* pDescriptorWrites, uint32_t descriptorCopyCount, const VkCopyDescriptorSet* pDescriptorCopies)
{
// dsUpdate will return VK_TRUE only if a bailout error occurs, so we want to call down tree when update returns VK_FALSE
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
VkBool32 rtn = dsUpdate(dev_data,
device,
descriptorWriteCount,
pDescriptorWrites,
descriptorCopyCount,
pDescriptorCopies);
loader_platform_thread_unlock_mutex(&globalLock);
if (!rtn) {
dev_data->device_dispatch_table->UpdateDescriptorSets(device, descriptorWriteCount, pDescriptorWrites, descriptorCopyCount, pDescriptorCopies);
}
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkAllocateCommandBuffers(VkDevice device, const VkCommandBufferAllocateInfo* pCreateInfo, VkCommandBuffer* pCommandBuffer)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->AllocateCommandBuffers(device, pCreateInfo, pCommandBuffer);
if (VK_SUCCESS == result) {
loader_platform_thread_lock_mutex(&globalLock);
for (uint32_t i = 0; i < pCreateInfo->commandBufferCount; i++) {
// Validate command pool
if (dev_data->commandPoolMap.find(pCreateInfo->commandPool) != dev_data->commandPoolMap.end()) {
// Add command buffer to its commandPool map
dev_data->commandPoolMap[pCreateInfo->commandPool].commandBuffers.push_back(pCommandBuffer[i]);
GLOBAL_CB_NODE* pCB = new GLOBAL_CB_NODE;
// Add command buffer to map
dev_data->commandBufferMap[pCommandBuffer[i]] = pCB;
resetCB(dev_data, pCommandBuffer[i]);
pCB->createInfo = *pCreateInfo;
pCB->device = device;
}
}
loader_platform_thread_unlock_mutex(&globalLock);
}
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkBeginCommandBuffer(VkCommandBuffer commandBuffer, const VkCommandBufferBeginInfo* pBeginInfo)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
// Validate command buffer level
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
if (pCB->createInfo.level != VK_COMMAND_BUFFER_LEVEL_PRIMARY) {
// Secondary Command Buffer
const VkCommandBufferInheritanceInfo *pInfo = pBeginInfo->pInheritanceInfo;
if (!pInfo) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, reinterpret_cast<uint64_t>(commandBuffer), __LINE__,
DRAWSTATE_BEGIN_CB_INVALID_STATE, "DS", "vkBeginCommandBuffer(): Secondary Command Buffer (%p) must have inheritance info.",
reinterpret_cast<void*>(commandBuffer));
} else {
if (pBeginInfo->flags & VK_COMMAND_BUFFER_USAGE_RENDER_PASS_CONTINUE_BIT) {
if (!pInfo->renderPass) { // renderpass should NOT be null for an Secondary CB
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, reinterpret_cast<uint64_t>(commandBuffer),
__LINE__, DRAWSTATE_BEGIN_CB_INVALID_STATE, "DS",
"vkBeginCommandBuffer(): Secondary Command Buffers (%p) must specify a valid renderpass parameter.", reinterpret_cast<void*>(commandBuffer));
}
if (!pInfo->framebuffer) { // framebuffer may be null for an Secondary CB, but this affects perf
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_WARNING_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, reinterpret_cast<uint64_t>(commandBuffer),
__LINE__, DRAWSTATE_BEGIN_CB_INVALID_STATE, "DS",
"vkBeginCommandBuffer(): Secondary Command Buffers (%p) may perform better if a valid framebuffer parameter is specified.",
reinterpret_cast<void*>(commandBuffer));
} else {
string errorString = "";
VkRenderPass fbRP = dev_data->frameBufferMap[pInfo->framebuffer]->renderPass;
if (!verify_renderpass_compatibility(dev_data, fbRP, pInfo->renderPass, errorString)) {
// renderPass that framebuffer was created with must be compatible with local renderPass
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, reinterpret_cast<uint64_t>(commandBuffer),
__LINE__, DRAWSTATE_RENDERPASS_INCOMPATIBLE, "DS",
"vkBeginCommandBuffer(): Secondary Command Buffer (%p) renderPass (%#" PRIxLEAST64 ") is incompatible w/ framebuffer (%#" PRIxLEAST64
") w/ render pass (%#" PRIxLEAST64 ") due to: %s", reinterpret_cast<void*>(commandBuffer), (uint64_t)(pInfo->renderPass),
(uint64_t)(pInfo->framebuffer), (uint64_t)(fbRP), errorString.c_str());
}
}
}
if ((pInfo->occlusionQueryEnable == VK_FALSE || dev_data->physDevProperties.features.occlusionQueryPrecise == VK_FALSE) && (pInfo->queryFlags & VK_QUERY_CONTROL_PRECISE_BIT)) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, reinterpret_cast<uint64_t>(commandBuffer),
__LINE__, DRAWSTATE_BEGIN_CB_INVALID_STATE, "DS",
"vkBeginCommandBuffer(): Secondary Command Buffer (%p) must not have VK_QUERY_CONTROL_PRECISE_BIT if occulusionQuery is disabled or the device does not "
"support precise occlusion queries.", reinterpret_cast<void*>(commandBuffer));
}
}
if (pInfo && pInfo->renderPass != VK_NULL_HANDLE) {
auto rp_data = dev_data->renderPassMap.find(pInfo->renderPass);
if (rp_data != dev_data->renderPassMap.end() && rp_data->second && rp_data->second->pCreateInfo) {
if (pInfo->subpass >= rp_data->second->pCreateInfo->subpassCount) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, (uint64_t)commandBuffer,
__LINE__, DRAWSTATE_BEGIN_CB_INVALID_STATE, "DS",
"vkBeginCommandBuffer(): Secondary Command Buffers (%p) must has a subpass index (%d) that is less than the number of subpasses (%d).",
(void*)commandBuffer, pInfo->subpass, rp_data->second->pCreateInfo->subpassCount);
}
}
}
}
if (CB_RECORDING == pCB->state) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, (uint64_t)commandBuffer, __LINE__, DRAWSTATE_BEGIN_CB_INVALID_STATE, "DS",
"vkBeginCommandBuffer(): Cannot call Begin on CB (%#" PRIxLEAST64 ") in the RECORDING state. Must first call vkEndCommandBuffer().", (uint64_t)commandBuffer);
} else if (CB_RECORDED == pCB->state) {
VkCommandPool cmdPool = pCB->createInfo.commandPool;
if (!(VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT & dev_data->commandPoolMap[cmdPool].createFlags)) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, (uint64_t)commandBuffer,
__LINE__, DRAWSTATE_INVALID_COMMAND_BUFFER_RESET, "DS",
"Call to vkBeginCommandBuffer() on command buffer (%#" PRIxLEAST64 ") attempts to implicitly reset cmdBuffer created from command pool (%#" PRIxLEAST64 ") that does NOT have the VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT bit set.",
(uint64_t) commandBuffer, (uint64_t) cmdPool);
}
resetCB(dev_data, commandBuffer);
}
// Set updated state here in case implicit reset occurs above
pCB->state = CB_RECORDING;
pCB->beginInfo = *pBeginInfo;
if (pCB->beginInfo.pInheritanceInfo) {
pCB->inheritanceInfo = *(pCB->beginInfo.pInheritanceInfo);
pCB->beginInfo.pInheritanceInfo = &pCB->inheritanceInfo;
}
} else {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, (uint64_t)commandBuffer, __LINE__, DRAWSTATE_INVALID_COMMAND_BUFFER, "DS",
"In vkBeginCommandBuffer() and unable to find CommandBuffer Node for CB %p!", (void*)commandBuffer);
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE != skipCall) {
return VK_ERROR_VALIDATION_FAILED_EXT;
}
VkResult result = dev_data->device_dispatch_table->BeginCommandBuffer(commandBuffer, pBeginInfo);
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkEndCommandBuffer(VkCommandBuffer commandBuffer)
{
VkBool32 skipCall = VK_FALSE;
VkResult result = VK_SUCCESS;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
if (pCB->state != CB_RECORDING) {
skipCall |= report_error_no_cb_begin(dev_data, commandBuffer, "vkEndCommandBuffer()");
}
for (auto query : pCB->activeQueries) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_QUERY, "DS",
"Ending command buffer with in progress query: queryPool %" PRIu64 ", index %d", (uint64_t)(query.pool), query.index);
}
}
if (VK_FALSE == skipCall) {
loader_platform_thread_unlock_mutex(&globalLock);
result = dev_data->device_dispatch_table->EndCommandBuffer(commandBuffer);
loader_platform_thread_lock_mutex(&globalLock);
if (VK_SUCCESS == result) {
pCB->state = CB_RECORDED;
// Reset CB status flags
pCB->status = 0;
printCB(dev_data, commandBuffer);
}
} else {
result = VK_ERROR_VALIDATION_FAILED_EXT;
}
loader_platform_thread_unlock_mutex(&globalLock);
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkResetCommandBuffer(VkCommandBuffer commandBuffer, VkCommandBufferResetFlags flags)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
VkCommandPool cmdPool = pCB->createInfo.commandPool;
if (!(VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT & dev_data->commandPoolMap[cmdPool].createFlags)) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, (uint64_t) commandBuffer,
__LINE__, DRAWSTATE_INVALID_COMMAND_BUFFER_RESET, "DS",
"Attempt to reset command buffer (%#" PRIxLEAST64 ") created from command pool (%#" PRIxLEAST64 ") that does NOT have the VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT bit set.",
(uint64_t) commandBuffer, (uint64_t) cmdPool);
}
if (dev_data->globalInFlightCmdBuffers.count(commandBuffer)) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, (uint64_t) commandBuffer,
__LINE__, DRAWSTATE_INVALID_COMMAND_BUFFER_RESET, "DS",
"Attempt to reset command buffer (%#" PRIxLEAST64 ") which is in use.", reinterpret_cast<uint64_t>(commandBuffer));
}
loader_platform_thread_unlock_mutex(&globalLock);
if (skipCall != VK_FALSE)
return VK_ERROR_VALIDATION_FAILED_EXT;
VkResult result = dev_data->device_dispatch_table->ResetCommandBuffer(commandBuffer, flags);
if (VK_SUCCESS == result) {
loader_platform_thread_lock_mutex(&globalLock);
resetCB(dev_data, commandBuffer);
loader_platform_thread_unlock_mutex(&globalLock);
}
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdBindPipeline(VkCommandBuffer commandBuffer, VkPipelineBindPoint pipelineBindPoint, VkPipeline pipeline)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_BINDPIPELINE, "vkCmdBindPipeline()");
if ((VK_PIPELINE_BIND_POINT_COMPUTE == pipelineBindPoint) && (pCB->activeRenderPass)) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_PIPELINE_EXT, (uint64_t) pipeline,
__LINE__, DRAWSTATE_INVALID_RENDERPASS_CMD, "DS",
"Incorrectly binding compute pipeline (%#" PRIxLEAST64 ") during active RenderPass (%#" PRIxLEAST64 ")",
(uint64_t) pipeline, (uint64_t) pCB->activeRenderPass);
} else if (VK_PIPELINE_BIND_POINT_GRAPHICS == pipelineBindPoint) {
skipCall |= outsideRenderPass(dev_data, pCB, "vkCmdBindPipeline");
}
PIPELINE_NODE* pPN = getPipeline(dev_data, pipeline);
if (pPN) {
pCB->lastBoundPipeline = pipeline;
set_cb_pso_status(pCB, pPN);
skipCall |= validatePipelineState(dev_data, pCB, pipelineBindPoint, pipeline);
} else {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_PIPELINE_EXT,
(uint64_t) pipeline, __LINE__, DRAWSTATE_INVALID_PIPELINE, "DS",
"Attempt to bind Pipeline %#" PRIxLEAST64 " that doesn't exist!", (uint64_t)(pipeline));
}
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdBindPipeline(commandBuffer, pipelineBindPoint, pipeline);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdSetViewport(
VkCommandBuffer commandBuffer,
uint32_t firstViewport,
uint32_t viewportCount,
const VkViewport* pViewports)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_SETVIEWPORTSTATE, "vkCmdSetViewport()");
pCB->status |= CBSTATUS_VIEWPORT_SET;
pCB->viewports.resize(viewportCount);
memcpy(pCB->viewports.data(), pViewports, viewportCount * sizeof(VkViewport));
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdSetViewport(commandBuffer, firstViewport, viewportCount, pViewports);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdSetScissor(
VkCommandBuffer commandBuffer,
uint32_t firstScissor,
uint32_t scissorCount,
const VkRect2D* pScissors)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_SETSCISSORSTATE, "vkCmdSetScissor()");
pCB->status |= CBSTATUS_SCISSOR_SET;
pCB->scissors.resize(scissorCount);
memcpy(pCB->scissors.data(), pScissors, scissorCount * sizeof(VkRect2D));
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdSetScissor(commandBuffer, firstScissor, scissorCount, pScissors);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdSetLineWidth(VkCommandBuffer commandBuffer, float lineWidth)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_SETLINEWIDTHSTATE, "vkCmdSetLineWidth()");
pCB->status |= CBSTATUS_LINE_WIDTH_SET;
pCB->lineWidth = lineWidth;
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdSetLineWidth(commandBuffer, lineWidth);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdSetDepthBias(
VkCommandBuffer commandBuffer,
float depthBiasConstantFactor,
float depthBiasClamp,
float depthBiasSlopeFactor)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_SETDEPTHBIASSTATE, "vkCmdSetDepthBias()");
pCB->status |= CBSTATUS_DEPTH_BIAS_SET;
pCB->depthBiasConstantFactor = depthBiasConstantFactor;
pCB->depthBiasClamp = depthBiasClamp;
pCB->depthBiasSlopeFactor = depthBiasSlopeFactor;
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdSetDepthBias(commandBuffer, depthBiasConstantFactor, depthBiasClamp, depthBiasSlopeFactor);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdSetBlendConstants(VkCommandBuffer commandBuffer, const float blendConstants[4])
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_SETBLENDSTATE, "vkCmdSetBlendConstants()");
pCB->status |= CBSTATUS_BLEND_SET;
memcpy(pCB->blendConstants, blendConstants, 4 * sizeof(float));
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdSetBlendConstants(commandBuffer, blendConstants);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdSetDepthBounds(
VkCommandBuffer commandBuffer,
float minDepthBounds,
float maxDepthBounds)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_SETDEPTHBOUNDSSTATE, "vkCmdSetDepthBounds()");
pCB->status |= CBSTATUS_DEPTH_BOUNDS_SET;
pCB->minDepthBounds = minDepthBounds;
pCB->maxDepthBounds = maxDepthBounds;
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdSetDepthBounds(commandBuffer, minDepthBounds, maxDepthBounds);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdSetStencilCompareMask(
VkCommandBuffer commandBuffer,
VkStencilFaceFlags faceMask,
uint32_t compareMask)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_SETSTENCILREADMASKSTATE, "vkCmdSetStencilCompareMask()");
if (faceMask & VK_STENCIL_FACE_FRONT_BIT) {
pCB->front.compareMask = compareMask;
}
if (faceMask & VK_STENCIL_FACE_BACK_BIT) {
pCB->back.compareMask = compareMask;
}
/* TODO: Do we need to track front and back separately? */
/* TODO: We aren't capturing the faceMask, do we need to? */
pCB->status |= CBSTATUS_STENCIL_READ_MASK_SET;
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdSetStencilCompareMask(commandBuffer, faceMask, compareMask);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdSetStencilWriteMask(
VkCommandBuffer commandBuffer,
VkStencilFaceFlags faceMask,
uint32_t writeMask)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_SETSTENCILWRITEMASKSTATE, "vkCmdSetStencilWriteMask()");
if (faceMask & VK_STENCIL_FACE_FRONT_BIT) {
pCB->front.writeMask = writeMask;
}
if (faceMask & VK_STENCIL_FACE_BACK_BIT) {
pCB->back.writeMask = writeMask;
}
pCB->status |= CBSTATUS_STENCIL_WRITE_MASK_SET;
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdSetStencilWriteMask(commandBuffer, faceMask, writeMask);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdSetStencilReference(
VkCommandBuffer commandBuffer,
VkStencilFaceFlags faceMask,
uint32_t reference)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_SETSTENCILREFERENCESTATE, "vkCmdSetStencilReference()");
if (faceMask & VK_STENCIL_FACE_FRONT_BIT) {
pCB->front.reference = reference;
}
if (faceMask & VK_STENCIL_FACE_BACK_BIT) {
pCB->back.reference = reference;
}
pCB->status |= CBSTATUS_STENCIL_REFERENCE_SET;
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdSetStencilReference(commandBuffer, faceMask, reference);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdBindDescriptorSets(VkCommandBuffer commandBuffer, VkPipelineBindPoint pipelineBindPoint, VkPipelineLayout layout, uint32_t firstSet, uint32_t setCount, const VkDescriptorSet* pDescriptorSets, uint32_t dynamicOffsetCount, const uint32_t* pDynamicOffsets)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
if (pCB->state == CB_RECORDING) {
if ((VK_PIPELINE_BIND_POINT_COMPUTE == pipelineBindPoint) && (pCB->activeRenderPass)) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_INVALID_RENDERPASS_CMD, "DS",
"Incorrectly binding compute DescriptorSets during active RenderPass (%#" PRIxLEAST64 ")", (uint64_t) pCB->activeRenderPass);
} else if (VK_PIPELINE_BIND_POINT_GRAPHICS == pipelineBindPoint) {
skipCall |= outsideRenderPass(dev_data, pCB, "vkCmdBindDescriptorSets");
}
if (VK_FALSE == skipCall) {
// Track total count of dynamic descriptor types to make sure we have an offset for each one
uint32_t totalDynamicDescriptors = 0;
string errorString = "";
uint32_t lastSetIndex = firstSet+setCount-1;
if (lastSetIndex >= pCB->boundDescriptorSets.size())
pCB->boundDescriptorSets.resize(lastSetIndex+1);
VkDescriptorSet oldFinalBoundSet = pCB->boundDescriptorSets[lastSetIndex];
for (uint32_t i=0; i<setCount; i++) {
SET_NODE* pSet = getSetNode(dev_data, pDescriptorSets[i]);
if (pSet) {
pCB->uniqueBoundSets.insert(pDescriptorSets[i]);
pSet->boundCmdBuffers.insert(commandBuffer);
pCB->lastBoundDescriptorSet = pDescriptorSets[i];
pCB->lastBoundPipelineLayout = layout;
pCB->boundDescriptorSets[i+firstSet] = pDescriptorSets[i];
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT, (uint64_t) pDescriptorSets[i], __LINE__, DRAWSTATE_NONE, "DS",
"DS %#" PRIxLEAST64 " bound on pipeline %s", (uint64_t) pDescriptorSets[i], string_VkPipelineBindPoint(pipelineBindPoint));
if (!pSet->pUpdateStructs && (pSet->descriptorCount != 0)) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_WARNING_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT, (uint64_t) pDescriptorSets[i], __LINE__, DRAWSTATE_DESCRIPTOR_SET_NOT_UPDATED, "DS",
"DS %#" PRIxLEAST64 " bound but it was never updated. You may want to either update it or not bind it.", (uint64_t) pDescriptorSets[i]);
}
// Verify that set being bound is compatible with overlapping setLayout of pipelineLayout
if (!verify_set_layout_compatibility(dev_data, pSet, layout, i+firstSet, errorString)) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT, (uint64_t) pDescriptorSets[i], __LINE__, DRAWSTATE_PIPELINE_LAYOUTS_INCOMPATIBLE, "DS",
"descriptorSet #%u being bound is not compatible with overlapping layout in pipelineLayout due to: %s", i, errorString.c_str());
}
if (pSet->pLayout->dynamicDescriptorCount) {
// First make sure we won't overstep bounds of pDynamicOffsets array
if ((totalDynamicDescriptors + pSet->pLayout->dynamicDescriptorCount) > dynamicOffsetCount) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT, (uint64_t) pDescriptorSets[i], __LINE__, DRAWSTATE_INVALID_DYNAMIC_OFFSET_COUNT, "DS",
"descriptorSet #%u (%#" PRIxLEAST64 ") requires %u dynamicOffsets, but only %u dynamicOffsets are left in pDynamicOffsets array. There must be one dynamic offset for each dynamic descriptor being bound.",
i, (uint64_t) pDescriptorSets[i], pSet->pLayout->dynamicDescriptorCount, (dynamicOffsetCount - totalDynamicDescriptors));
} else { // Validate and store dynamic offsets with the set
// Validate Dynamic Offset Minimums
uint32_t cur_dyn_offset = totalDynamicDescriptors;
for (uint32_t d = 0; d < pSet->descriptorCount; d++) {
if (pSet->pLayout->descriptorTypes[d] == VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC) {
if (vk_safe_modulo(pDynamicOffsets[cur_dyn_offset], dev_data->physDevProperties.properties.limits.minUniformBufferOffsetAlignment) != 0) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_PHYSICAL_DEVICE_EXT, 0,
__LINE__, DRAWSTATE_INVALID_UNIFORM_BUFFER_OFFSET, "DS",
"vkCmdBindDescriptorSets(): pDynamicOffsets[%d] is %d but must be a multiple of device limit minUniformBufferOffsetAlignment %#" PRIxLEAST64,
cur_dyn_offset, pDynamicOffsets[cur_dyn_offset], dev_data->physDevProperties.properties.limits.minUniformBufferOffsetAlignment);
}
cur_dyn_offset++;
} else if (pSet->pLayout->descriptorTypes[d] == VK_DESCRIPTOR_TYPE_STORAGE_BUFFER_DYNAMIC) {
if (vk_safe_modulo(pDynamicOffsets[cur_dyn_offset], dev_data->physDevProperties.properties.limits.minStorageBufferOffsetAlignment) != 0) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_PHYSICAL_DEVICE_EXT, 0,
__LINE__, DRAWSTATE_INVALID_STORAGE_BUFFER_OFFSET, "DS",
"vkCmdBindDescriptorSets(): pDynamicOffsets[%d] is %d but must be a multiple of device limit minStorageBufferOffsetAlignment %#" PRIxLEAST64,
cur_dyn_offset, pDynamicOffsets[cur_dyn_offset], dev_data->physDevProperties.properties.limits.minStorageBufferOffsetAlignment);
}
cur_dyn_offset++;
}
}
// Keep running total of dynamic descriptor count to verify at the end
totalDynamicDescriptors += pSet->pLayout->dynamicDescriptorCount;
}
}
} else {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT, (uint64_t) pDescriptorSets[i], __LINE__, DRAWSTATE_INVALID_SET, "DS",
"Attempt to bind DS %#" PRIxLEAST64 " that doesn't exist!", (uint64_t) pDescriptorSets[i]);
}
}
skipCall |= addCmd(dev_data, pCB, CMD_BINDDESCRIPTORSETS, "vkCmdBindDescrsiptorSets()");
// For any previously bound sets, need to set them to "invalid" if they were disturbed by this update
if (firstSet > 0) { // Check set #s below the first bound set
for (uint32_t i=0; i<firstSet; ++i) {
if (pCB->boundDescriptorSets[i] && !verify_set_layout_compatibility(dev_data, dev_data->setMap[pCB->boundDescriptorSets[i]], layout, i, errorString)) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT, (uint64_t) pCB->boundDescriptorSets[i], __LINE__, DRAWSTATE_NONE, "DS",
"DescriptorSetDS %#" PRIxLEAST64 " previously bound as set #%u was disturbed by newly bound pipelineLayout (%#" PRIxLEAST64 ")", (uint64_t) pCB->boundDescriptorSets[i], i, (uint64_t) layout);
pCB->boundDescriptorSets[i] = VK_NULL_HANDLE;
}
}
}
// Check if newly last bound set invalidates any remaining bound sets
if ((pCB->boundDescriptorSets.size()-1) > (lastSetIndex)) {
if (oldFinalBoundSet && !verify_set_layout_compatibility(dev_data, dev_data->setMap[oldFinalBoundSet], layout, lastSetIndex, errorString)) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DESCRIPTOR_SET_EXT, (uint64_t) oldFinalBoundSet, __LINE__, DRAWSTATE_NONE, "DS",
"DescriptorSetDS %#" PRIxLEAST64 " previously bound as set #%u is incompatible with set %#" PRIxLEAST64 " newly bound as set #%u so set #%u and any subsequent sets were disturbed by newly bound pipelineLayout (%#" PRIxLEAST64 ")", (uint64_t) oldFinalBoundSet, lastSetIndex, (uint64_t) pCB->boundDescriptorSets[lastSetIndex], lastSetIndex, lastSetIndex+1, (uint64_t) layout);
pCB->boundDescriptorSets.resize(lastSetIndex+1);
}
}
// dynamicOffsetCount must equal the total number of dynamic descriptors in the sets being bound
if (totalDynamicDescriptors != dynamicOffsetCount) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, (uint64_t) commandBuffer, __LINE__, DRAWSTATE_INVALID_DYNAMIC_OFFSET_COUNT, "DS",
"Attempting to bind %u descriptorSets with %u dynamic descriptors, but dynamicOffsetCount is %u. It should exactly match the number of dynamic descriptors.", setCount, totalDynamicDescriptors, dynamicOffsetCount);
}
if (dynamicOffsetCount) {
// Save dynamicOffsets bound to this CB
pCB->dynamicOffsets.assign(pDynamicOffsets, pDynamicOffsets + dynamicOffsetCount);
}
}
} else {
skipCall |= report_error_no_cb_begin(dev_data, commandBuffer, "vkCmdBindDescriptorSets()");
}
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdBindDescriptorSets(commandBuffer, pipelineBindPoint, layout, firstSet, setCount, pDescriptorSets, dynamicOffsetCount, pDynamicOffsets);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdBindIndexBuffer(VkCommandBuffer commandBuffer, VkBuffer buffer, VkDeviceSize offset, VkIndexType indexType)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_BINDINDEXBUFFER, "vkCmdBindIndexBuffer()");
VkDeviceSize offset_align = 0;
switch (indexType) {
case VK_INDEX_TYPE_UINT16:
offset_align = 2;
break;
case VK_INDEX_TYPE_UINT32:
offset_align = 4;
break;
default:
// ParamChecker should catch bad enum, we'll also throw alignment error below if offset_align stays 0
break;
}
if (!offset_align || (offset % offset_align)) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_VTX_INDEX_ALIGNMENT_ERROR, "DS",
"vkCmdBindIndexBuffer() offset (%#" PRIxLEAST64 ") does not fall on alignment (%s) boundary.", offset, string_VkIndexType(indexType));
}
pCB->status |= CBSTATUS_INDEX_BUFFER_BOUND;
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdBindIndexBuffer(commandBuffer, buffer, offset, indexType);
}
void updateResourceTracking(GLOBAL_CB_NODE* pCB, uint32_t firstBinding, uint32_t bindingCount, const VkBuffer* pBuffers) {
uint32_t end = firstBinding + bindingCount;
if (pCB->currentDrawData.buffers.size() < end) {
pCB->currentDrawData.buffers.resize(end);
}
for (uint32_t i = 0; i < bindingCount; ++i) {
pCB->currentDrawData.buffers[i + firstBinding] = pBuffers[i];
}
}
void updateResourceTrackingOnDraw(GLOBAL_CB_NODE* pCB) {
pCB->drawData.push_back(pCB->currentDrawData);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdBindVertexBuffers(
VkCommandBuffer commandBuffer,
uint32_t firstBinding,
uint32_t bindingCount,
const VkBuffer *pBuffers,
const VkDeviceSize *pOffsets)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
addCmd(dev_data, pCB, CMD_BINDVERTEXBUFFER, "vkCmdBindVertexBuffer()");
updateResourceTracking(pCB, firstBinding, bindingCount, pBuffers);
} else {
skipCall |= report_error_no_cb_begin(dev_data, commandBuffer, "vkCmdBindVertexBuffer()");
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdBindVertexBuffers(commandBuffer, firstBinding, bindingCount, pBuffers, pOffsets);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdDraw(VkCommandBuffer commandBuffer, uint32_t vertexCount, uint32_t instanceCount, uint32_t firstVertex, uint32_t firstInstance)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_DRAW, "vkCmdDraw()");
pCB->drawCount[DRAW]++;
skipCall |= validate_draw_state(dev_data, pCB, VK_FALSE);
// TODO : Need to pass commandBuffer as srcObj here
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, 0, __LINE__, DRAWSTATE_NONE, "DS",
"vkCmdDraw() call #%" PRIu64 ", reporting DS state:", g_drawCount[DRAW]++);
skipCall |= synchAndPrintDSConfig(dev_data, commandBuffer);
if (VK_FALSE == skipCall) {
updateResourceTrackingOnDraw(pCB);
}
skipCall |= outsideRenderPass(dev_data, pCB, "vkCmdDraw");
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdDraw(commandBuffer, vertexCount, instanceCount, firstVertex, firstInstance);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdDrawIndexed(VkCommandBuffer commandBuffer, uint32_t indexCount, uint32_t instanceCount, uint32_t firstIndex, int32_t vertexOffset, uint32_t firstInstance)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
VkBool32 skipCall = VK_FALSE;
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_DRAWINDEXED, "vkCmdDrawIndexed()");
pCB->drawCount[DRAW_INDEXED]++;
skipCall |= validate_draw_state(dev_data, pCB, VK_TRUE);
// TODO : Need to pass commandBuffer as srcObj here
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, 0, __LINE__, DRAWSTATE_NONE, "DS",
"vkCmdDrawIndexed() call #%" PRIu64 ", reporting DS state:", g_drawCount[DRAW_INDEXED]++);
skipCall |= synchAndPrintDSConfig(dev_data, commandBuffer);
if (VK_FALSE == skipCall) {
updateResourceTrackingOnDraw(pCB);
}
skipCall |= outsideRenderPass(dev_data, pCB, "vkCmdDrawIndexed");
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdDrawIndexed(commandBuffer, indexCount, instanceCount, firstIndex, vertexOffset, firstInstance);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdDrawIndirect(VkCommandBuffer commandBuffer, VkBuffer buffer, VkDeviceSize offset, uint32_t count, uint32_t stride)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
VkBool32 skipCall = VK_FALSE;
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_DRAWINDIRECT, "vkCmdDrawIndirect()");
pCB->drawCount[DRAW_INDIRECT]++;
skipCall |= validate_draw_state(dev_data, pCB, VK_FALSE);
// TODO : Need to pass commandBuffer as srcObj here
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, 0, __LINE__, DRAWSTATE_NONE, "DS",
"vkCmdDrawIndirect() call #%" PRIu64 ", reporting DS state:", g_drawCount[DRAW_INDIRECT]++);
skipCall |= synchAndPrintDSConfig(dev_data, commandBuffer);
if (VK_FALSE == skipCall) {
updateResourceTrackingOnDraw(pCB);
}
skipCall |= outsideRenderPass(dev_data, pCB, "vkCmdDrawIndirect");
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdDrawIndirect(commandBuffer, buffer, offset, count, stride);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdDrawIndexedIndirect(VkCommandBuffer commandBuffer, VkBuffer buffer, VkDeviceSize offset, uint32_t count, uint32_t stride)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_DRAWINDEXEDINDIRECT, "vkCmdDrawIndexedIndirect()");
pCB->drawCount[DRAW_INDEXED_INDIRECT]++;
loader_platform_thread_unlock_mutex(&globalLock);
skipCall |= validate_draw_state(dev_data, pCB, VK_TRUE);
loader_platform_thread_lock_mutex(&globalLock);
// TODO : Need to pass commandBuffer as srcObj here
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_INFORMATION_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, 0, __LINE__, DRAWSTATE_NONE, "DS",
"vkCmdDrawIndexedIndirect() call #%" PRIu64 ", reporting DS state:", g_drawCount[DRAW_INDEXED_INDIRECT]++);
skipCall |= synchAndPrintDSConfig(dev_data, commandBuffer);
if (VK_FALSE == skipCall) {
updateResourceTrackingOnDraw(pCB);
}
skipCall |= outsideRenderPass(dev_data, pCB, "vkCmdDrawIndexedIndirect");
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdDrawIndexedIndirect(commandBuffer, buffer, offset, count, stride);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdDispatch(VkCommandBuffer commandBuffer, uint32_t x, uint32_t y, uint32_t z)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_DISPATCH, "vkCmdDispatch()");
skipCall |= insideRenderPass(dev_data, pCB, "vkCmdDispatch");
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdDispatch(commandBuffer, x, y, z);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdDispatchIndirect(VkCommandBuffer commandBuffer, VkBuffer buffer, VkDeviceSize offset)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_DISPATCHINDIRECT, "vkCmdDispatchIndirect()");
skipCall |= insideRenderPass(dev_data, pCB, "vkCmdDispatchIndirect");
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdDispatchIndirect(commandBuffer, buffer, offset);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdCopyBuffer(VkCommandBuffer commandBuffer, VkBuffer srcBuffer, VkBuffer dstBuffer, uint32_t regionCount, const VkBufferCopy* pRegions)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_COPYBUFFER, "vkCmdCopyBuffer()");
skipCall |= insideRenderPass(dev_data, pCB, "vkCmdCopyBuffer");
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdCopyBuffer(commandBuffer, srcBuffer, dstBuffer, regionCount, pRegions);
}
VkBool32 VerifySourceImageLayout(VkCommandBuffer cmdBuffer, VkImage srcImage, VkImageSubresourceLayers subLayers, VkImageLayout srcImageLayout) {
VkBool32 skip_call = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(cmdBuffer), layer_data_map);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, cmdBuffer);
for (uint32_t i = 0; i < subLayers.layerCount; ++i) {
uint32_t layer = i + subLayers.baseArrayLayer;
VkImageSubresource sub = {subLayers.aspectMask, subLayers.mipLevel, layer};
IMAGE_CMD_BUF_NODE node;
if (!FindLayout(pCB, srcImage, sub, node)) {
SetLayout(pCB, srcImage, sub, {srcImageLayout, srcImageLayout});
continue;
}
if (node.layout != srcImageLayout) {
// TODO: Improve log message in the next pass
skip_call |=
log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, 0,
__LINE__, DRAWSTATE_INVALID_IMAGE_LAYOUT, "DS",
"Cannot copy from an image whose source layout is %s "
"and doesn't match the current layout %s.",
string_VkImageLayout(srcImageLayout),
string_VkImageLayout(node.layout));
}
}
if (srcImageLayout != VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL) {
if (srcImageLayout == VK_IMAGE_LAYOUT_GENERAL) {
// LAYOUT_GENERAL is allowed, but may not be performance optimal, flag as perf warning.
skip_call |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_IMAGE_LAYOUT, "DS",
"Layout for input image should be TRANSFER_SRC_OPTIMAL instead of GENERAL.");
} else {
skip_call |=
log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
(VkDebugReportObjectTypeEXT)0, 0, __LINE__,
DRAWSTATE_INVALID_IMAGE_LAYOUT, "DS",
"Layout for input image is %s but can only be "
"TRANSFER_SRC_OPTIMAL or GENERAL.",
string_VkImageLayout(srcImageLayout));
}
}
return skip_call;
}
VkBool32 VerifyDestImageLayout(VkCommandBuffer cmdBuffer, VkImage destImage, VkImageSubresourceLayers subLayers, VkImageLayout destImageLayout) {
VkBool32 skip_call = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(cmdBuffer), layer_data_map);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, cmdBuffer);
for (uint32_t i = 0; i < subLayers.layerCount; ++i) {
uint32_t layer = i + subLayers.baseArrayLayer;
VkImageSubresource sub = {subLayers.aspectMask, subLayers.mipLevel, layer};
IMAGE_CMD_BUF_NODE node;
if (!FindLayout(pCB, destImage, sub, node)) {
SetLayout(pCB, destImage, sub, {destImageLayout, destImageLayout});
continue;
}
if (node.layout != destImageLayout) {
skip_call |=
log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, 0,
__LINE__, DRAWSTATE_INVALID_IMAGE_LAYOUT, "DS",
"Cannot copy from an image whose dest layout is %s and "
"doesn't match the current layout %s.",
string_VkImageLayout(destImageLayout),
string_VkImageLayout(node.layout));
}
}
if (destImageLayout != VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL) {
if (destImageLayout == VK_IMAGE_LAYOUT_GENERAL) {
// LAYOUT_GENERAL is allowed, but may not be performance optimal, flag as perf warning.
skip_call |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_PERFORMANCE_WARNING_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_IMAGE_LAYOUT, "DS",
"Layout for output image should be TRANSFER_DST_OPTIMAL instead of GENERAL.");
} else {
skip_call |=
log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
(VkDebugReportObjectTypeEXT)0, 0, __LINE__,
DRAWSTATE_INVALID_IMAGE_LAYOUT, "DS",
"Layout for output image is %s but can only be "
"TRANSFER_DST_OPTIMAL or GENERAL.",
string_VkImageLayout(destImageLayout));
}
}
return skip_call;
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdCopyImage(VkCommandBuffer commandBuffer,
VkImage srcImage,
VkImageLayout srcImageLayout,
VkImage dstImage,
VkImageLayout dstImageLayout,
uint32_t regionCount, const VkImageCopy* pRegions)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_COPYIMAGE, "vkCmdCopyImage()");
skipCall |= insideRenderPass(dev_data, pCB, "vkCmdCopyImage");
for (uint32_t i = 0; i < regionCount; ++i) {
skipCall |= VerifySourceImageLayout(commandBuffer, srcImage, pRegions[i].srcSubresource, srcImageLayout);
skipCall |= VerifyDestImageLayout(commandBuffer, dstImage, pRegions[i].dstSubresource, dstImageLayout);
}
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdCopyImage(commandBuffer, srcImage, srcImageLayout, dstImage, dstImageLayout, regionCount, pRegions);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdBlitImage(VkCommandBuffer commandBuffer,
VkImage srcImage, VkImageLayout srcImageLayout,
VkImage dstImage, VkImageLayout dstImageLayout,
uint32_t regionCount, const VkImageBlit* pRegions,
VkFilter filter)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_BLITIMAGE, "vkCmdBlitImage()");
skipCall |= insideRenderPass(dev_data, pCB, "vkCmdBlitImage");
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdBlitImage(commandBuffer, srcImage, srcImageLayout, dstImage, dstImageLayout, regionCount, pRegions, filter);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdCopyBufferToImage(VkCommandBuffer commandBuffer,
VkBuffer srcBuffer,
VkImage dstImage, VkImageLayout dstImageLayout,
uint32_t regionCount, const VkBufferImageCopy* pRegions)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_COPYBUFFERTOIMAGE, "vkCmdCopyBufferToImage()");
skipCall |= insideRenderPass(dev_data, pCB, "vkCmdCopyBufferToImage");
for (uint32_t i = 0; i < regionCount; ++i) {
skipCall |= VerifySourceImageLayout(commandBuffer, dstImage,
pRegions[i].imageSubresource,
dstImageLayout);
}
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdCopyBufferToImage(commandBuffer, srcBuffer, dstImage, dstImageLayout, regionCount, pRegions);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdCopyImageToBuffer(VkCommandBuffer commandBuffer,
VkImage srcImage, VkImageLayout srcImageLayout,
VkBuffer dstBuffer,
uint32_t regionCount, const VkBufferImageCopy* pRegions)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_COPYIMAGETOBUFFER, "vkCmdCopyImageToBuffer()");
skipCall |= insideRenderPass(dev_data, pCB, "vkCmdCopyImageToBuffer");
for (uint32_t i = 0; i < regionCount; ++i) {
skipCall |= VerifySourceImageLayout(commandBuffer, srcImage,
pRegions[i].imageSubresource,
srcImageLayout);
}
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdCopyImageToBuffer(commandBuffer, srcImage, srcImageLayout, dstBuffer, regionCount, pRegions);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdUpdateBuffer(VkCommandBuffer commandBuffer, VkBuffer dstBuffer, VkDeviceSize dstOffset, VkDeviceSize dataSize, const uint32_t* pData)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_UPDATEBUFFER, "vkCmdUpdateBuffer()");
skipCall |= insideRenderPass(dev_data, pCB, "vkCmdCopyUpdateBuffer");
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdUpdateBuffer(commandBuffer, dstBuffer, dstOffset, dataSize, pData);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdFillBuffer(VkCommandBuffer commandBuffer, VkBuffer dstBuffer, VkDeviceSize dstOffset, VkDeviceSize size, uint32_t data)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_FILLBUFFER, "vkCmdFillBuffer()");
skipCall |= insideRenderPass(dev_data, pCB, "vkCmdCopyFillBuffer");
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdFillBuffer(commandBuffer, dstBuffer, dstOffset, size, data);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdClearAttachments(
VkCommandBuffer commandBuffer,
uint32_t attachmentCount,
const VkClearAttachment* pAttachments,
uint32_t rectCount,
const VkClearRect* pRects)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_CLEARATTACHMENTS, "vkCmdClearAttachments()");
// Warn if this is issued prior to Draw Cmd and clearing the entire attachment
if (!hasDrawCmd(pCB) &&
(pCB->activeRenderPassBeginInfo.renderArea.extent.width == pRects[0].rect.extent.width) &&
(pCB->activeRenderPassBeginInfo.renderArea.extent.height == pRects[0].rect.extent.height)) {
// TODO : commandBuffer should be srcObj
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_WARNING_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, 0, __LINE__, DRAWSTATE_CLEAR_CMD_BEFORE_DRAW, "DS",
"vkCmdClearAttachments() issued on CB object 0x%" PRIxLEAST64 " prior to any Draw Cmds."
" It is recommended you use RenderPass LOAD_OP_CLEAR on Attachments prior to any Draw.", (uint64_t)(commandBuffer));
}
skipCall |= outsideRenderPass(dev_data, pCB, "vkCmdClearAttachments");
}
// Validate that attachment is in reference list of active subpass
if (pCB->activeRenderPass) {
const VkRenderPassCreateInfo *pRPCI = dev_data->renderPassMap[pCB->activeRenderPass]->pCreateInfo;
const VkSubpassDescription *pSD = &pRPCI->pSubpasses[pCB->activeSubpass];
for (uint32_t attachment_idx = 0; attachment_idx < attachmentCount; attachment_idx++) {
const VkClearAttachment *attachment = &pAttachments[attachment_idx];
if (attachment->aspectMask & VK_IMAGE_ASPECT_COLOR_BIT) {
VkBool32 found = VK_FALSE;
for (uint32_t i = 0; i < pSD->colorAttachmentCount; i++) {
if (attachment->colorAttachment == pSD->pColorAttachments[i].attachment) {
found = VK_TRUE;
break;
}
}
if (VK_FALSE == found) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT,
(uint64_t)commandBuffer, __LINE__, DRAWSTATE_MISSING_ATTACHMENT_REFERENCE, "DS",
"vkCmdClearAttachments() attachment index %d not found in attachment reference array of active subpass %d",
attachment->colorAttachment, pCB->activeSubpass);
}
} else if (attachment->aspectMask & (VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT)) {
if (!pSD->pDepthStencilAttachment || // Says no DS will be used in active subpass
(pSD->pDepthStencilAttachment->attachment == VK_ATTACHMENT_UNUSED)) { // Says no DS will be used in active subpass
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT,
(uint64_t)commandBuffer, __LINE__, DRAWSTATE_MISSING_ATTACHMENT_REFERENCE, "DS",
"vkCmdClearAttachments() attachment index %d does not match depthStencilAttachment.attachment (%d) found in active subpass %d",
attachment->colorAttachment,
(pSD->pDepthStencilAttachment) ? pSD->pDepthStencilAttachment->attachment : VK_ATTACHMENT_UNUSED,
pCB->activeSubpass);
}
}
}
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdClearAttachments(commandBuffer, attachmentCount, pAttachments, rectCount, pRects);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdClearColorImage(
VkCommandBuffer commandBuffer,
VkImage image, VkImageLayout imageLayout,
const VkClearColorValue *pColor,
uint32_t rangeCount, const VkImageSubresourceRange* pRanges)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_CLEARCOLORIMAGE, "vkCmdClearColorImage()");
skipCall |= insideRenderPass(dev_data, pCB, "vkCmdClearColorImage");
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdClearColorImage(commandBuffer, image, imageLayout, pColor, rangeCount, pRanges);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdClearDepthStencilImage(
VkCommandBuffer commandBuffer,
VkImage image, VkImageLayout imageLayout,
const VkClearDepthStencilValue *pDepthStencil,
uint32_t rangeCount,
const VkImageSubresourceRange* pRanges)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_CLEARDEPTHSTENCILIMAGE, "vkCmdClearDepthStencilImage()");
skipCall |= insideRenderPass(dev_data, pCB, "vkCmdClearDepthStencilImage");
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdClearDepthStencilImage(commandBuffer, image, imageLayout, pDepthStencil, rangeCount, pRanges);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdResolveImage(VkCommandBuffer commandBuffer,
VkImage srcImage, VkImageLayout srcImageLayout,
VkImage dstImage, VkImageLayout dstImageLayout,
uint32_t regionCount, const VkImageResolve* pRegions)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_RESOLVEIMAGE, "vkCmdResolveImage()");
skipCall |= insideRenderPass(dev_data, pCB, "vkCmdResolveImage");
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdResolveImage(commandBuffer, srcImage, srcImageLayout, dstImage, dstImageLayout, regionCount, pRegions);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdSetEvent(VkCommandBuffer commandBuffer, VkEvent event, VkPipelineStageFlags stageMask)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_SETEVENT, "vkCmdSetEvent()");
skipCall |= insideRenderPass(dev_data, pCB, "vkCmdSetEvent");
pCB->events.push_back(event);
pCB->eventToStageMap[event] = stageMask;
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdSetEvent(commandBuffer, event, stageMask);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdResetEvent(VkCommandBuffer commandBuffer, VkEvent event, VkPipelineStageFlags stageMask)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_RESETEVENT, "vkCmdResetEvent()");
skipCall |= insideRenderPass(dev_data, pCB, "vkCmdResetEvent");
pCB->events.push_back(event);
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdResetEvent(commandBuffer, event, stageMask);
}
VkBool32 TransitionImageLayouts(VkCommandBuffer cmdBuffer, uint32_t memBarrierCount, const VkImageMemoryBarrier* pImgMemBarriers) {
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(cmdBuffer), layer_data_map);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, cmdBuffer);
VkBool32 skip = VK_FALSE;
for (uint32_t i = 0; i < memBarrierCount; ++i) {
auto mem_barrier = &pImgMemBarriers[i];
if (!mem_barrier)
continue;
// TODO: Do not iterate over every possibility - consolidate where
// possible
for (uint32_t j = 0; j < mem_barrier->subresourceRange.levelCount;
j++) {
uint32_t level = mem_barrier->subresourceRange.baseMipLevel + j;
for (uint32_t k = 0; k < mem_barrier->subresourceRange.layerCount;
k++) {
uint32_t layer =
mem_barrier->subresourceRange.baseArrayLayer + k;
VkImageSubresource sub = {
mem_barrier->subresourceRange.aspectMask, level, layer};
IMAGE_CMD_BUF_NODE node;
if (!FindLayout(pCB, mem_barrier->image, sub, node)) {
SetLayout(pCB, mem_barrier->image, sub,
{mem_barrier->oldLayout, mem_barrier->newLayout});
continue;
}
if (node.layout != mem_barrier->oldLayout) {
skip |= log_msg(
dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
(VkDebugReportObjectTypeEXT)0, 0, __LINE__,
DRAWSTATE_INVALID_IMAGE_LAYOUT, "DS",
"You cannot transition the layout from %s "
"when current layout is %s.",
string_VkImageLayout(mem_barrier->oldLayout),
string_VkImageLayout(node.layout));
}
SetLayout(pCB, mem_barrier->image, sub, mem_barrier->newLayout);
}
}
}
return skip;
}
// Print readable FlagBits in FlagMask
std::string string_VkAccessFlags(VkAccessFlags accessMask)
{
std::string result;
std::string separator;
if (accessMask == 0) {
result = "[None]";
} else {
result = "[";
for (auto i = 0; i < 32; i++) {
if (accessMask & (1 << i)) {
result = result + separator + string_VkAccessFlagBits((VkAccessFlagBits)(1 << i));
separator = " | ";
}
}
result = result + "]";
}
return result;
}
// AccessFlags MUST have 'required_bit' set, and may have one or more of 'optional_bits' set.
// If required_bit is zero, accessMask must have at least one of 'optional_bits' set
// TODO: Add tracking to ensure that at least one barrier has been set for these layout transitions
VkBool32 ValidateMaskBits(const layer_data* my_data, VkCommandBuffer cmdBuffer, const VkAccessFlags& accessMask, const VkImageLayout& layout,
VkAccessFlags required_bit, VkAccessFlags optional_bits, const char* type) {
VkBool32 skip_call = VK_FALSE;
if ((accessMask & required_bit) || (!required_bit && (accessMask & optional_bits))) {
if (accessMask & !(required_bit | optional_bits)) {
// TODO: Verify against Valid Use
skip_call |= log_msg(my_data->report_data, VK_DEBUG_REPORT_WARNING_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_BARRIER, "DS",
"Additional bits in %s accessMask %d %s are specified when layout is %s.",
type, accessMask, string_VkAccessFlags(accessMask).c_str(), string_VkImageLayout(layout));
}
} else {
if (!required_bit) {
skip_call |= log_msg(my_data->report_data, VK_DEBUG_REPORT_WARNING_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_BARRIER, "DS",
"%s AccessMask %d %s must contain at least one of access bits %d %s when layout is %s, unless the app has previously added a barrier for this transition.",
type, accessMask, string_VkAccessFlags(accessMask).c_str(), optional_bits,
string_VkAccessFlags(optional_bits).c_str(), string_VkImageLayout(layout));
} else {
std::string opt_bits;
if (optional_bits != 0) {
std::stringstream ss;
ss << optional_bits;
opt_bits = "and may have optional bits " + ss.str() + ' ' + string_VkAccessFlags(optional_bits);
}
skip_call |= log_msg(my_data->report_data, VK_DEBUG_REPORT_WARNING_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_BARRIER, "DS",
"%s AccessMask %d %s must have required access bit %d %s %s when layout is %s, unless the app has previously added a barrier for this transition.",
type, accessMask, string_VkAccessFlags(accessMask).c_str(),
required_bit, string_VkAccessFlags(required_bit).c_str(),
opt_bits.c_str(), string_VkImageLayout(layout));
}
}
return skip_call;
}
VkBool32 ValidateMaskBitsFromLayouts(const layer_data* my_data, VkCommandBuffer cmdBuffer, const VkAccessFlags& accessMask, const VkImageLayout& layout, const char* type) {
VkBool32 skip_call = VK_FALSE;
switch (layout) {
case VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL: {
skip_call |= ValidateMaskBits(my_data, cmdBuffer, accessMask, layout, VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT, VK_ACCESS_COLOR_ATTACHMENT_READ_BIT, type);
break;
}
case VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL: {
skip_call |= ValidateMaskBits(my_data, cmdBuffer, accessMask, layout, VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT, VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT, type);
break;
}
case VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL: {
skip_call |= ValidateMaskBits(my_data, cmdBuffer, accessMask, layout, VK_ACCESS_TRANSFER_WRITE_BIT, 0, type);
break;
}
case VK_IMAGE_LAYOUT_PREINITIALIZED: {
skip_call |= ValidateMaskBits(my_data, cmdBuffer, accessMask, layout, VK_ACCESS_HOST_WRITE_BIT, 0, type);
break;
}
case VK_IMAGE_LAYOUT_DEPTH_STENCIL_READ_ONLY_OPTIMAL: {
skip_call |= ValidateMaskBits(my_data, cmdBuffer, accessMask, layout, 0, VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT | VK_ACCESS_SHADER_READ_BIT, type);
break;
}
case VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL: {
skip_call |= ValidateMaskBits(my_data, cmdBuffer, accessMask, layout, 0, VK_ACCESS_INPUT_ATTACHMENT_READ_BIT | VK_ACCESS_SHADER_READ_BIT, type);
break;
}
case VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL: {
skip_call |= ValidateMaskBits(my_data, cmdBuffer, accessMask, layout, VK_ACCESS_TRANSFER_READ_BIT, 0, type);
break;
}
case VK_IMAGE_LAYOUT_UNDEFINED: {
if (accessMask != 0) {
// TODO: Verify against Valid Use section spec
skip_call |= log_msg(my_data->report_data, VK_DEBUG_REPORT_WARNING_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_BARRIER, "DS",
"Additional bits in %s accessMask %d %s are specified when layout is %s.", type, accessMask, string_VkAccessFlags(accessMask).c_str(),
string_VkImageLayout(layout));
}
break;
}
case VK_IMAGE_LAYOUT_GENERAL:
default: {
break;
}
}
return skip_call;
}
VkBool32 ValidateBarriers(VkCommandBuffer cmdBuffer, uint32_t memBarrierCount,
const VkMemoryBarrier *pMemBarriers,
uint32_t bufferBarrierCount,
const VkBufferMemoryBarrier *pBufferMemBarriers,
uint32_t imageMemBarrierCount,
const VkImageMemoryBarrier *pImageMemBarriers) {
VkBool32 skip_call = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(cmdBuffer), layer_data_map);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, cmdBuffer);
if (pCB->activeRenderPass && memBarrierCount) {
if (!dev_data->renderPassMap[pCB->activeRenderPass]->hasSelfDependency[pCB->activeSubpass]) {
skip_call |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_BARRIER, "DS",
"Barriers cannot be set during subpass %d with no self dependency specified.", pCB->activeSubpass);
}
}
for (uint32_t i = 0; i < imageMemBarrierCount; ++i) {
auto mem_barrier = &pImageMemBarriers[i];
if (pCB->activeRenderPass) {
skip_call |=
log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
(VkDebugReportObjectTypeEXT)0, 0, __LINE__,
DRAWSTATE_INVALID_BARRIER, "DS",
"Image Barriers cannot be used during a render pass.");
}
if (mem_barrier && mem_barrier->sType == VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER) {
skip_call |= ValidateMaskBitsFromLayouts(dev_data, cmdBuffer, mem_barrier->srcAccessMask, mem_barrier->oldLayout, "Source");
skip_call |= ValidateMaskBitsFromLayouts(dev_data, cmdBuffer, mem_barrier->dstAccessMask, mem_barrier->newLayout, "Dest");
}
}
for (uint32_t i = 0; i < bufferBarrierCount; ++i) {
auto mem_barrier = &pBufferMemBarriers[i];
if (pCB->activeRenderPass) {
skip_call |=
log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
(VkDebugReportObjectTypeEXT)0, 0, __LINE__,
DRAWSTATE_INVALID_BARRIER, "DS",
"Buffer Barriers cannot be used during a render pass.");
}
if (!mem_barrier)
continue;
auto buffer_data = dev_data->bufferMap.find(mem_barrier->buffer);
uint64_t buffer_size = buffer_data->second.create_info
? reinterpret_cast<uint64_t &>(
buffer_data->second.create_info->size)
: 0;
if (buffer_data != dev_data->bufferMap.end() &&
mem_barrier->offset + mem_barrier->size > buffer_size) {
skip_call |=
log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
(VkDebugReportObjectTypeEXT)0, 0, __LINE__,
DRAWSTATE_INVALID_BARRIER, "DS",
"Buffer Barrier 0x%" PRIx64 " has offset %" PRIu64
" and size %" PRIu64
" whose sum is greater than total size %" PRIu64 ".",
reinterpret_cast<const uint64_t &>(mem_barrier->buffer),
reinterpret_cast<const uint64_t &>(mem_barrier->offset),
reinterpret_cast<const uint64_t &>(mem_barrier->size),
buffer_size);
}
}
return skip_call;
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdWaitEvents(
VkCommandBuffer commandBuffer, uint32_t eventCount, const VkEvent* pEvents,
VkPipelineStageFlags sourceStageMask, VkPipelineStageFlags dstStageMask,
uint32_t memoryBarrierCount, const VkMemoryBarrier* pMemoryBarriers,
uint32_t bufferMemoryBarrierCount, const VkBufferMemoryBarrier* pBufferMemoryBarriers,
uint32_t imageMemoryBarrierCount, const VkImageMemoryBarrier* pImageMemoryBarriers)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
VkPipelineStageFlags stageMask = 0;
for (uint32_t i = 0; i < eventCount; ++i) {
pCB->waitedEvents.push_back(pEvents[i]);
pCB->events.push_back(pEvents[i]);
auto event_data = pCB->eventToStageMap.find(pEvents[i]);
if (event_data != pCB->eventToStageMap.end()) {
stageMask |= event_data->second;
} else {
auto global_event_data = dev_data->eventMap.find(pEvents[i]);
if (global_event_data == dev_data->eventMap.end()) {
skipCall |= log_msg(
dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_EVENT_EXT,
reinterpret_cast<const uint64_t &>(pEvents[i]),
__LINE__, DRAWSTATE_INVALID_FENCE, "DS",
"Fence 0x%" PRIx64
" cannot be waited on if it has never been set.",
reinterpret_cast<const uint64_t &>(pEvents[i]));
} else {
stageMask |= global_event_data->second.stageMask;
}
}
}
if (sourceStageMask != stageMask) {
skipCall |= log_msg(
dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
(VkDebugReportObjectTypeEXT)0, 0, __LINE__,
DRAWSTATE_INVALID_FENCE, "DS",
"srcStageMask in vkCmdWaitEvents must be the bitwise OR of the "
"stageMask parameters used in calls to vkCmdSetEvent and "
"VK_PIPELINE_STAGE_HOST_BIT if used with vkSetEvent.");
}
if (pCB->state == CB_RECORDING) {
skipCall |= addCmd(dev_data, pCB, CMD_WAITEVENTS, "vkCmdWaitEvents()");
} else {
skipCall |= report_error_no_cb_begin(dev_data, commandBuffer, "vkCmdWaitEvents()");
}
skipCall |= TransitionImageLayouts(commandBuffer, imageMemoryBarrierCount, pImageMemoryBarriers);
skipCall |=
ValidateBarriers(commandBuffer, memoryBarrierCount, pMemoryBarriers,
bufferMemoryBarrierCount, pBufferMemoryBarriers,
imageMemoryBarrierCount, pImageMemoryBarriers);
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdWaitEvents(commandBuffer, eventCount, pEvents, sourceStageMask, dstStageMask,
memoryBarrierCount, pMemoryBarriers,
bufferMemoryBarrierCount, pBufferMemoryBarriers,
imageMemoryBarrierCount, pImageMemoryBarriers);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdPipelineBarrier(
VkCommandBuffer commandBuffer, VkPipelineStageFlags srcStageMask,
VkPipelineStageFlags dstStageMask, VkDependencyFlags dependencyFlags,
uint32_t memoryBarrierCount, const VkMemoryBarrier* pMemoryBarriers,
uint32_t bufferMemoryBarrierCount, const VkBufferMemoryBarrier* pBufferMemoryBarriers,
uint32_t imageMemoryBarrierCount, const VkImageMemoryBarrier* pImageMemoryBarriers)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_PIPELINEBARRIER, "vkCmdPipelineBarrier()");
skipCall |= TransitionImageLayouts(commandBuffer, imageMemoryBarrierCount, pImageMemoryBarriers);
skipCall |=
ValidateBarriers(commandBuffer, memoryBarrierCount, pMemoryBarriers,
bufferMemoryBarrierCount, pBufferMemoryBarriers,
imageMemoryBarrierCount, pImageMemoryBarriers);
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdPipelineBarrier(commandBuffer, srcStageMask, dstStageMask, dependencyFlags,
memoryBarrierCount, pMemoryBarriers,
bufferMemoryBarrierCount, pBufferMemoryBarriers,
imageMemoryBarrierCount, pImageMemoryBarriers);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdBeginQuery(VkCommandBuffer commandBuffer, VkQueryPool queryPool, uint32_t slot, VkFlags flags)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
QueryObject query = {queryPool, slot};
pCB->activeQueries.insert(query);
if (!pCB->startedQueries.count(query)) {
pCB->startedQueries.insert(query);
}
skipCall |= addCmd(dev_data, pCB, CMD_BEGINQUERY, "vkCmdBeginQuery()");
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdBeginQuery(commandBuffer, queryPool, slot, flags);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdEndQuery(VkCommandBuffer commandBuffer, VkQueryPool queryPool, uint32_t slot)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
QueryObject query = {queryPool, slot};
if (!pCB->activeQueries.count(query)) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_QUERY, "DS",
"Ending a query before it was started: queryPool %" PRIu64 ", index %d", (uint64_t)(queryPool), slot);
} else {
pCB->activeQueries.erase(query);
}
pCB->queryToStateMap[query] = 1;
if (pCB->state == CB_RECORDING) {
skipCall |= addCmd(dev_data, pCB, CMD_ENDQUERY, "VkCmdEndQuery()");
} else {
skipCall |= report_error_no_cb_begin(dev_data, commandBuffer, "vkCmdEndQuery()");
}
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdEndQuery(commandBuffer, queryPool, slot);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdResetQueryPool(VkCommandBuffer commandBuffer, VkQueryPool queryPool, uint32_t firstQuery, uint32_t queryCount)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
for (uint32_t i = 0; i < queryCount; i++) {
QueryObject query = {queryPool, firstQuery + i};
pCB->waitedEventsBeforeQueryReset[query] = pCB->waitedEvents;
pCB->queryToStateMap[query] = 0;
}
if (pCB->state == CB_RECORDING) {
skipCall |= addCmd(dev_data, pCB, CMD_RESETQUERYPOOL, "VkCmdResetQueryPool()");
} else {
skipCall |= report_error_no_cb_begin(dev_data, commandBuffer, "vkCmdResetQueryPool()");
}
skipCall |= insideRenderPass(dev_data, pCB, "vkCmdQueryPool");
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdResetQueryPool(commandBuffer, queryPool, firstQuery, queryCount);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdCopyQueryPoolResults(VkCommandBuffer commandBuffer, VkQueryPool queryPool, uint32_t firstQuery,
uint32_t queryCount, VkBuffer dstBuffer, VkDeviceSize dstOffset,
VkDeviceSize stride, VkQueryResultFlags flags)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
for (uint32_t i = 0; i < queryCount; i++) {
QueryObject query = {queryPool, firstQuery + i};
if(!pCB->queryToStateMap[query]) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_QUERY, "DS",
"Requesting a copy from query to buffer with invalid query: queryPool %" PRIu64 ", index %d", (uint64_t)(queryPool), firstQuery + i);
}
}
if (pCB->state == CB_RECORDING) {
skipCall |= addCmd(dev_data, pCB, CMD_COPYQUERYPOOLRESULTS, "vkCmdCopyQueryPoolResults()");
} else {
skipCall |= report_error_no_cb_begin(dev_data, commandBuffer, "vkCmdCopyQueryPoolResults()");
}
skipCall |= insideRenderPass(dev_data, pCB, "vkCmdCopyQueryPoolResults");
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdCopyQueryPoolResults(commandBuffer, queryPool,
firstQuery, queryCount, dstBuffer, dstOffset, stride, flags);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdWriteTimestamp(VkCommandBuffer commandBuffer, VkPipelineStageFlagBits pipelineStage, VkQueryPool queryPool, uint32_t slot)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
QueryObject query = {queryPool, slot};
pCB->queryToStateMap[query] = 1;
if (pCB->state == CB_RECORDING) {
skipCall |= addCmd(dev_data, pCB, CMD_WRITETIMESTAMP, "vkCmdWriteTimestamp()");
} else {
skipCall |= report_error_no_cb_begin(dev_data, commandBuffer, "vkCmdWriteTimestamp()");
}
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdWriteTimestamp(commandBuffer, pipelineStage, queryPool, slot);
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkCreateFramebuffer(VkDevice device, const VkFramebufferCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkFramebuffer* pFramebuffer)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->CreateFramebuffer(device, pCreateInfo, pAllocator, pFramebuffer);
if (VK_SUCCESS == result) {
// Shadow create info and store in map
VkFramebufferCreateInfo* localFBCI = new VkFramebufferCreateInfo(*pCreateInfo);
if (pCreateInfo->pAttachments) {
localFBCI->pAttachments = new VkImageView[localFBCI->attachmentCount];
memcpy((void*)localFBCI->pAttachments, pCreateInfo->pAttachments, localFBCI->attachmentCount*sizeof(VkImageView));
}
loader_platform_thread_lock_mutex(&globalLock);
dev_data->frameBufferMap[*pFramebuffer] = localFBCI;
loader_platform_thread_unlock_mutex(&globalLock);
}
return result;
}
// Store the DAG.
struct DAGNode {
uint32_t pass;
std::vector<uint32_t> prev;
std::vector<uint32_t> next;
};
VkBool32 FindDependency(const int index, const int dependent, const std::vector<DAGNode>& subpass_to_node, std::unordered_set<uint32_t>& processed_nodes) {
// If we have already checked this node we have not found a dependency path so return false.
if (processed_nodes.count(index))
return VK_FALSE;
processed_nodes.insert(index);
const DAGNode& node = subpass_to_node[index];
// Look for a dependency path. If one exists return true else recurse on the previous nodes.
if (std::find(node.prev.begin(), node.prev.end(), dependent) == node.prev.end()) {
for (auto elem : node.prev) {
if (FindDependency(elem, dependent, subpass_to_node, processed_nodes))
return VK_TRUE;
}
} else {
return VK_TRUE;
}
return VK_FALSE;
}
VkBool32 CheckDependencyExists(const layer_data* my_data, VkDevice device, const int subpass, const std::vector<uint32_t>& dependent_subpasses, const std::vector<DAGNode>& subpass_to_node, VkBool32& skip_call) {
VkBool32 result = VK_TRUE;
// Loop through all subpasses that share the same attachment and make sure a dependency exists
for (uint32_t k = 0; k < dependent_subpasses.size(); ++k) {
if (subpass == dependent_subpasses[k])
continue;
const DAGNode& node = subpass_to_node[subpass];
// Check for a specified dependency between the two nodes. If one exists we are done.
auto prev_elem = std::find(node.prev.begin(), node.prev.end(), dependent_subpasses[k]);
auto next_elem = std::find(node.next.begin(), node.next.end(), dependent_subpasses[k]);
if (prev_elem == node.prev.end() && next_elem == node.next.end()) {
// If no dependency exits an implicit dependency still might. If so, warn and if not throw an error.
std::unordered_set<uint32_t> processed_nodes;
if (FindDependency(subpass, dependent_subpasses[k], subpass_to_node, processed_nodes) ||
FindDependency(dependent_subpasses[k], subpass, subpass_to_node, processed_nodes)) {
// TODO: Verify against Valid Use section of spec
skip_call |= log_msg(my_data->report_data, VK_DEBUG_REPORT_WARNING_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_RENDERPASS, "DS",
"A dependency between subpasses %d and %d must exist but only an implicit one is specified.",
subpass, dependent_subpasses[k]);
} else {
skip_call |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_RENDERPASS, "DS",
"A dependency between subpasses %d and %d must exist but one is not specified.",
subpass, dependent_subpasses[k]);
result = VK_FALSE;
}
}
}
return result;
}
VkBool32 CheckPreserved(const layer_data* my_data, VkDevice device, const VkRenderPassCreateInfo* pCreateInfo, const int index, const uint32_t attachment, const std::vector<DAGNode>& subpass_to_node, int depth, VkBool32& skip_call) {
const DAGNode& node = subpass_to_node[index];
// If this node writes to the attachment return true as next nodes need to preserve the attachment.
const VkSubpassDescription& subpass = pCreateInfo->pSubpasses[index];
for (uint32_t j = 0; j < subpass.colorAttachmentCount; ++j) {
if (attachment == subpass.pColorAttachments[j].attachment)
return VK_TRUE;
}
if (subpass.pDepthStencilAttachment &&
subpass.pDepthStencilAttachment->attachment != VK_ATTACHMENT_UNUSED) {
if (attachment == subpass.pDepthStencilAttachment->attachment)
return VK_TRUE;
}
VkBool32 result = VK_FALSE;
// Loop through previous nodes and see if any of them write to the attachment.
for (auto elem : node.prev) {
result |= CheckPreserved(my_data, device, pCreateInfo, elem, attachment, subpass_to_node, depth + 1, skip_call);
}
// If the attachment was written to by a previous node than this node needs to preserve it.
if (result && depth > 0) {
const VkSubpassDescription& subpass = pCreateInfo->pSubpasses[index];
VkBool32 has_preserved = VK_FALSE;
for (uint32_t j = 0; j < subpass.preserveAttachmentCount; ++j) {
if (subpass.pPreserveAttachments[j] == attachment) {
has_preserved = VK_TRUE;
break;
}
}
if (has_preserved == VK_FALSE) {
skip_call |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_RENDERPASS, "DS",
"Attachment %d is used by a later subpass and must be preserved in subpass %d.", attachment, index);
}
}
return result;
}
VkBool32 ValidateDependencies(const layer_data* my_data, VkDevice device, const VkRenderPassCreateInfo* pCreateInfo, const std::vector<DAGNode>& subpass_to_node) {
VkBool32 skip_call = VK_FALSE;
std::vector<std::vector<uint32_t>> output_attachment_to_subpass(pCreateInfo->attachmentCount);
std::vector<std::vector<uint32_t>> input_attachment_to_subpass(pCreateInfo->attachmentCount);
// Find for each attachment the subpasses that use them.
for (uint32_t i = 0; i < pCreateInfo->subpassCount; ++i) {
const VkSubpassDescription& subpass = pCreateInfo->pSubpasses[i];
for (uint32_t j = 0; j < subpass.inputAttachmentCount; ++j) {
input_attachment_to_subpass[subpass.pInputAttachments[j].attachment].push_back(i);
}
for (uint32_t j = 0; j < subpass.colorAttachmentCount; ++j) {
output_attachment_to_subpass[subpass.pColorAttachments[j].attachment].push_back(i);
}
if (subpass.pDepthStencilAttachment && subpass.pDepthStencilAttachment->attachment != VK_ATTACHMENT_UNUSED) {
output_attachment_to_subpass[subpass.pDepthStencilAttachment->attachment].push_back(i);
}
}
// If there is a dependency needed make sure one exists
for (uint32_t i = 0; i < pCreateInfo->subpassCount; ++i) {
const VkSubpassDescription& subpass = pCreateInfo->pSubpasses[i];
// If the attachment is an input then all subpasses that output must have a dependency relationship
for (uint32_t j = 0; j < subpass.inputAttachmentCount; ++j) {
const uint32_t& attachment = subpass.pInputAttachments[j].attachment;
CheckDependencyExists(my_data, device, i, output_attachment_to_subpass[attachment], subpass_to_node, skip_call);
}
// If the attachment is an output then all subpasses that use the attachment must have a dependency relationship
for (uint32_t j = 0; j < subpass.colorAttachmentCount; ++j) {
const uint32_t& attachment = subpass.pColorAttachments[j].attachment;
CheckDependencyExists(my_data, device, i, output_attachment_to_subpass[attachment], subpass_to_node, skip_call);
CheckDependencyExists(my_data, device, i, input_attachment_to_subpass[attachment], subpass_to_node, skip_call);
}
if (subpass.pDepthStencilAttachment && subpass.pDepthStencilAttachment->attachment != VK_ATTACHMENT_UNUSED) {
const uint32_t& attachment = subpass.pDepthStencilAttachment->attachment;
CheckDependencyExists(my_data, device, i, output_attachment_to_subpass[attachment], subpass_to_node, skip_call);
CheckDependencyExists(my_data, device, i, input_attachment_to_subpass[attachment], subpass_to_node, skip_call);
}
}
// Loop through implicit dependencies, if this pass reads make sure the attachment is preserved for all passes after it was written.
for (uint32_t i = 0; i < pCreateInfo->subpassCount; ++i) {
const VkSubpassDescription& subpass = pCreateInfo->pSubpasses[i];
for (uint32_t j = 0; j < subpass.inputAttachmentCount; ++j) {
CheckPreserved(my_data, device, pCreateInfo, i, subpass.pInputAttachments[j].attachment, subpass_to_node, 0, skip_call);
}
}
return skip_call;
}
VkBool32 ValidateLayouts(const layer_data* my_data, VkDevice device, const VkRenderPassCreateInfo* pCreateInfo) {
VkBool32 skip = VK_FALSE;
for (uint32_t i = 0; i < pCreateInfo->subpassCount; ++i) {
const VkSubpassDescription& subpass = pCreateInfo->pSubpasses[i];
for (uint32_t j = 0; j < subpass.inputAttachmentCount; ++j) {
if (subpass.pInputAttachments[j].layout != VK_IMAGE_LAYOUT_DEPTH_STENCIL_READ_ONLY_OPTIMAL &&
subpass.pInputAttachments[j].layout != VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL) {
if (subpass.pInputAttachments[j].layout == VK_IMAGE_LAYOUT_GENERAL) {
// TODO: Verify Valid Use in spec. I believe this is allowed (valid) but may not be optimal performance
skip |= log_msg(my_data->report_data, VK_DEBUG_REPORT_WARNING_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_IMAGE_LAYOUT, "DS",
"Layout for input attachment is GENERAL but should be READ_ONLY_OPTIMAL.");
} else {
skip |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_IMAGE_LAYOUT, "DS",
"Layout for input attachment is %d but can only be READ_ONLY_OPTIMAL or GENERAL.", subpass.pInputAttachments[j].attachment);
}
}
}
for (uint32_t j = 0; j < subpass.colorAttachmentCount; ++j) {
if (subpass.pColorAttachments[j].layout != VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL) {
if (subpass.pColorAttachments[j].layout == VK_IMAGE_LAYOUT_GENERAL) {
// TODO: Verify Valid Use in spec. I believe this is allowed (valid) but may not be optimal performance
skip |= log_msg(my_data->report_data, VK_DEBUG_REPORT_WARNING_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_IMAGE_LAYOUT, "DS",
"Layout for color attachment is GENERAL but should be COLOR_ATTACHMENT_OPTIMAL.");
} else {
skip |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_IMAGE_LAYOUT, "DS",
"Layout for color attachment is %d but can only be COLOR_ATTACHMENT_OPTIMAL or GENERAL.", subpass.pColorAttachments[j].attachment);
}
}
}
if ((subpass.pDepthStencilAttachment != NULL) &&
(subpass.pDepthStencilAttachment->attachment != VK_ATTACHMENT_UNUSED)) {
if (subpass.pDepthStencilAttachment->layout != VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL) {
if (subpass.pDepthStencilAttachment->layout == VK_IMAGE_LAYOUT_GENERAL) {
// TODO: Verify Valid Use in spec. I believe this is allowed (valid) but may not be optimal performance
skip |= log_msg(my_data->report_data, VK_DEBUG_REPORT_WARNING_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_IMAGE_LAYOUT, "DS",
"Layout for depth attachment is GENERAL but should be DEPTH_STENCIL_ATTACHMENT_OPTIMAL.");
} else {
skip |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_IMAGE_LAYOUT, "DS",
"Layout for depth attachment is %d but can only be DEPTH_STENCIL_ATTACHMENT_OPTIMAL or GENERAL.", subpass.pDepthStencilAttachment->attachment);
}
}
}
}
return skip;
}
VkBool32 CreatePassDAG(const layer_data* my_data, VkDevice device, const VkRenderPassCreateInfo* pCreateInfo, std::vector<DAGNode>& subpass_to_node, std::vector<bool>& has_self_dependency) {
VkBool32 skip_call = VK_FALSE;
for (uint32_t i = 0; i < pCreateInfo->subpassCount; ++i) {
DAGNode& subpass_node = subpass_to_node[i];
subpass_node.pass = i;
}
for (uint32_t i = 0; i < pCreateInfo->dependencyCount; ++i) {
const VkSubpassDependency& dependency = pCreateInfo->pDependencies[i];
if (dependency.srcSubpass > dependency.dstSubpass && dependency.srcSubpass != VK_SUBPASS_EXTERNAL && dependency.dstSubpass != VK_SUBPASS_EXTERNAL) {
skip_call |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_RENDERPASS, "DS",
"Depedency graph must be specified such that an earlier pass cannot depend on a later pass.");
} else if (dependency.srcSubpass == VK_SUBPASS_EXTERNAL && dependency.dstSubpass == VK_SUBPASS_EXTERNAL) {
skip_call |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_RENDERPASS, "DS",
"The src and dest subpasses cannot both be external.");
} else if (dependency.srcSubpass == dependency.dstSubpass) {
has_self_dependency[dependency.srcSubpass] = true;
}
if (dependency.dstSubpass != VK_SUBPASS_EXTERNAL) {
subpass_to_node[dependency.dstSubpass].prev.push_back(dependency.srcSubpass);
}
if (dependency.srcSubpass != VK_SUBPASS_EXTERNAL) {
subpass_to_node[dependency.srcSubpass].next.push_back(dependency.dstSubpass);
}
}
return skip_call;
}
// TODOSC : Add intercept of vkCreateShaderModule
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkCreateShaderModule(
VkDevice device,
const VkShaderModuleCreateInfo *pCreateInfo,
const VkAllocationCallbacks* pAllocator,
VkShaderModule *pShaderModule)
{
layer_data *my_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkBool32 skip_call = VK_FALSE;
if (!shader_is_spirv(pCreateInfo)) {
skip_call |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_DEVICE_EXT,
/* dev */ 0, __LINE__, SHADER_CHECKER_NON_SPIRV_SHADER, "SC",
"Shader is not SPIR-V");
}
if (VK_FALSE != skip_call)
return VK_ERROR_VALIDATION_FAILED_EXT;
VkResult res = my_data->device_dispatch_table->CreateShaderModule(device, pCreateInfo, pAllocator, pShaderModule);
if (res == VK_SUCCESS) {
loader_platform_thread_lock_mutex(&globalLock);
my_data->shaderModuleMap[*pShaderModule] = new shader_module(pCreateInfo);
loader_platform_thread_unlock_mutex(&globalLock);
}
return res;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkCreateRenderPass(VkDevice device, const VkRenderPassCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkRenderPass* pRenderPass)
{
VkBool32 skip_call = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
// Create DAG
std::vector<bool> has_self_dependency(pCreateInfo->subpassCount);
std::vector<DAGNode> subpass_to_node(pCreateInfo->subpassCount);
skip_call |= CreatePassDAG(dev_data, device, pCreateInfo, subpass_to_node, has_self_dependency);
// Validate using DAG
skip_call |= ValidateDependencies(dev_data, device, pCreateInfo, subpass_to_node);
skip_call |= ValidateLayouts(dev_data, device, pCreateInfo);
if (VK_FALSE != skip_call) {
return VK_ERROR_VALIDATION_FAILED_EXT;
}
loader_platform_thread_unlock_mutex(&globalLock);
VkResult result = dev_data->device_dispatch_table->CreateRenderPass(device, pCreateInfo, pAllocator, pRenderPass);
if (VK_SUCCESS == result) {
loader_platform_thread_lock_mutex(&globalLock);
// TODOSC : Merge in tracking of renderpass from ShaderChecker
// Shadow create info and store in map
VkRenderPassCreateInfo* localRPCI = new VkRenderPassCreateInfo(*pCreateInfo);
if (pCreateInfo->pAttachments) {
localRPCI->pAttachments = new VkAttachmentDescription[localRPCI->attachmentCount];
memcpy((void*)localRPCI->pAttachments, pCreateInfo->pAttachments, localRPCI->attachmentCount*sizeof(VkAttachmentDescription));
}
if (pCreateInfo->pSubpasses) {
localRPCI->pSubpasses = new VkSubpassDescription[localRPCI->subpassCount];
memcpy((void*)localRPCI->pSubpasses, pCreateInfo->pSubpasses, localRPCI->subpassCount*sizeof(VkSubpassDescription));
for (uint32_t i = 0; i < localRPCI->subpassCount; i++) {
VkSubpassDescription *subpass = (VkSubpassDescription *) &localRPCI->pSubpasses[i];
const uint32_t attachmentCount = subpass->inputAttachmentCount +
subpass->colorAttachmentCount * (1 + (subpass->pResolveAttachments?1:0)) +
((subpass->pDepthStencilAttachment) ? 1 : 0) + subpass->preserveAttachmentCount;
VkAttachmentReference *attachments = new VkAttachmentReference[attachmentCount];
memcpy(attachments, subpass->pInputAttachments,
sizeof(attachments[0]) * subpass->inputAttachmentCount);
subpass->pInputAttachments = attachments;
attachments += subpass->inputAttachmentCount;
memcpy(attachments, subpass->pColorAttachments,
sizeof(attachments[0]) * subpass->colorAttachmentCount);
subpass->pColorAttachments = attachments;
attachments += subpass->colorAttachmentCount;
if (subpass->pResolveAttachments) {
memcpy(attachments, subpass->pResolveAttachments,
sizeof(attachments[0]) * subpass->colorAttachmentCount);
subpass->pResolveAttachments = attachments;
attachments += subpass->colorAttachmentCount;
}
if (subpass->pDepthStencilAttachment) {
memcpy(attachments, subpass->pDepthStencilAttachment,
sizeof(attachments[0]) * 1);
subpass->pDepthStencilAttachment = attachments;
attachments += 1;
}
memcpy(attachments, subpass->pPreserveAttachments,
sizeof(attachments[0]) * subpass->preserveAttachmentCount);
subpass->pPreserveAttachments = &attachments->attachment;
}
}
if (pCreateInfo->pDependencies) {
localRPCI->pDependencies = new VkSubpassDependency[localRPCI->dependencyCount];
memcpy((void*)localRPCI->pDependencies, pCreateInfo->pDependencies, localRPCI->dependencyCount*sizeof(VkSubpassDependency));
}
dev_data->renderPassMap[*pRenderPass] = new RENDER_PASS_NODE(localRPCI);
dev_data->renderPassMap[*pRenderPass]->hasSelfDependency = has_self_dependency;
loader_platform_thread_unlock_mutex(&globalLock);
}
return result;
}
// Free the renderpass shadow
static void deleteRenderPasses(layer_data* my_data)
{
if (my_data->renderPassMap.size() <= 0)
return;
for (auto ii=my_data->renderPassMap.begin(); ii!=my_data->renderPassMap.end(); ++ii) {
const VkRenderPassCreateInfo* pRenderPassInfo = (*ii).second->pCreateInfo;
if (pRenderPassInfo->pAttachments) {
delete[] pRenderPassInfo->pAttachments;
}
if (pRenderPassInfo->pSubpasses) {
for (uint32_t i=0; i<pRenderPassInfo->subpassCount; ++i) {
// Attachements are all allocated in a block, so just need to
// find the first non-null one to delete
if (pRenderPassInfo->pSubpasses[i].pInputAttachments) {
delete[] pRenderPassInfo->pSubpasses[i].pInputAttachments;
} else if (pRenderPassInfo->pSubpasses[i].pColorAttachments) {
delete[] pRenderPassInfo->pSubpasses[i].pColorAttachments;
} else if (pRenderPassInfo->pSubpasses[i].pResolveAttachments) {
delete[] pRenderPassInfo->pSubpasses[i].pResolveAttachments;
} else if (pRenderPassInfo->pSubpasses[i].pPreserveAttachments) {
delete[] pRenderPassInfo->pSubpasses[i].pPreserveAttachments;
}
}
delete[] pRenderPassInfo->pSubpasses;
}
if (pRenderPassInfo->pDependencies) {
delete[] pRenderPassInfo->pDependencies;
}
delete pRenderPassInfo;
delete (*ii).second;
}
my_data->renderPassMap.clear();
}
VkBool32 VerifyFramebufferAndRenderPassLayouts(VkCommandBuffer cmdBuffer, const VkRenderPassBeginInfo* pRenderPassBegin) {
VkBool32 skip_call = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(cmdBuffer), layer_data_map);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, cmdBuffer);
const VkRenderPassCreateInfo* pRenderPassInfo = dev_data->renderPassMap[pRenderPassBegin->renderPass]->pCreateInfo;
const VkFramebufferCreateInfo* pFramebufferInfo = dev_data->frameBufferMap[pRenderPassBegin->framebuffer];
if (pRenderPassInfo->attachmentCount != pFramebufferInfo->attachmentCount) {
skip_call |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_RENDERPASS, "DS",
"You cannot start a render pass using a framebuffer with a different number of attachments.");
}
for (uint32_t i = 0; i < pRenderPassInfo->attachmentCount; ++i) {
const VkImageView& image_view = pFramebufferInfo->pAttachments[i];
auto image_data = dev_data->imageViewMap.find(image_view);
assert(image_data != dev_data->imageViewMap.end());
const VkImage& image = image_data->second->image;
const VkImageSubresourceRange& subRange = image_data->second->subresourceRange;
IMAGE_CMD_BUF_NODE newNode = {pRenderPassInfo->pAttachments[i].initialLayout, pRenderPassInfo->pAttachments[i].initialLayout};
// TODO: Do not iterate over every possibility - consolidate where possible
for (uint32_t j = 0; j < subRange.levelCount; j++) {
uint32_t level = subRange.baseMipLevel + j;
for (uint32_t k = 0; k < subRange.layerCount; k++) {
uint32_t layer = subRange.baseArrayLayer + k;
VkImageSubresource sub = {subRange.aspectMask, level, layer};
IMAGE_CMD_BUF_NODE node;
if (!FindLayout(pCB, image, sub, node)) {
SetLayout(pCB, image, sub, newNode);
continue;
}
if (newNode.layout != node.layout) {
skip_call |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_RENDERPASS, "DS",
"You cannot start a render pass using attachment %i where the intial layout differs from the starting layout.", i);
}
}
}
}
return skip_call;
}
void TransitionSubpassLayouts(VkCommandBuffer cmdBuffer, const VkRenderPassBeginInfo* pRenderPassBegin, const int subpass_index) {
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(cmdBuffer), layer_data_map);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, cmdBuffer);
auto render_pass_data = dev_data->renderPassMap.find(pRenderPassBegin->renderPass);
if (render_pass_data == dev_data->renderPassMap.end()) {
return;
}
const VkRenderPassCreateInfo* pRenderPassInfo = render_pass_data->second->pCreateInfo;
auto framebuffer_data = dev_data->frameBufferMap.find(pRenderPassBegin->framebuffer);
if (framebuffer_data == dev_data->frameBufferMap.end()) {
return;
}
const VkFramebufferCreateInfo* pFramebufferInfo = framebuffer_data->second;
const VkSubpassDescription& subpass = pRenderPassInfo->pSubpasses[subpass_index];
for (uint32_t j = 0; j < subpass.inputAttachmentCount; ++j) {
const VkImageView& image_view = pFramebufferInfo->pAttachments[subpass.pInputAttachments[j].attachment];
SetLayout(dev_data, pCB, image_view,
subpass.pInputAttachments[j].layout);
}
for (uint32_t j = 0; j < subpass.colorAttachmentCount; ++j) {
const VkImageView& image_view = pFramebufferInfo->pAttachments[subpass.pColorAttachments[j].attachment];
SetLayout(dev_data, pCB, image_view,
subpass.pColorAttachments[j].layout);
}
if ((subpass.pDepthStencilAttachment != NULL) &&
(subpass.pDepthStencilAttachment->attachment != VK_ATTACHMENT_UNUSED)) {
const VkImageView& image_view = pFramebufferInfo->pAttachments[subpass.pDepthStencilAttachment->attachment];
SetLayout(dev_data, pCB, image_view,
subpass.pDepthStencilAttachment->layout);
}
}
VkBool32 validatePrimaryCommandBuffer(const layer_data* my_data, const GLOBAL_CB_NODE* pCB, const std::string& cmd_name) {
VkBool32 skip_call = VK_FALSE;
if (pCB->createInfo.level != VK_COMMAND_BUFFER_LEVEL_PRIMARY) {
skip_call |= log_msg(my_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT)0, 0, __LINE__, DRAWSTATE_INVALID_COMMAND_BUFFER, "DS",
"Cannot execute command %s on a secondary command buffer.", cmd_name.c_str());
}
return skip_call;
}
void TransitionFinalSubpassLayouts(VkCommandBuffer cmdBuffer, const VkRenderPassBeginInfo* pRenderPassBegin) {
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(cmdBuffer), layer_data_map);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, cmdBuffer);
auto render_pass_data = dev_data->renderPassMap.find(pRenderPassBegin->renderPass);
if (render_pass_data == dev_data->renderPassMap.end()) {
return;
}
const VkRenderPassCreateInfo* pRenderPassInfo = render_pass_data->second->pCreateInfo;
auto framebuffer_data = dev_data->frameBufferMap.find(pRenderPassBegin->framebuffer);
if (framebuffer_data == dev_data->frameBufferMap.end()) {
return;
}
const VkFramebufferCreateInfo* pFramebufferInfo = framebuffer_data->second;
for (uint32_t i = 0; i < pRenderPassInfo->attachmentCount; ++i) {
const VkImageView& image_view = pFramebufferInfo->pAttachments[i];
SetLayout(dev_data, pCB, image_view,
pRenderPassInfo->pAttachments[i].finalLayout);
}
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdBeginRenderPass(VkCommandBuffer commandBuffer, const VkRenderPassBeginInfo *pRenderPassBegin, VkSubpassContents contents)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
if (pRenderPassBegin && pRenderPassBegin->renderPass) {
skipCall |= VerifyFramebufferAndRenderPassLayouts(commandBuffer, pRenderPassBegin);
skipCall |= insideRenderPass(dev_data, pCB, "vkCmdBeginRenderPass");
skipCall |= validatePrimaryCommandBuffer(dev_data, pCB, "vkCmdBeginRenderPass");
skipCall |= addCmd(dev_data, pCB, CMD_BEGINRENDERPASS, "vkCmdBeginRenderPass()");
pCB->activeRenderPass = pRenderPassBegin->renderPass;
// This is a shallow copy as that is all that is needed for now
pCB->activeRenderPassBeginInfo = *pRenderPassBegin;
pCB->activeSubpass = 0;
pCB->activeSubpassContents = contents;
pCB->framebuffer = pRenderPassBegin->framebuffer;
} else {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_INVALID_RENDERPASS, "DS",
"You cannot use a NULL RenderPass object in vkCmdBeginRenderPass()");
}
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall) {
dev_data->device_dispatch_table->CmdBeginRenderPass(commandBuffer, pRenderPassBegin, contents);
loader_platform_thread_lock_mutex(&globalLock);
// This is a shallow copy as that is all that is needed for now
dev_data->renderPassBeginInfo = *pRenderPassBegin;
dev_data->currentSubpass = 0;
loader_platform_thread_unlock_mutex(&globalLock);
}
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdNextSubpass(VkCommandBuffer commandBuffer, VkSubpassContents contents)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
TransitionSubpassLayouts(commandBuffer, &dev_data->renderPassBeginInfo, ++dev_data->currentSubpass);
if (pCB) {
skipCall |= validatePrimaryCommandBuffer(dev_data, pCB, "vkCmdNextSubpass");
skipCall |= addCmd(dev_data, pCB, CMD_NEXTSUBPASS, "vkCmdNextSubpass()");
pCB->activeSubpass++;
pCB->activeSubpassContents = contents;
TransitionSubpassLayouts(commandBuffer, &pCB->activeRenderPassBeginInfo, pCB->activeSubpass);
if (pCB->lastBoundPipeline) {
skipCall |= validatePipelineState(dev_data, pCB, VK_PIPELINE_BIND_POINT_GRAPHICS, pCB->lastBoundPipeline);
}
skipCall |= outsideRenderPass(dev_data, pCB, "vkCmdNextSubpass");
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdNextSubpass(commandBuffer, contents);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdEndRenderPass(VkCommandBuffer commandBuffer)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
TransitionFinalSubpassLayouts(commandBuffer, &dev_data->renderPassBeginInfo);
if (pCB) {
skipCall |= outsideRenderPass(dev_data, pCB, "vkCmdEndRenderpass");
skipCall |= validatePrimaryCommandBuffer(dev_data, pCB, "vkCmdEndRenderPass");
skipCall |= addCmd(dev_data, pCB, CMD_ENDRENDERPASS, "vkCmdEndRenderPass()");
TransitionFinalSubpassLayouts(commandBuffer, &pCB->activeRenderPassBeginInfo);
pCB->activeRenderPass = 0;
pCB->activeSubpass = 0;
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdEndRenderPass(commandBuffer);
}
bool logInvalidAttachmentMessage(layer_data* dev_data, VkCommandBuffer secondaryBuffer, VkRenderPass secondaryPass, VkRenderPass primaryPass, uint32_t primaryAttach, uint32_t secondaryAttach, const char* msg) {
return log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_INVALID_SECONDARY_COMMAND_BUFFER, "DS",
"vkCmdExecuteCommands() called w/ invalid Cmd Buffer %p which has a render pass %" PRIx64 " that is not compatible with the current render pass %" PRIx64 "."
"Attachment %" PRIu32 " is not compatable with %" PRIu32 ". %s",
(void*)secondaryBuffer, (uint64_t)(secondaryPass), (uint64_t)(primaryPass), primaryAttach, secondaryAttach, msg);
}
bool validateAttachmentCompatibility(layer_data* dev_data, VkCommandBuffer primaryBuffer, VkRenderPass primaryPass, uint32_t primaryAttach, VkCommandBuffer secondaryBuffer, VkRenderPass secondaryPass, uint32_t secondaryAttach, bool is_multi) {
bool skip_call = false;
auto primary_data = dev_data->renderPassMap.find(primaryPass);
auto secondary_data = dev_data->renderPassMap.find(secondaryPass);
if (primary_data->second->pCreateInfo->attachmentCount <= primaryAttach) {
primaryAttach = VK_ATTACHMENT_UNUSED;
}
if (secondary_data->second->pCreateInfo->attachmentCount <= secondaryAttach) {
secondaryAttach = VK_ATTACHMENT_UNUSED;
}
if (primaryAttach == VK_ATTACHMENT_UNUSED && secondaryAttach == VK_ATTACHMENT_UNUSED) {
return skip_call;
}
if (primaryAttach == VK_ATTACHMENT_UNUSED) {
skip_call |= logInvalidAttachmentMessage(dev_data, secondaryBuffer, secondaryPass, primaryPass, primaryAttach, secondaryAttach, "The first is unused while the second is not.");
return skip_call;
}
if (secondaryAttach == VK_ATTACHMENT_UNUSED) {
skip_call |= logInvalidAttachmentMessage(dev_data, secondaryBuffer, secondaryPass, primaryPass, primaryAttach, secondaryAttach, "The second is unused while the first is not.");
return skip_call;
}
if (primary_data->second->pCreateInfo->pAttachments[primaryAttach].format != secondary_data->second->pCreateInfo->pAttachments[secondaryAttach].format) {
skip_call |= logInvalidAttachmentMessage(dev_data, secondaryBuffer, secondaryPass, primaryPass, primaryAttach, secondaryAttach, "They have different formats.");
}
if (primary_data->second->pCreateInfo->pAttachments[primaryAttach].samples != secondary_data->second->pCreateInfo->pAttachments[secondaryAttach].samples) {
skip_call |= logInvalidAttachmentMessage(dev_data, secondaryBuffer, secondaryPass, primaryPass, primaryAttach, secondaryAttach, "They have different samples.");
}
if (is_multi && primary_data->second->pCreateInfo->pAttachments[primaryAttach].flags != secondary_data->second->pCreateInfo->pAttachments[secondaryAttach].flags) {
skip_call |= logInvalidAttachmentMessage(dev_data, secondaryBuffer, secondaryPass, primaryPass, primaryAttach, secondaryAttach, "They have different flags.");
}
return skip_call;
}
bool validateSubpassCompatibility(layer_data* dev_data, VkCommandBuffer primaryBuffer, VkRenderPass primaryPass, VkCommandBuffer secondaryBuffer, VkRenderPass secondaryPass, const int subpass, bool is_multi) {
bool skip_call = false;
auto primary_data = dev_data->renderPassMap.find(primaryPass);
auto secondary_data = dev_data->renderPassMap.find(secondaryPass);
const VkSubpassDescription& primary_desc = primary_data->second->pCreateInfo->pSubpasses[subpass];
const VkSubpassDescription& secondary_desc = secondary_data->second->pCreateInfo->pSubpasses[subpass];
uint32_t maxInputAttachmentCount = std::max(primary_desc.inputAttachmentCount, secondary_desc.inputAttachmentCount);
for (uint32_t i = 0; i < maxInputAttachmentCount; ++i) {
uint32_t primary_input_attach = VK_ATTACHMENT_UNUSED, secondary_input_attach = VK_ATTACHMENT_UNUSED;
if (i < primary_desc.inputAttachmentCount) {
primary_input_attach = primary_desc.pInputAttachments[i].attachment;
}
if (i < secondary_desc.inputAttachmentCount) {
secondary_input_attach = secondary_desc.pInputAttachments[i].attachment;
}
skip_call |= validateAttachmentCompatibility(dev_data, primaryBuffer, primaryPass, primary_input_attach, secondaryBuffer, secondaryPass, secondary_input_attach, is_multi);
}
maxInputAttachmentCount = std::max(primary_desc.colorAttachmentCount, secondary_desc.colorAttachmentCount);
for (uint32_t i = 0; i < maxInputAttachmentCount; ++i) {
uint32_t primary_color_attach = VK_ATTACHMENT_UNUSED, secondary_color_attach = VK_ATTACHMENT_UNUSED;
if (i < primary_desc.colorAttachmentCount) {
primary_color_attach = primary_desc.pColorAttachments[i].attachment;
}
if (i < secondary_desc.colorAttachmentCount) {
secondary_color_attach = secondary_desc.pColorAttachments[i].attachment;
}
skip_call |= validateAttachmentCompatibility(dev_data, primaryBuffer, primaryPass, primary_color_attach, secondaryBuffer, secondaryPass, secondary_color_attach, is_multi);
uint32_t primary_resolve_attach = VK_ATTACHMENT_UNUSED, secondary_resolve_attach = VK_ATTACHMENT_UNUSED;
if (i < primary_desc.colorAttachmentCount && primary_desc.pResolveAttachments) {
primary_resolve_attach = primary_desc.pResolveAttachments[i].attachment;
}
if (i < secondary_desc.colorAttachmentCount && secondary_desc.pResolveAttachments) {
secondary_resolve_attach = secondary_desc.pResolveAttachments[i].attachment;
}
skip_call |= validateAttachmentCompatibility(dev_data, primaryBuffer, primaryPass, primary_resolve_attach, secondaryBuffer, secondaryPass, secondary_resolve_attach, is_multi);
}
uint32_t primary_depthstencil_attach = VK_ATTACHMENT_UNUSED, secondary_depthstencil_attach = VK_ATTACHMENT_UNUSED;
if (primary_desc.pDepthStencilAttachment) {
primary_depthstencil_attach = primary_desc.pDepthStencilAttachment[0].attachment;
}
if (secondary_desc.pDepthStencilAttachment) {
secondary_depthstencil_attach = secondary_desc.pDepthStencilAttachment[0].attachment;
}
skip_call |= validateAttachmentCompatibility(dev_data, primaryBuffer, primaryPass, primary_depthstencil_attach, secondaryBuffer, secondaryPass, secondary_depthstencil_attach, is_multi);
return skip_call;
}
bool validateRenderPassCompatibility(layer_data* dev_data, VkCommandBuffer primaryBuffer, VkRenderPass primaryPass, VkCommandBuffer secondaryBuffer, VkRenderPass secondaryPass) {
bool skip_call = false;
// Early exit if renderPass objects are identical (and therefore compatible)
if (primaryPass == secondaryPass)
return skip_call;
auto primary_data = dev_data->renderPassMap.find(primaryPass);
auto secondary_data = dev_data->renderPassMap.find(secondaryPass);
if (primary_data == dev_data->renderPassMap.end() || primary_data->second == nullptr) {
skip_call |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_INVALID_SECONDARY_COMMAND_BUFFER, "DS",
"vkCmdExecuteCommands() called w/ invalid current Cmd Buffer %p which has invalid render pass %" PRIx64 ".",
(void*)primaryBuffer, (uint64_t)(primaryPass));
return skip_call;
}
if (secondary_data == dev_data->renderPassMap.end() || secondary_data->second == nullptr) {
skip_call |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_INVALID_SECONDARY_COMMAND_BUFFER, "DS",
"vkCmdExecuteCommands() called w/ invalid secondary Cmd Buffer %p which has invalid render pass %" PRIx64 ".",
(void*)secondaryBuffer, (uint64_t)(secondaryPass));
return skip_call;
}
if (primary_data->second->pCreateInfo->subpassCount != secondary_data->second->pCreateInfo->subpassCount) {
skip_call |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_INVALID_SECONDARY_COMMAND_BUFFER, "DS",
"vkCmdExecuteCommands() called w/ invalid Cmd Buffer %p which has a render pass %" PRIx64 " that is not compatible with the current render pass %" PRIx64 "."
"They have a different number of subpasses.",
(void*)secondaryBuffer, (uint64_t)(secondaryPass), (uint64_t)(primaryPass));
return skip_call;
}
bool is_multi = primary_data->second->pCreateInfo->subpassCount > 1;
for (uint32_t i = 0; i < primary_data->second->pCreateInfo->subpassCount; ++i) {
skip_call |= validateSubpassCompatibility(dev_data, primaryBuffer, primaryPass, secondaryBuffer, secondaryPass, i, is_multi);
}
return skip_call;
}
bool validateFramebuffer(layer_data* dev_data, VkCommandBuffer primaryBuffer, const GLOBAL_CB_NODE* pCB, VkCommandBuffer secondaryBuffer, const GLOBAL_CB_NODE* pSubCB) {
bool skip_call = false;
if (!pSubCB->beginInfo.pInheritanceInfo) {
return skip_call;
}
VkFramebuffer primary_fb = pCB->framebuffer;
VkFramebuffer secondary_fb = pSubCB->beginInfo.pInheritanceInfo->framebuffer;
if (secondary_fb != VK_NULL_HANDLE) {
if (primary_fb != secondary_fb) {
skip_call |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_INVALID_SECONDARY_COMMAND_BUFFER, "DS",
"vkCmdExecuteCommands() called w/ invalid Cmd Buffer %p which has a framebuffer %" PRIx64 " that is not compatible with the current framebuffer %" PRIx64 ".",
(void*)secondaryBuffer, (uint64_t)(secondary_fb), (uint64_t)(primary_fb));
}
auto fb_data = dev_data->frameBufferMap.find(secondary_fb);
if (fb_data == dev_data->frameBufferMap.end() || !fb_data->second) {
skip_call |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_INVALID_SECONDARY_COMMAND_BUFFER, "DS",
"vkCmdExecuteCommands() called w/ invalid Cmd Buffer %p which has invalid framebuffer %" PRIx64 ".",
(void*)secondaryBuffer, (uint64_t)(secondary_fb));
return skip_call;
}
skip_call |= validateRenderPassCompatibility(dev_data, secondaryBuffer, fb_data->second->renderPass, secondaryBuffer, pSubCB->beginInfo.pInheritanceInfo->renderPass);
}
return skip_call;
}
bool validateSecondaryCommandBufferState(layer_data *dev_data,
GLOBAL_CB_NODE *pCB,
GLOBAL_CB_NODE *pSubCB) {
bool skipCall = false;
unordered_set<int> activeTypes;
for (auto queryObject : pCB->activeQueries) {
auto queryPoolData = dev_data->queryPoolMap.find(queryObject.pool);
if (queryPoolData != dev_data->queryPoolMap.end()) {
if (queryPoolData->second.createInfo.queryType ==
VK_QUERY_TYPE_PIPELINE_STATISTICS &&
pSubCB->beginInfo.pInheritanceInfo) {
VkQueryPipelineStatisticFlags cmdBufStatistics =
pSubCB->beginInfo.pInheritanceInfo->pipelineStatistics;
if ((cmdBufStatistics &
queryPoolData->second.createInfo.pipelineStatistics) !=
cmdBufStatistics) {
skipCall |= log_msg(
dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
(VkDebugReportObjectTypeEXT)0, 0, __LINE__,
DRAWSTATE_INVALID_SECONDARY_COMMAND_BUFFER, "DS",
"vkCmdExecuteCommands() called w/ invalid Cmd Buffer %p "
"which has invalid active query pool %" PRIx64
". Pipeline statistics is being queried so the command "
"buffer must have all bits set on the queryPool.",
reinterpret_cast<void *>(pCB->commandBuffer),
reinterpret_cast<uint64_t>(queryPoolData->first));
}
}
activeTypes.insert(queryPoolData->second.createInfo.queryType);
}
}
for (auto queryObject : pSubCB->startedQueries) {
auto queryPoolData = dev_data->queryPoolMap.find(queryObject.pool);
if (queryPoolData != dev_data->queryPoolMap.end() &&
activeTypes.count(queryPoolData->second.createInfo.queryType)) {
skipCall |=
log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
(VkDebugReportObjectTypeEXT)0, 0, __LINE__,
DRAWSTATE_INVALID_SECONDARY_COMMAND_BUFFER, "DS",
"vkCmdExecuteCommands() called w/ invalid Cmd Buffer %p "
"which has invalid active query pool %" PRIx64
"of type %d but a query of that type has been started on "
"secondary Cmd Buffer %p.",
reinterpret_cast<void *>(pCB->commandBuffer),
reinterpret_cast<uint64_t>(queryPoolData->first),
queryPoolData->second.createInfo.queryType,
reinterpret_cast<void *>(pSubCB->commandBuffer));
}
}
return skipCall;
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdExecuteCommands(VkCommandBuffer commandBuffer, uint32_t commandBuffersCount, const VkCommandBuffer* pCommandBuffers)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (pCB) {
GLOBAL_CB_NODE* pSubCB = NULL;
for (uint32_t i=0; i<commandBuffersCount; i++) {
pSubCB = getCBNode(dev_data, pCommandBuffers[i]);
if (!pSubCB) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_INVALID_SECONDARY_COMMAND_BUFFER, "DS",
"vkCmdExecuteCommands() called w/ invalid Cmd Buffer %p in element %u of pCommandBuffers array.", (void*)pCommandBuffers[i], i);
} else if (VK_COMMAND_BUFFER_LEVEL_PRIMARY == pSubCB->createInfo.level) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, (VkDebugReportObjectTypeEXT) 0, 0, __LINE__, DRAWSTATE_INVALID_SECONDARY_COMMAND_BUFFER, "DS",
"vkCmdExecuteCommands() called w/ Primary Cmd Buffer %p in element %u of pCommandBuffers array. All cmd buffers in pCommandBuffers array must be secondary.", (void*)pCommandBuffers[i], i);
} else if (pCB->activeRenderPass) { // Secondary CB w/i RenderPass must have *CONTINUE_BIT set
if (!(pSubCB->beginInfo.flags & VK_COMMAND_BUFFER_USAGE_RENDER_PASS_CONTINUE_BIT)) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, (uint64_t)pCommandBuffers[i], __LINE__, DRAWSTATE_BEGIN_CB_INVALID_STATE, "DS",
"vkCmdExecuteCommands(): Secondary Command Buffer (%p) executed within render pass (%#" PRIxLEAST64 ") must have had vkBeginCommandBuffer() called w/ VK_COMMAND_BUFFER_USAGE_RENDER_PASS_CONTINUE_BIT set.", (void*)pCommandBuffers[i], (uint64_t)pCB->activeRenderPass);
} else {
// Make sure render pass is compatible with parent command buffer pass if has continue
skipCall |= validateRenderPassCompatibility(dev_data, commandBuffer, pCB->activeRenderPass, pCommandBuffers[i], pSubCB->beginInfo.pInheritanceInfo->renderPass);
skipCall |= validateFramebuffer(dev_data, commandBuffer, pCB, pCommandBuffers[i], pSubCB);
}
string errorString = "";
if (!verify_renderpass_compatibility(dev_data, pCB->activeRenderPass, pSubCB->beginInfo.pInheritanceInfo->renderPass, errorString)) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, (uint64_t)pCommandBuffers[i], __LINE__, DRAWSTATE_RENDERPASS_INCOMPATIBLE, "DS",
"vkCmdExecuteCommands(): Secondary Command Buffer (%p) w/ render pass (%#" PRIxLEAST64 ") is incompatible w/ primary command buffer (%p) w/ render pass (%#" PRIxLEAST64 ") due to: %s",
(void*)pCommandBuffers[i], (uint64_t)pSubCB->beginInfo.pInheritanceInfo->renderPass, (void*)commandBuffer, (uint64_t)pCB->activeRenderPass, errorString.c_str());
}
// If framebuffer for secondary CB is not NULL, then it must match FB from vkCmdBeginRenderPass()
// that this CB will be executed in AND framebuffer must have been created w/ RP compatible w/ renderpass
if (pSubCB->beginInfo.pInheritanceInfo->framebuffer) {
if (pSubCB->beginInfo.pInheritanceInfo->framebuffer != pCB->activeRenderPassBeginInfo.framebuffer) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, (uint64_t)pCommandBuffers[i], __LINE__, DRAWSTATE_FRAMEBUFFER_INCOMPATIBLE, "DS",
"vkCmdExecuteCommands(): Secondary Command Buffer (%p) references framebuffer (%#" PRIxLEAST64 ") that does not match framebuffer (%#" PRIxLEAST64 ") in active renderpass (%#" PRIxLEAST64 ").",
(void*)pCommandBuffers[i], (uint64_t)pSubCB->beginInfo.pInheritanceInfo->framebuffer, (uint64_t)pCB->activeRenderPassBeginInfo.framebuffer, (uint64_t)pCB->activeRenderPass);
}
}
}
// TODO(mlentine): Move more logic into this method
skipCall |=
validateSecondaryCommandBufferState(dev_data, pCB, pSubCB);
skipCall |= validateCommandBufferState(dev_data, pSubCB);
// Secondary cmdBuffers are considered pending execution starting w/
// being recorded
if (!(pSubCB->beginInfo.flags &
VK_COMMAND_BUFFER_USAGE_SIMULTANEOUS_USE_BIT)) {
if (dev_data->globalInFlightCmdBuffers.find(
pSubCB->commandBuffer) !=
dev_data->globalInFlightCmdBuffers.end()) {
skipCall |= log_msg(
dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT,
(uint64_t)(pCB->commandBuffer), __LINE__,
DRAWSTATE_INVALID_CB_SIMULTANEOUS_USE, "DS",
"Attempt to simultaneously execute CB %#" PRIxLEAST64
" w/o VK_COMMAND_BUFFER_USAGE_SIMULTANEOUS_USE_BIT "
"set!",
(uint64_t)(pCB->commandBuffer));
}
if (pCB->beginInfo.flags & VK_COMMAND_BUFFER_USAGE_SIMULTANEOUS_USE_BIT) {
// Warn that non-simultaneous secondary cmd buffer renders primary non-simultaneous
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_WARNING_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, (uint64_t)(pCommandBuffers[i]), __LINE__, DRAWSTATE_INVALID_CB_SIMULTANEOUS_USE, "DS",
"vkCmdExecuteCommands(): Secondary Command Buffer (%#" PRIxLEAST64 ") does not have VK_COMMAND_BUFFER_USAGE_SIMULTANEOUS_USE_BIT set and will cause primary command buffer (%#" PRIxLEAST64 ") to be treated as if it does not have VK_COMMAND_BUFFER_USAGE_SIMULTANEOUS_USE_BIT set, even though it does.",
(uint64_t)(pCommandBuffers[i]), (uint64_t)(pCB->commandBuffer));
pCB->beginInfo.flags &= ~VK_COMMAND_BUFFER_USAGE_SIMULTANEOUS_USE_BIT;
}
}
if (!pCB->activeQueries.empty() &&
!dev_data->physDevProperties.features.inheritedQueries) {
skipCall |= log_msg(
dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT,
reinterpret_cast<uint64_t>(pCommandBuffers[i]), __LINE__,
DRAWSTATE_INVALID_COMMAND_BUFFER, "DS",
"vkCmdExecuteCommands(): Secondary Command Buffer "
"(%#" PRIxLEAST64 ") cannot be submitted with a query in "
"flight and inherited queries not "
"supported on this device.",
reinterpret_cast<uint64_t>(pCommandBuffers[i]));
}
pSubCB->primaryCommandBuffer = pCB->commandBuffer;
pCB->secondaryCommandBuffers.insert(pSubCB->commandBuffer);
dev_data->globalInFlightCmdBuffers.insert(pSubCB->commandBuffer);
}
skipCall |= validatePrimaryCommandBuffer(dev_data, pCB, "vkCmdExecuteComands");
skipCall |= addCmd(dev_data, pCB, CMD_EXECUTECOMMANDS, "vkCmdExecuteComands()");
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
dev_data->device_dispatch_table->CmdExecuteCommands(commandBuffer, commandBuffersCount, pCommandBuffers);
}
VkBool32 ValidateMapImageLayouts(VkDevice device, VkDeviceMemory mem) {
VkBool32 skip_call = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
auto mem_data = dev_data->memImageMap.find(mem);
if (mem_data != dev_data->memImageMap.end()) {
std::vector<VkImageLayout> layouts;
if (FindLayouts(dev_data, mem_data->second, layouts)) {
for (auto layout : layouts) {
if (layout != VK_IMAGE_LAYOUT_PREINITIALIZED &&
layout != VK_IMAGE_LAYOUT_GENERAL) {
skip_call |= log_msg(
dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
(VkDebugReportObjectTypeEXT)0, 0, __LINE__,
DRAWSTATE_INVALID_IMAGE_LAYOUT, "DS",
"Cannot map an image with layout %s. Only "
"GENERAL or PREINITIALIZED are supported.",
string_VkImageLayout(layout));
}
}
}
}
return skip_call;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkMapMemory(
VkDevice device,
VkDeviceMemory mem,
VkDeviceSize offset,
VkDeviceSize size,
VkFlags flags,
void **ppData)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkBool32 skip_call = VK_FALSE;
loader_platform_thread_lock_mutex(&globalLock);
skip_call = ValidateMapImageLayouts(device, mem);
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skip_call) {
return dev_data->device_dispatch_table->MapMemory(device, mem, offset, size, flags, ppData);
}
return VK_ERROR_VALIDATION_FAILED_EXT;
}
VKAPI_ATTR VkResult VKAPI_CALL vkBindImageMemory(
VkDevice device,
VkImage image,
VkDeviceMemory mem,
VkDeviceSize memOffset)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->BindImageMemory(device, image, mem, memOffset);
loader_platform_thread_lock_mutex(&globalLock);
dev_data->memImageMap[mem] = image;
loader_platform_thread_unlock_mutex(&globalLock);
return result;
}
VKAPI_ATTR VkResult VKAPI_CALL vkSetEvent(VkDevice device, VkEvent event) {
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
dev_data->eventMap[event].needsSignaled = false;
dev_data->eventMap[event].stageMask = VK_PIPELINE_STAGE_HOST_BIT;
loader_platform_thread_unlock_mutex(&globalLock);
VkResult result = dev_data->device_dispatch_table->SetEvent(device, event);
return result;
}
VKAPI_ATTR VkResult VKAPI_CALL vkQueueBindSparse(
VkQueue queue,
uint32_t bindInfoCount,
const VkBindSparseInfo* pBindInfo,
VkFence fence)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(queue), layer_data_map);
VkBool32 skip_call = VK_FALSE;
loader_platform_thread_lock_mutex(&globalLock);
for (uint32_t bindIdx=0; bindIdx < bindInfoCount; ++bindIdx) {
const VkBindSparseInfo& bindInfo = pBindInfo[bindIdx];
for (uint32_t i=0; i < bindInfo.waitSemaphoreCount; ++i) {
if (dev_data->semaphoreMap[bindInfo.pWaitSemaphores[i]].signaled) {
dev_data->semaphoreMap[bindInfo.pWaitSemaphores[i]].signaled =
0;
} else {
skip_call |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, 0, __LINE__, DRAWSTATE_QUEUE_FORWARD_PROGRESS, "DS",
"Queue %#" PRIx64 " is waiting on semaphore %#" PRIx64 " that has no way to be signaled.",
(uint64_t)(queue), (uint64_t)(bindInfo.pWaitSemaphores[i]));
}
}
for (uint32_t i=0; i < bindInfo.signalSemaphoreCount; ++i) {
dev_data->semaphoreMap[bindInfo.pSignalSemaphores[i]].signaled = 1;
}
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skip_call)
return dev_data->device_dispatch_table->QueueBindSparse(queue, bindInfoCount, pBindInfo, fence);
else
return VK_ERROR_VALIDATION_FAILED_EXT;
}
VKAPI_ATTR VkResult VKAPI_CALL vkCreateSemaphore(
VkDevice device,
const VkSemaphoreCreateInfo* pCreateInfo,
const VkAllocationCallbacks* pAllocator,
VkSemaphore* pSemaphore)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->CreateSemaphore(device, pCreateInfo, pAllocator, pSemaphore);
if (result == VK_SUCCESS) {
loader_platform_thread_lock_mutex(&globalLock);
dev_data->semaphoreMap[*pSemaphore].signaled = 0;
dev_data->semaphoreMap[*pSemaphore].in_use.store(0);
loader_platform_thread_unlock_mutex(&globalLock);
}
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkCreateSwapchainKHR(
VkDevice device,
const VkSwapchainCreateInfoKHR *pCreateInfo,
const VkAllocationCallbacks *pAllocator,
VkSwapchainKHR *pSwapchain)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->CreateSwapchainKHR(device, pCreateInfo, pAllocator, pSwapchain);
if (VK_SUCCESS == result) {
SWAPCHAIN_NODE *swapchain_data = new SWAPCHAIN_NODE(pCreateInfo);
loader_platform_thread_lock_mutex(&globalLock);
dev_data->device_extensions.swapchainMap[*pSwapchain] = swapchain_data;
loader_platform_thread_unlock_mutex(&globalLock);
}
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkDestroySwapchainKHR(
VkDevice device,
VkSwapchainKHR swapchain,
const VkAllocationCallbacks *pAllocator)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
auto swapchain_data = dev_data->device_extensions.swapchainMap.find(swapchain);
if (swapchain_data != dev_data->device_extensions.swapchainMap.end()) {
if (swapchain_data->second->images.size() > 0) {
for (auto swapchain_image : swapchain_data->second->images) {
auto image_sub =
dev_data->imageSubresourceMap.find(swapchain_image);
if (image_sub != dev_data->imageSubresourceMap.end()) {
for (auto imgsubpair : image_sub->second) {
auto image_item =
dev_data->imageLayoutMap.find(imgsubpair);
if (image_item != dev_data->imageLayoutMap.end()) {
dev_data->imageLayoutMap.erase(image_item);
}
}
dev_data->imageSubresourceMap.erase(image_sub);
}
}
}
delete swapchain_data->second;
dev_data->device_extensions.swapchainMap.erase(swapchain);
}
loader_platform_thread_unlock_mutex(&globalLock);
dev_data->device_dispatch_table->DestroySwapchainKHR(device, swapchain, pAllocator);
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkGetSwapchainImagesKHR(
VkDevice device,
VkSwapchainKHR swapchain,
uint32_t* pCount,
VkImage* pSwapchainImages)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->GetSwapchainImagesKHR(device, swapchain, pCount, pSwapchainImages);
if (result == VK_SUCCESS && pSwapchainImages != NULL) {
// This should never happen and is checked by param checker.
if (!pCount) return result;
loader_platform_thread_lock_mutex(&globalLock);
for (uint32_t i = 0; i < *pCount; ++i) {
IMAGE_NODE image_node;
image_node.layout = VK_IMAGE_LAYOUT_UNDEFINED;
auto swapchain_node = dev_data->device_extensions.swapchainMap[swapchain];
image_node.format = swapchain_node->createInfo.imageFormat;
swapchain_node->images.push_back(pSwapchainImages[i]);
ImageSubresourcePair subpair = {pSwapchainImages[i], false,
VkImageSubresource()};
dev_data->imageSubresourceMap[pSwapchainImages[i]].push_back(
subpair);
dev_data->imageLayoutMap[subpair] = image_node;
}
loader_platform_thread_unlock_mutex(&globalLock);
}
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkQueuePresentKHR(VkQueue queue, const VkPresentInfoKHR* pPresentInfo)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(queue), layer_data_map);
VkBool32 skip_call = VK_FALSE;
if (pPresentInfo) {
loader_platform_thread_lock_mutex(&globalLock);
for (uint32_t i=0; i < pPresentInfo->waitSemaphoreCount; ++i) {
if (dev_data->semaphoreMap[pPresentInfo->pWaitSemaphores[i]]
.signaled) {
dev_data->semaphoreMap[pPresentInfo->pWaitSemaphores[i]]
.signaled = 0;
} else {
skip_call |= log_msg(
dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, 0, __LINE__,
DRAWSTATE_QUEUE_FORWARD_PROGRESS, "DS",
"Queue %#" PRIx64 " is waiting on semaphore %#" PRIx64
" that has no way to be signaled.",
(uint64_t)(queue),
(uint64_t)(pPresentInfo->pWaitSemaphores[i]));
}
}
for (uint32_t i = 0; i < pPresentInfo->swapchainCount; ++i) {
auto swapchain_data = dev_data->device_extensions.swapchainMap.find(pPresentInfo->pSwapchains[i]);
if (swapchain_data != dev_data->device_extensions.swapchainMap.end() && pPresentInfo->pImageIndices[i] < swapchain_data->second->images.size()) {
VkImage image = swapchain_data->second->images[pPresentInfo->pImageIndices[i]];
vector<VkImageLayout> layouts;
if (FindLayouts(dev_data, image, layouts)) {
for (auto layout : layouts) {
if (layout != VK_IMAGE_LAYOUT_PRESENT_SRC_KHR) {
skip_call |= log_msg(
dev_data->report_data,
VK_DEBUG_REPORT_ERROR_BIT_EXT,
VK_DEBUG_REPORT_OBJECT_TYPE_QUEUE_EXT,
reinterpret_cast<uint64_t &>(queue), __LINE__,
DRAWSTATE_INVALID_IMAGE_LAYOUT, "DS",
"Images passed to present must be in layout "
"PRESENT_SOURCE_KHR but is in %s",
string_VkImageLayout(layout));
}
}
}
}
}
loader_platform_thread_unlock_mutex(&globalLock);
}
if (VK_FALSE == skip_call)
return dev_data->device_dispatch_table->QueuePresentKHR(queue, pPresentInfo);
return VK_ERROR_VALIDATION_FAILED_EXT;
}
VKAPI_ATTR VkResult VKAPI_CALL vkAcquireNextImageKHR(
VkDevice device,
VkSwapchainKHR swapchain,
uint64_t timeout,
VkSemaphore semaphore,
VkFence fence,
uint32_t* pImageIndex)
{
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(device), layer_data_map);
VkResult result = dev_data->device_dispatch_table->AcquireNextImageKHR(device, swapchain, timeout, semaphore, fence, pImageIndex);
loader_platform_thread_lock_mutex(&globalLock);
// FIXME/TODO: Need to add some thing code the "fence" parameter
dev_data->semaphoreMap[semaphore].signaled = 1;
loader_platform_thread_unlock_mutex(&globalLock);
return result;
}
VK_LAYER_EXPORT VKAPI_ATTR VkResult VKAPI_CALL vkCreateDebugReportCallbackEXT(
VkInstance instance,
const VkDebugReportCallbackCreateInfoEXT* pCreateInfo,
const VkAllocationCallbacks* pAllocator,
VkDebugReportCallbackEXT* pMsgCallback)
{
layer_data* my_data = get_my_data_ptr(get_dispatch_key(instance), layer_data_map);
VkLayerInstanceDispatchTable *pTable = my_data->instance_dispatch_table;
VkResult res = pTable->CreateDebugReportCallbackEXT(instance, pCreateInfo, pAllocator, pMsgCallback);
if (VK_SUCCESS == res) {
loader_platform_thread_lock_mutex(&globalLock);
res = layer_create_msg_callback(my_data->report_data, pCreateInfo, pAllocator, pMsgCallback);
loader_platform_thread_unlock_mutex(&globalLock);
}
return res;
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkDestroyDebugReportCallbackEXT(
VkInstance instance,
VkDebugReportCallbackEXT msgCallback,
const VkAllocationCallbacks* pAllocator)
{
layer_data* my_data = get_my_data_ptr(get_dispatch_key(instance), layer_data_map);
VkLayerInstanceDispatchTable *pTable = my_data->instance_dispatch_table;
pTable->DestroyDebugReportCallbackEXT(instance, msgCallback, pAllocator);
loader_platform_thread_lock_mutex(&globalLock);
layer_destroy_msg_callback(my_data->report_data, msgCallback, pAllocator);
loader_platform_thread_unlock_mutex(&globalLock);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkDebugReportMessageEXT(
VkInstance instance,
VkDebugReportFlagsEXT flags,
VkDebugReportObjectTypeEXT objType,
uint64_t object,
size_t location,
int32_t msgCode,
const char* pLayerPrefix,
const char* pMsg)
{
layer_data *my_data = get_my_data_ptr(get_dispatch_key(instance), layer_data_map);
my_data->instance_dispatch_table->DebugReportMessageEXT(instance, flags, objType, object, location, msgCode, pLayerPrefix, pMsg);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdDbgMarkerBegin(VkCommandBuffer commandBuffer, const char* pMarker)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (!dev_data->device_extensions.debug_marker_enabled) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, (uint64_t)commandBuffer, __LINE__, DRAWSTATE_INVALID_EXTENSION, "DS",
"Attempt to use CmdDbgMarkerBegin but extension disabled!");
return;
} else if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_DBGMARKERBEGIN, "vkCmdDbgMarkerBegin()");
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
debug_marker_dispatch_table(commandBuffer)->CmdDbgMarkerBegin(commandBuffer, pMarker);
}
VK_LAYER_EXPORT VKAPI_ATTR void VKAPI_CALL vkCmdDbgMarkerEnd(VkCommandBuffer commandBuffer)
{
VkBool32 skipCall = VK_FALSE;
layer_data* dev_data = get_my_data_ptr(get_dispatch_key(commandBuffer), layer_data_map);
loader_platform_thread_lock_mutex(&globalLock);
GLOBAL_CB_NODE* pCB = getCBNode(dev_data, commandBuffer);
if (!dev_data->device_extensions.debug_marker_enabled) {
skipCall |= log_msg(dev_data->report_data, VK_DEBUG_REPORT_ERROR_BIT_EXT, VK_DEBUG_REPORT_OBJECT_TYPE_COMMAND_BUFFER_EXT, (uint64_t)commandBuffer, __LINE__, DRAWSTATE_INVALID_EXTENSION, "DS",
"Attempt to use CmdDbgMarkerEnd but extension disabled!");
return;
} else if (pCB) {
skipCall |= addCmd(dev_data, pCB, CMD_DBGMARKEREND, "vkCmdDbgMarkerEnd()");
}
loader_platform_thread_unlock_mutex(&globalLock);
if (VK_FALSE == skipCall)
debug_marker_dispatch_table(commandBuffer)->CmdDbgMarkerEnd(commandBuffer);
}
VK_LAYER_EXPORT VKAPI_ATTR PFN_vkVoidFunction VKAPI_CALL vkGetDeviceProcAddr(VkDevice dev, const char* funcName)
{
if (!strcmp(funcName, "vkGetDeviceProcAddr"))
return (PFN_vkVoidFunction) vkGetDeviceProcAddr;
if (!strcmp(funcName, "vkDestroyDevice"))
return (PFN_vkVoidFunction) vkDestroyDevice;
if (!strcmp(funcName, "vkQueueSubmit"))
return (PFN_vkVoidFunction) vkQueueSubmit;
if (!strcmp(funcName, "vkWaitForFences"))
return (PFN_vkVoidFunction) vkWaitForFences;
if (!strcmp(funcName, "vkGetFenceStatus"))
return (PFN_vkVoidFunction) vkGetFenceStatus;
if (!strcmp(funcName, "vkQueueWaitIdle"))
return (PFN_vkVoidFunction) vkQueueWaitIdle;
if (!strcmp(funcName, "vkDeviceWaitIdle"))
return (PFN_vkVoidFunction) vkDeviceWaitIdle;
if (!strcmp(funcName, "vkGetDeviceQueue"))
return (PFN_vkVoidFunction) vkGetDeviceQueue;
if (!strcmp(funcName, "vkDestroyInstance"))
return (PFN_vkVoidFunction) vkDestroyInstance;
if (!strcmp(funcName, "vkDestroyDevice"))
return (PFN_vkVoidFunction) vkDestroyDevice;
if (!strcmp(funcName, "vkDestroyFence"))
return (PFN_vkVoidFunction) vkDestroyFence;
if (!strcmp(funcName, "vkResetFences"))
return (PFN_vkVoidFunction)vkResetFences;
if (!strcmp(funcName, "vkDestroySemaphore"))
return (PFN_vkVoidFunction) vkDestroySemaphore;
if (!strcmp(funcName, "vkDestroyEvent"))
return (PFN_vkVoidFunction) vkDestroyEvent;
if (!strcmp(funcName, "vkDestroyQueryPool"))
return (PFN_vkVoidFunction) vkDestroyQueryPool;
if (!strcmp(funcName, "vkDestroyBuffer"))
return (PFN_vkVoidFunction) vkDestroyBuffer;
if (!strcmp(funcName, "vkDestroyBufferView"))
return (PFN_vkVoidFunction) vkDestroyBufferView;
if (!strcmp(funcName, "vkDestroyImage"))
return (PFN_vkVoidFunction) vkDestroyImage;
if (!strcmp(funcName, "vkDestroyImageView"))
return (PFN_vkVoidFunction) vkDestroyImageView;
if (!strcmp(funcName, "vkDestroyShaderModule"))
return (PFN_vkVoidFunction) vkDestroyShaderModule;
if (!strcmp(funcName, "vkDestroyPipeline"))
return (PFN_vkVoidFunction) vkDestroyPipeline;
if (!strcmp(funcName, "vkDestroyPipelineLayout"))
return (PFN_vkVoidFunction) vkDestroyPipelineLayout;
if (!strcmp(funcName, "vkDestroySampler"))
return (PFN_vkVoidFunction) vkDestroySampler;
if (!strcmp(funcName, "vkDestroyDescriptorSetLayout"))
return (PFN_vkVoidFunction) vkDestroyDescriptorSetLayout;
if (!strcmp(funcName, "vkDestroyDescriptorPool"))
return (PFN_vkVoidFunction) vkDestroyDescriptorPool;
if (!strcmp(funcName, "vkDestroyFramebuffer"))
return (PFN_vkVoidFunction) vkDestroyFramebuffer;
if (!strcmp(funcName, "vkDestroyRenderPass"))
return (PFN_vkVoidFunction) vkDestroyRenderPass;
if (!strcmp(funcName, "vkCreateBuffer"))
return (PFN_vkVoidFunction) vkCreateBuffer;
if (!strcmp(funcName, "vkCreateBufferView"))
return (PFN_vkVoidFunction) vkCreateBufferView;
if (!strcmp(funcName, "vkCreateImage"))
return (PFN_vkVoidFunction) vkCreateImage;
if (!strcmp(funcName, "vkCreateImageView"))
return (PFN_vkVoidFunction) vkCreateImageView;
if (!strcmp(funcName, "vkCreateFence"))
return (PFN_vkVoidFunction) vkCreateFence;
if (!strcmp(funcName, "CreatePipelineCache"))
return (PFN_vkVoidFunction) vkCreatePipelineCache;
if (!strcmp(funcName, "DestroyPipelineCache"))
return (PFN_vkVoidFunction) vkDestroyPipelineCache;
if (!strcmp(funcName, "GetPipelineCacheData"))
return (PFN_vkVoidFunction) vkGetPipelineCacheData;
if (!strcmp(funcName, "MergePipelineCaches"))
return (PFN_vkVoidFunction) vkMergePipelineCaches;
if (!strcmp(funcName, "vkCreateGraphicsPipelines"))
return (PFN_vkVoidFunction) vkCreateGraphicsPipelines;
if (!strcmp(funcName, "vkCreateComputePipelines"))
return (PFN_vkVoidFunction) vkCreateComputePipelines;
if (!strcmp(funcName, "vkCreateSampler"))
return (PFN_vkVoidFunction) vkCreateSampler;
if (!strcmp(funcName, "vkCreateDescriptorSetLayout"))
return (PFN_vkVoidFunction) vkCreateDescriptorSetLayout;
if (!strcmp(funcName, "vkCreatePipelineLayout"))
return (PFN_vkVoidFunction) vkCreatePipelineLayout;
if (!strcmp(funcName, "vkCreateDescriptorPool"))
return (PFN_vkVoidFunction) vkCreateDescriptorPool;
if (!strcmp(funcName, "vkResetDescriptorPool"))
return (PFN_vkVoidFunction) vkResetDescriptorPool;
if (!strcmp(funcName, "vkAllocateDescriptorSets"))
return (PFN_vkVoidFunction) vkAllocateDescriptorSets;
if (!strcmp(funcName, "vkFreeDescriptorSets"))
return (PFN_vkVoidFunction) vkFreeDescriptorSets;
if (!strcmp(funcName, "vkUpdateDescriptorSets"))
return (PFN_vkVoidFunction) vkUpdateDescriptorSets;
if (!strcmp(funcName, "vkCreateCommandPool"))
return (PFN_vkVoidFunction) vkCreateCommandPool;
if (!strcmp(funcName, "vkDestroyCommandPool"))
return (PFN_vkVoidFunction) vkDestroyCommandPool;
if (!strcmp(funcName, "vkResetCommandPool"))
return (PFN_vkVoidFunction) vkResetCommandPool;
if (!strcmp(funcName, "vkCreateQueryPool"))
return (PFN_vkVoidFunction)vkCreateQueryPool;
if (!strcmp(funcName, "vkAllocateCommandBuffers"))
return (PFN_vkVoidFunction) vkAllocateCommandBuffers;
if (!strcmp(funcName, "vkFreeCommandBuffers"))
return (PFN_vkVoidFunction) vkFreeCommandBuffers;
if (!strcmp(funcName, "vkBeginCommandBuffer"))
return (PFN_vkVoidFunction) vkBeginCommandBuffer;
if (!strcmp(funcName, "vkEndCommandBuffer"))
return (PFN_vkVoidFunction) vkEndCommandBuffer;
if (!strcmp(funcName, "vkResetCommandBuffer"))
return (PFN_vkVoidFunction) vkResetCommandBuffer;
if (!strcmp(funcName, "vkCmdBindPipeline"))
return (PFN_vkVoidFunction) vkCmdBindPipeline;
if (!strcmp(funcName, "vkCmdSetViewport"))
return (PFN_vkVoidFunction) vkCmdSetViewport;
if (!strcmp(funcName, "vkCmdSetScissor"))
return (PFN_vkVoidFunction) vkCmdSetScissor;
if (!strcmp(funcName, "vkCmdSetLineWidth"))
return (PFN_vkVoidFunction) vkCmdSetLineWidth;
if (!strcmp(funcName, "vkCmdSetDepthBias"))
return (PFN_vkVoidFunction) vkCmdSetDepthBias;
if (!strcmp(funcName, "vkCmdSetBlendConstants"))
return (PFN_vkVoidFunction) vkCmdSetBlendConstants;
if (!strcmp(funcName, "vkCmdSetDepthBounds"))
return (PFN_vkVoidFunction) vkCmdSetDepthBounds;
if (!strcmp(funcName, "vkCmdSetStencilCompareMask"))
return (PFN_vkVoidFunction) vkCmdSetStencilCompareMask;
if (!strcmp(funcName, "vkCmdSetStencilWriteMask"))
return (PFN_vkVoidFunction) vkCmdSetStencilWriteMask;
if (!strcmp(funcName, "vkCmdSetStencilReference"))
return (PFN_vkVoidFunction) vkCmdSetStencilReference;
if (!strcmp(funcName, "vkCmdBindDescriptorSets"))
return (PFN_vkVoidFunction) vkCmdBindDescriptorSets;
if (!strcmp(funcName, "vkCmdBindVertexBuffers"))
return (PFN_vkVoidFunction) vkCmdBindVertexBuffers;
if (!strcmp(funcName, "vkCmdBindIndexBuffer"))
return (PFN_vkVoidFunction) vkCmdBindIndexBuffer;
if (!strcmp(funcName, "vkCmdDraw"))
return (PFN_vkVoidFunction) vkCmdDraw;
if (!strcmp(funcName, "vkCmdDrawIndexed"))
return (PFN_vkVoidFunction) vkCmdDrawIndexed;
if (!strcmp(funcName, "vkCmdDrawIndirect"))
return (PFN_vkVoidFunction) vkCmdDrawIndirect;
if (!strcmp(funcName, "vkCmdDrawIndexedIndirect"))
return (PFN_vkVoidFunction) vkCmdDrawIndexedIndirect;
if (!strcmp(funcName, "vkCmdDispatch"))
return (PFN_vkVoidFunction) vkCmdDispatch;
if (!strcmp(funcName, "vkCmdDispatchIndirect"))
return (PFN_vkVoidFunction) vkCmdDispatchIndirect;
if (!strcmp(funcName, "vkCmdCopyBuffer"))
return (PFN_vkVoidFunction) vkCmdCopyBuffer;
if (!strcmp(funcName, "vkCmdCopyImage"))
return (PFN_vkVoidFunction) vkCmdCopyImage;
if (!strcmp(funcName, "vkCmdCopyBufferToImage"))
return (PFN_vkVoidFunction) vkCmdCopyBufferToImage;
if (!strcmp(funcName, "vkCmdCopyImageToBuffer"))
return (PFN_vkVoidFunction) vkCmdCopyImageToBuffer;
if (!strcmp(funcName, "vkCmdUpdateBuffer"))
return (PFN_vkVoidFunction) vkCmdUpdateBuffer;
if (!strcmp(funcName, "vkCmdFillBuffer"))
return (PFN_vkVoidFunction) vkCmdFillBuffer;
if (!strcmp(funcName, "vkCmdClearColorImage"))
return (PFN_vkVoidFunction) vkCmdClearColorImage;
if (!strcmp(funcName, "vkCmdClearDepthStencilImage"))
return (PFN_vkVoidFunction) vkCmdClearDepthStencilImage;
if (!strcmp(funcName, "vkCmdClearAttachments"))
return (PFN_vkVoidFunction) vkCmdClearAttachments;
if (!strcmp(funcName, "vkCmdResolveImage"))
return (PFN_vkVoidFunction) vkCmdResolveImage;
if (!strcmp(funcName, "vkCmdSetEvent"))
return (PFN_vkVoidFunction) vkCmdSetEvent;
if (!strcmp(funcName, "vkCmdResetEvent"))
return (PFN_vkVoidFunction) vkCmdResetEvent;
if (!strcmp(funcName, "vkCmdWaitEvents"))
return (PFN_vkVoidFunction) vkCmdWaitEvents;
if (!strcmp(funcName, "vkCmdPipelineBarrier"))
return (PFN_vkVoidFunction) vkCmdPipelineBarrier;
if (!strcmp(funcName, "vkCmdBeginQuery"))
return (PFN_vkVoidFunction) vkCmdBeginQuery;
if (!strcmp(funcName, "vkCmdEndQuery"))
return (PFN_vkVoidFunction) vkCmdEndQuery;
if (!strcmp(funcName, "vkCmdResetQueryPool"))
return (PFN_vkVoidFunction) vkCmdResetQueryPool;
if (!strcmp(funcName, "vkCmdWriteTimestamp"))
return (PFN_vkVoidFunction) vkCmdWriteTimestamp;
if (!strcmp(funcName, "vkCreateFramebuffer"))
return (PFN_vkVoidFunction) vkCreateFramebuffer;
if (!strcmp(funcName, "vkCreateShaderModule"))
return (PFN_vkVoidFunction) vkCreateShaderModule;
if (!strcmp(funcName, "vkCreateRenderPass"))
return (PFN_vkVoidFunction) vkCreateRenderPass;
if (!strcmp(funcName, "vkCmdBeginRenderPass"))
return (PFN_vkVoidFunction) vkCmdBeginRenderPass;
if (!strcmp(funcName, "vkCmdNextSubpass"))
return (PFN_vkVoidFunction) vkCmdNextSubpass;
if (!strcmp(funcName, "vkCmdEndRenderPass"))
return (PFN_vkVoidFunction) vkCmdEndRenderPass;
if (!strcmp(funcName, "vkCmdExecuteCommands"))
return (PFN_vkVoidFunction) vkCmdExecuteCommands;
if (!strcmp(funcName, "vkSetEvent"))
return (PFN_vkVoidFunction) vkSetEvent;
if (!strcmp(funcName, "vkMapMemory"))
return (PFN_vkVoidFunction) vkMapMemory;
if (!strcmp(funcName, "vkGetQueryPoolResults"))
return (PFN_vkVoidFunction) vkGetQueryPoolResults;
if (!strcmp(funcName, "vkBindImageMemory"))
return (PFN_vkVoidFunction) vkBindImageMemory;
if (!strcmp(funcName, "vkQueueBindSparse"))
return (PFN_vkVoidFunction) vkQueueBindSparse;
if (!strcmp(funcName, "vkCreateSemaphore"))
return (PFN_vkVoidFunction) vkCreateSemaphore;
if (dev == NULL)
return NULL;
layer_data *dev_data;
dev_data = get_my_data_ptr(get_dispatch_key(dev), layer_data_map);
if (dev_data->device_extensions.wsi_enabled)
{
if (!strcmp(funcName, "vkCreateSwapchainKHR"))
return (PFN_vkVoidFunction) vkCreateSwapchainKHR;
if (!strcmp(funcName, "vkDestroySwapchainKHR"))
return (PFN_vkVoidFunction) vkDestroySwapchainKHR;
if (!strcmp(funcName, "vkGetSwapchainImagesKHR"))
return (PFN_vkVoidFunction) vkGetSwapchainImagesKHR;
if (!strcmp(funcName, "vkAcquireNextImageKHR"))
return (PFN_vkVoidFunction) vkAcquireNextImageKHR;
if (!strcmp(funcName, "vkQueuePresentKHR"))
return (PFN_vkVoidFunction) vkQueuePresentKHR;
}
VkLayerDispatchTable* pTable = dev_data->device_dispatch_table;
if (dev_data->device_extensions.debug_marker_enabled)
{
if (!strcmp(funcName, "vkCmdDbgMarkerBegin"))
return (PFN_vkVoidFunction) vkCmdDbgMarkerBegin;
if (!strcmp(funcName, "vkCmdDbgMarkerEnd"))
return (PFN_vkVoidFunction) vkCmdDbgMarkerEnd;
}
{
if (pTable->GetDeviceProcAddr == NULL)
return NULL;
return pTable->GetDeviceProcAddr(dev, funcName);
}
}
VK_LAYER_EXPORT VKAPI_ATTR PFN_vkVoidFunction VKAPI_CALL vkGetInstanceProcAddr(VkInstance instance, const char* funcName)
{
if (!strcmp(funcName, "vkGetInstanceProcAddr"))
return (PFN_vkVoidFunction) vkGetInstanceProcAddr;
if (!strcmp(funcName, "vkGetDeviceProcAddr"))
return (PFN_vkVoidFunction) vkGetDeviceProcAddr;
if (!strcmp(funcName, "vkCreateInstance"))
return (PFN_vkVoidFunction) vkCreateInstance;
if (!strcmp(funcName, "vkCreateDevice"))
return (PFN_vkVoidFunction) vkCreateDevice;
if (!strcmp(funcName, "vkDestroyInstance"))
return (PFN_vkVoidFunction) vkDestroyInstance;
if (!strcmp(funcName, "vkEnumerateInstanceLayerProperties"))
return (PFN_vkVoidFunction) vkEnumerateInstanceLayerProperties;
if (!strcmp(funcName, "vkEnumerateInstanceExtensionProperties"))
return (PFN_vkVoidFunction) vkEnumerateInstanceExtensionProperties;
if (!strcmp(funcName, "vkEnumerateDeviceLayerProperties"))
return (PFN_vkVoidFunction) vkEnumerateDeviceLayerProperties;
if (!strcmp(funcName, "vkEnumerateDeviceExtensionProperties"))
return (PFN_vkVoidFunction) vkEnumerateDeviceExtensionProperties;
if (instance == NULL)
return NULL;
PFN_vkVoidFunction fptr;
layer_data* my_data;
my_data = get_my_data_ptr(get_dispatch_key(instance), layer_data_map);
fptr = debug_report_get_instance_proc_addr(my_data->report_data, funcName);
if (fptr)
return fptr;
VkLayerInstanceDispatchTable* pTable = my_data->instance_dispatch_table;
if (pTable->GetInstanceProcAddr == NULL)
return NULL;
return pTable->GetInstanceProcAddr(instance, funcName);
}